Method for correlating energy consumption and value loss of public institution assets and collaborative early warning system

Abstract

The application discloses a public institution asset energy consumption and value loss correlation method and a collaborative early warning system, wherein the method comprises the following steps: S1, standardizing classification of assets, constructing an energy consumption and value loss factor set, determining multi-dimensional correlation influence factors, and establishing a multi-dimensional coupling correlation mechanism; S2, based on the mechanism, a dynamic self-correcting coupling quantization model is constructed, a three-level topology graph of 'public institution -> building -> core asset' is taken as a basis, topological and time sequence characteristics are fused, dynamic correction factors and fixed asset accelerated depreciation policy factors are embedded, and precise quantization of value loss of single asset and multi-asset linkage is realized; S3, a double-target collaborative optimization model is constructed, and a collaborative control strategy is solved and generated; and S4, based on the strategy, collaborative early warning of energy consumption abnormality and value loss acceleration is triggered. The application solves the problems of energy consumption and value control fragmentation, large quantization error and uncoordinated optimization, has high prediction accuracy, and has remarkable energy-saving and value-protecting effects.

Description

Technical Field

[0001] This invention relates to the intersection of public institution asset management and energy-saving technology, and in particular to a method for analyzing the correlation mechanism between energy consumption and value loss of public institution assets, as well as a collaborative monitoring and early warning system. This system is used for energy consumption monitoring, value assessment, and collaborative management of core assets of public institutions, providing technical support for public institutions to achieve energy conservation and consumption reduction while preserving and increasing the value of their assets. Background Technology

[0002] Public institutions, as a significant sector of energy consumption, encompass government agencies, public institutions, and social organizations that use all or part of their funds. These institutions, including government agencies, schools, hospitals, and research institutes, operate in diverse settings. Their energy management levels not only affect their own operating costs but also play a leading role in promoting green and low-carbon development across society. Currently, driven by both the "dual-carbon" strategy and the comprehensive conservation strategy, my country's public institutions are deeply implementing the comprehensive conservation strategy, actively promoting market-based mechanisms for energy conservation and carbon reduction, achieving a "double decrease" in both total energy and resource consumption and intensity. Simultaneously, they are gradually establishing a framework for full life-cycle asset management, promoting the standardized management and revitalization of administrative and institutional state-owned assets.

[0003] In terms of energy management, while public institutions have gradually promoted the digitalization of energy consumption monitoring and the adoption of market-oriented energy-saving models such as energy cost management, their energy management often focuses on single objectives such as total energy consumption control and energy-saving renovation implementation. They primarily concentrate on the statistics and control of resource consumption data such as water, electricity, and coal, failing to establish a link between energy consumption and the value loss of assets. This makes it difficult to accurately identify the impact of high-energy-consuming behavior on asset value decline and to achieve asset value preservation and appreciation through energy consumption regulation. Furthermore, the complex energy consumption characteristics of public institutions, with numerous energy-consuming terminals, long time spans, and multiple energy-consuming departments, increase the difficulty of energy conservation management. Moreover, the advancement of digital and intelligent energy conservation faces technical bottlenecks and data barriers, further hindering the coordinated development of energy consumption management and asset value management.

[0004] In asset valuation, existing technologies primarily rely on financial depreciation methods to assess the value of public institution assets. These methods focus on static indicators such as initial acquisition cost and useful life, emphasizing the entire lifecycle management from "entry" to "exit," while neglecting the accelerating effect of dynamic factors like energy consumption on asset value depreciation. For example, long-term high-energy-consumption operation of buildings accelerates the aging of components such as walls and air conditioning systems; high fuel consumption in official vehicles speeds up the wear and tear on core components like engines; and prolonged standby energy consumption in office equipment shortens its lifespan. These energy-related losses are not included in the asset valuation system, leading to discrepancies between the valuation results and actual depreciation, and failing to provide accurate basis for asset maintenance and disposal decisions.

[0005] In existing technologies, there have been some attempts to correlate energy consumption data with equipment value, but most of them focus on the simple energy consumption-value correspondence of enterprise production equipment, only establishing a shallow relationship between total energy consumption and depreciation or cost. They have not built a multi-dimensional coupling and correlation mechanism for the multi-category and multi-level characteristics of public institution assets, nor have they considered the compound impact and offsetting effect of maintenance, renovation and environmental factors on asset value loss.

[0006] Another study on the depreciation calculation of power grid assets introduced environmental factors and time-series characteristics to predict the remaining useful life. However, this method mainly focuses on power equipment and does not cover asset types such as public institution buildings, vehicles, and office equipment. It does not construct a multi-level asset topology to reflect the linkage between assets, nor does it incorporate the policy factor of accelerated depreciation of fixed assets for adaptive dynamic correction.

[0007] Other solutions involve constructing value models around building energy consumption assets and conducting multi-asset linkage analysis. However, their optimization objectives are mostly focused on equipment maintenance plans or single energy consumption indicators. They lack a dual-objective collaborative optimization model oriented towards maximizing energy saving ratio and asset value preservation ratio, and they lack a collaborative early warning mechanism for abnormal energy consumption and accelerated value loss. It is evident that existing technologies have not formed a complete technical system applicable to all types of assets in public institutions, encompassing "correlation mechanism - quantitative model - optimization method - collaborative early warning." This fails to address the core pain points of public institution asset energy consumption and value management being disconnected, lacking clear correlation, having large quantitative errors, lacking collaborative optimization, and having single early warning systems. A complete technical solution is urgently needed to achieve deep coupling and collaborative management between the two, which is the core technical problem that this invention aims to solve. Summary of the Invention

[0008] To address the existing problems, this invention provides a method for analyzing the correlation mechanism between energy consumption and value loss of public institution assets and a collaborative monitoring and early warning system, the specific solution of which is as follows:

[0009] The method for linking energy consumption and value loss of public institution assets includes the following steps:

[0010] S1. After standardizing and classifying the assets of public institutions, data is collected, and energy consumption factor set and value loss factor set are constructed based on the collected data. Energy consumption factors, operation status factors, maintenance factors and environmental factors are identified as related influencing factors, thereby constructing a multi-dimensional coupling and correlation mechanism between the asset energy consumption and value loss of public institutions.

[0011] S2. Based on the multi-dimensional coupling and correlation mechanism, a dynamic self-correcting coupling quantitative model is constructed. This model is based on a three-level asset topology map of "public institutions → buildings → core assets", integrates topological features and temporal features, and embeds dynamic correction factors and fixed asset accelerated depreciation policy factors to quantify the value loss of single assets and multi-asset linkage.

[0012] S3. Based on the quantification results of the aforementioned dynamic self-correcting coupled quantification model, construct an energy consumption-value dual-objective collaborative optimization model with the objectives of maximizing energy saving rate and maximizing asset value preservation rate, and solve to generate a collaborative management and control strategy.

[0013] S4. Based on the aforementioned collaborative management and control strategy, trigger and execute a collaborative early warning for abnormal energy consumption and accelerated value loss.

[0014] This invention also discloses a collaborative early warning system for energy consumption and value loss of public institution assets, used to implement the method described in any of the above claims, comprising a perception layer, a network layer, and a platform layer working collaboratively, wherein:

[0015] The perception layer is used to deploy multiple types of intelligent monitoring units to collect multi-source heterogeneous data related to asset energy consumption, operation, maintenance, environment and value, and to perform preliminary verification of the collected data.

[0016] The network layer is used to deploy smart gateways to perform multi-protocol conversion, data fusion, edge computing, breakpoint resume, and encrypted transmission on data from the perception layer.

[0017] The platform layer is used to integrate data cleaning, correlation analysis, model calculation, collaborative early warning, and visual management and control functions.

[0018] The beneficial effects of this invention are as follows:

[0019] First, a multi-dimensional coupling and correlation mechanism between energy consumption and value was proposed, which solved the problem of unclear correlation. Second, a dynamic self-correcting coupling quantification model was constructed to overcome the defect of large quantification error. Third, a dual-objective collaborative optimization method was constructed to solve the pain point of non-coordinated optimization. Fourth, a three-layer collaborative monitoring and early warning system was constructed to realize the complete implementation of the technology system. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 The flowchart of the multi-dimensional coupling and correlation mechanism for energy consumption and value loss of public institution assets of the present invention is shown below.

[0022] Figure 2 This invention provides a dynamic, self-correcting energy consumption-value loss coupled quantification process.

[0023] Figure 3This is a flowchart of the intelligent energy-saving management and value collaborative optimization method for public institution assets based on energy consumption-value correlation of the present invention.

[0024] Figure 4 This is a block diagram illustrating the principle of the energy consumption and value loss collaborative monitoring and early warning system based on multi-source heterogeneous data fusion of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example 1

[0027] like Figure 1 As shown in this embodiment, a method for constructing a multi-dimensional coupling and correlation mechanism between energy consumption and value loss of core assets in public institutions is provided. This method is applied to comprehensive public institutions (covering administrative office areas and ancillary service facilities) under the jurisdiction of government affairs management departments. Addressing the technical pain points of existing technologies, such as the disconnect between energy conservation management and value management of public institution assets, the lack of clarity regarding the intrinsic relationship between energy consumption and value loss, and the single dimension of correlation analysis not covering all categories of core assets, this method overcomes the limitations of traditional single-dimensional correlation and static benchmark calibration. By combining the asset decay law under the second law of thermodynamics with the operational characteristics of public institution assets, a quantifiable, reusable, and scenario-based multi-dimensional coupling and correlation mechanism is constructed. This solves the technical problem that existing technologies cannot deeply bind asset energy consumption with physical decay and financial depreciation, providing core support for subsequent coordinated optimization of asset energy consumption control and value preservation. Specifically, it includes:

[0028] S11. Asset Classification and Quantitative Factor Library Construction:

[0029] First, by connecting with the target public institution's fixed asset management system, energy consumption monitoring platform, financial accounting system, and operation and maintenance management system, all basic data, including fixed asset ledgers, real-time energy consumption monitoring data, financial depreciation vouchers, and maintenance and renovation records, are acquired in batches. Abnormal data (such as missing values ​​and abnormal fluctuation values) are removed, and missing data is supplemented using linear interpolation. Breaking through the limitations of existing technologies that only classify single types of assets, the core assets of public institutions are standardized into three categories according to their physical attributes, energy consumption characteristics, and value loss patterns. Simultaneously, a quantifiable energy consumption factor set E and a value loss factor set V are constructed for each asset category, clarifying the quantitative collection standards and calculation methods for each factor, as detailed below:

[0030] Category 1: Building assets, including main civil engineering structure, supporting air conditioning and heating system, water supply and drainage system, and distributed photovoltaic system (covering core building types of public institutions).

[0031] (1) Energy consumption factor set Annual energy consumption of air conditioning and heating systems (quantification standard: quarterly statistics, unit kWh, accurate to building level), annual energy consumption of conventional building electricity (quantification standard: distinguishing between office area and ancillary facilities, unit kWh), and self-consumption rate of photovoltaic system (quantification standard: self-consumption of electricity / total power generation, dimensionless, accurate to 3 decimal places).

[0032] (2) Value loss factor set : Initial original value of assets (according to the accounting standards for fixed assets of administrative and public institutions, accurate to yuan), natural depreciation rate (determined according to the statutory depreciation period), physical attenuation of the main structure (quantitative standard: measured by ultrasonic testing method, unit mm, reflecting the degree of structural wear), functional depreciation rate of supporting systems (quantitative standard: (energy consumption of new system - energy consumption of current system) / energy consumption of new system, dimensionless), annual maintenance investment (according to yuan, distinguishing between routine maintenance and special maintenance), energy-saving renovation investment (according to yuan, calculated according to the details of renovation projects).

[0033] Category 2: Official vehicle assets, including fuel-powered official vehicles and new energy official vehicles (covering mainstream models of public institutions).

[0034] (1) Energy consumption factor set Vehicle fuel / electricity consumption per 100km (quantification standard: L / 100km for fuel vehicles, kWh / 100km for new energy vehicles, calculated based on monthly average), annual mileage (km, accurate to 1km), annual idling energy consumption (L for fuel vehicles, kWh for new energy vehicles, calculated based on data collected by on-board terminal).

[0035] (2) Value loss factor set Initial asset value (accurate to yuan), natural depreciation rate (determined according to the statutory depreciation period), physical wear and tear of core components (quantitative standard: wear degree of engine / motor, measured by professional testing equipment, dimensionless, value from 0 to 1), annual maintenance investment (accurate to yuan), and energy-saving modification investment (such as new energy modification and replacement of energy-saving parts, accurate to yuan).

[0036] Category 3: Office equipment and furniture assets, including office electronic equipment (computers, printers, copiers, etc.) and fixed office furniture (covering all office scenarios).

[0037] (1) Energy consumption factor set Annual standby power consumption (unit: kWh, calculated based on standby time × standby power), rated operating power consumption (unit: kWh, calculated based on operating time × rated power), annual operating time (unit: h, accurate to 1h).

[0038] (2) Value loss factor set Initial asset value (accurate to yuan), natural depreciation rate (determined according to the statutory depreciation period), physical aging loss (quantitative standard: appearance wear and performance degradation, dimensionless, value 0~1), annual maintenance investment (accurate to yuan).

[0039] This step completes the standardized classification of assets and the construction of the factor library, providing asset types and basic parameters to support the subsequent benchmarking of value loss, and realizing a logical closed loop of data-classification-factors.

[0040] S12. Determine the relevant influencing factors:

[0041] Energy consumption factors, operating status factors, maintenance factors, and environmental factors were identified as influencing factors. Operating status factors include operating time, operating temperature, and vibration frequency; environmental factors include ambient temperature and ambient humidity. These factors will be used in subsequent coupling and correlation calculations.

[0042] S13. Construct a multi-dimensional coupling and association mechanism

[0043] Based on the aforementioned asset classification, factor set, and related influencing factors, a multi-dimensional coupling mechanism for the energy consumption and value loss of public institution assets is constructed. This mechanism clarifies the intrinsic relationship between multiple factors such as energy consumption, operation, maintenance, and environment and asset value loss, laying a theoretical foundation for subsequent quantitative models.

[0044] Step S2: Construct a dynamic self-correcting coupled quantization model

[0045] S21. Based on the natural decay law of assets, an environmental correction factor is introduced to construct a dynamic benchmark model for the natural value loss of assets:

[0046] Existing technologies mostly use static depreciation formulas to determine the benchmark for asset value loss, without considering the impact of environmental factors and operation and maintenance levels on the natural depreciation of assets. This embodiment is based on the irreversible energy loss and material decay laws of the second law of thermodynamics, combined with the actual scenario of public institution asset operation, to determine the benchmark for the natural value loss of assets under standard operation and maintenance conditions without additional energy consumption impact, establish a benchmark anchor point with a correlation mechanism, and introduce an environmental correction factor to achieve dynamic adaptation of the benchmark. The dynamic benchmark model for the natural value loss of assets throughout their entire life cycle is shown below:

[0047]

[0048] Explanation of parameters in the formula: The natural baseline residual value of the asset in the t-th year of use, in yuan; The initial recorded value of the asset (in accordance with the accounting standards of Government Accounting Standard No. 3 - Fixed Assets), unit: yuan; This is the annual natural depreciation factor for assets (based on the benchmark annual depreciation rate under standard conditions with no additional energy consumption impact), strictly matching the depreciation period requirements for fixed assets of administrative and public institutions. Specific values ​​are taken from the main building's civil engineering structure. Take 2% (depreciation period of 50 years), official vehicles Take 10% (depreciation period of 10 years), office electronic equipment Take 20% (depreciation period of 5 years), office furniture Take 10% (depreciation period of 10 years); The period in which the asset has been in use, in years, with a range of values. , The statutory depreciation period for assets; This is an environmental correction factor used to adapt to the environmental differences of different public institutions, quantifying the impact of temperature, humidity, and air quality on the natural degradation of assets. The value ranges from 0.95 to 1.05, with the exception of high-temperature and high-humidity environments (such as public institutions in the southern coastal areas). Use a concentration of 0.95~0.98, in a dry environment at room temperature (such as public institutions in inland northern regions). Use 1.00~1.02 for special environments (such as high-temperature workshops and high-dust office areas). The value is set to 0.92~0.95, and the least squares method is used to calibrate the value based on the environmental monitoring data of the target public institution over the past three years.

[0049] This step, based on the asset type and initial value from step S1, completes the dynamic natural value benchmark calibration, providing a benchmark anchor for subsequent coupled correlation calculations.

[0050] S22-S23, calculate the additional attenuation coefficient and attenuation offset coefficient, and construct a multi-dimensional coupled correlation model of energy consumption and value loss.

[0051] This step is the core of the correlation mechanism, breaking through the limitations of traditional single-dimensional (only linking total energy consumption with depreciation years) technology and lacking operation and maintenance (O&M) retrofit offsetting logic. It deeply couples energy consumption factors, value loss factors, and environmental factors to construct a three-dimensional correlation model of "natural baseline loss + energy consumption-driven incremental loss - O&M retrofit offsetting loss". This achieves a deep binding between asset energy consumption and physical degradation and financial depreciation, solving the defect of existing technologies that do not include energy-saving retrofits and maintenance energy consumption in the correlation logic. The actual residual value of the asset is coupled and calculated as follows:

[0052]

[0053] The core correlation parameters in the formula are defined as follows: For the t-th year of use, the annual additional depreciation factor of the asset driven by the deviation of actual energy consumption from the benchmark energy consumption; For the t-th year of use, the annual depreciation offset coefficient caused by asset maintenance and energy-saving renovation investment.

[0054] Among them, the annual attenuation offset coefficient breaks through the limitation of existing technologies that do not quantify the benefits of operation and maintenance transformation. It incorporates the investment in energy-saving transformation and maintenance costs into the correlation logic, quantifies the mitigation effect of energy-saving operation and maintenance on asset value loss, and is calculated as follows:

[0055]

[0056] Parameter description: , To maintain the investment and the benefit weighting coefficient of energy-saving renovation investment (to meet the following requirements) Calibration example: , (emphasizing the long-term benefits of energy-saving renovations). The actual annual maintenance expenditure for the asset is expressed in yuan. The annual benchmark maintenance investment amount for similar assets, in yuan (to be uniformly verified by the administrative affairs management department); The amount of annual special energy-saving renovation investment for assets, in yuan; The energy-saving renovation benefit coefficient quantifies the sustainability of the renovation effect, with a value range of 0.8 to 1.2. The better the renovation effect, the larger the δ value. It is jointly calibrated by the energy consumption reduction rate and the value preservation rate 1-2 years after the renovation.

[0057] Among them, the annual additional depreciation coefficient of assets, that is, the additional value loss increment caused by high energy consumption behavior, is the core correlation between energy consumption factor and value loss. It breaks through the limitations of the existing technology to calculate uniformly, calculates it differently according to asset classification, and introduces the energy consumption deviation threshold to avoid abnormal energy consumption data from affecting the calculation accuracy.

[0058] (1) The annual additional depreciation coefficient for building construction assets is calculated as follows:

[0059]

[0060] Parameter description: , , The weighting coefficients for each energy consumption factor in building categories (calibrated using the Analytic Hierarchy Process (AHP) to satisfy...) Specific calibration example: , , (Adaptable to the energy consumption characteristics of most public institution buildings). The actual annual energy consumption of the air conditioning and heating system is expressed in kWh. The annual benchmark energy consumption for air conditioning and heating systems in public buildings of the same type and area (strictly in accordance with the national standard GB / T51161-2016 "Energy Consumption Quota for Public Institutions", verified according to building orientation, area, and insulation level); The actual annual energy consumption for regular building electricity use, in kWh; The benchmark annual energy consumption for regular electricity use in public buildings of the same type and area; The self-consumption rate of the photovoltaic system is dimensionless and ranges from 0 to 1. It is a threshold function to ensure that the additional attenuation coefficient is non-negative, thus avoiding the unreasonable situation where the value loss decreases when the energy consumption is lower than the benchmark.

[0061] (2) The annual additional depreciation coefficient for official vehicle assets is calculated as follows:

[0062]

[0063] Parameter description: , , The weighting coefficients for each energy consumption factor of official vehicles (satisfying) Calibration example: , , (Highlighting the core impact of fuel consumption / electricity consumption). This represents the vehicle's actual annual fuel / electricity consumption per 100 kilometers. The fuel / electricity consumption per 100 kilometers is the benchmark fuel / electricity consumption set by the Ministry of Industry and Information Technology for the same model of vehicle. The vehicle's actual annual mileage, in km; The annual benchmark mileage for official vehicles as approved by the administrative affairs management department; This refers to the vehicle's actual annual idling energy consumption. The annual idle speed benchmark energy consumption for the same model of vehicle (standard value provided by the car manufacturer).

[0064] (3) The annual depreciation coefficient for office equipment and furniture assets is calculated as follows:

[0065]

[0066] Parameter description: , , Weighting coefficients (electronic devices satisfy) Calibration example: , Office furniture ); The actual annual standby energy consumption of the equipment / furniture, in kWh; The annual standby baseline energy consumption for similar equipment / furniture; Energy consumption of electronic equipment under rated operating conditions for the year, unit: kWh; This is the annual rated operating condition benchmark energy consumption for similar electronic equipment.

[0067] The multi-dimensional coupled correlation model of "natural baseline loss + energy consumption-driven incremental loss - operation and maintenance transformation offset loss" realizes three-dimensional correlation and completes the basic quantification of asset value loss.

[0068] S24. Parameter calibration and model validation:

[0069] S24.1: Attenuation rate parameter calibration and scenario adaptation

[0070] To address the differences in the depreciation characteristics of the three major asset classes, and overcome the shortcomings of existing technologies such as coarse parameter calibration and poor scenario adaptability, the parameters of the correlation mechanism have been finely calibrated and adapted to market-oriented energy-saving scenarios. This ensures the practicality and relevance of the correlation mechanism, as detailed below:

[0071] (1) Parameter refinement calibration: Using the historical asset data, energy consumption data, asset appraisal data, and environmental monitoring data of the target public institution over the past 3-5 years, the least squares regression analysis is used, combined with the analytic hierarchy process (AHP), to complete all the weight coefficients in the previous steps ( , The precise calibration of environmental correction factor ξ and energy-saving renovation benefit coefficient δ; and the setting of reasonable thresholds for the additional attenuation coefficients of different assets to avoid the model calculation results deviating from the actual loss patterns of assets: (Building category) The upper limit is 0.5%, for official vehicles. The upper limit is 3%, office equipment category The upper limit is 5%. When the calculated value exceeds the threshold, the value is taken according to the threshold to ensure the rationality of the model.

[0072] (2) Market-oriented scenario adaptation: For mainstream market-oriented energy-saving scenarios in public institutions, such as energy-saving renovation, energy performance contracting, and energy cost management, the corresponding investment in energy-saving renovation will be adjusted accordingly. Weighting coefficient and energy-saving renovation benefit coefficient δ: In the context of energy performance contracting, Adjusted to 0.7~0.8, δ adjusted to 1.0~1.2; in the energy cost management scenario, Adjusted to 0.6~0.7, δ adjusted to 0.9~1.1; in the scenario of independent energy-saving renovation, Maintaining a value of 0.6 and adjusting δ to 0.8~1.0 adapts to the asset value change rules under different energy-saving modes, filling the gap in existing technologies that are not adapted to market-oriented energy-saving scenarios.

[0073] S24.2: Validation and Rule Consolidation

[0074] Overcoming the limitations of existing technologies that rely solely on single-dimensional verification, this approach employs a two-dimensional method: "real-world data verification + comparative verification." It utilizes real-world data from the target public institution to verify the effectiveness of the correlation mechanism, ensuring its accuracy and feasibility. The specific process is as follows:

[0075] (1) Sample selection: Select existing assets within the target public institution that have completed third-party residual value assessment, covering three major categories: buildings, official vehicles, and office equipment and furniture. Select no less than 5 samples for each category of assets to ensure the representativeness of the samples;

[0076] (2) Actual calculation: The actual residual value of each sample asset is calculated using the coupled model of this embodiment. Simultaneously record the values ​​of various parameters during the calculation process;

[0077] (3) Two-dimensional verification: ① Deviation verification: Compare the calculated value with the residual value assessed by the third party. When the calculation deviation rate is ≤5%, it is determined that the correlation mechanism is effective for this type of asset; ② Comparison verification: Compare the calculation results of this embodiment with the calculation results of existing technologies (single financial depreciation method, simple energy consumption correlation method) to verify the accuracy and superiority of the correlation mechanism of this embodiment.

[0078] (4) Rule solidification: The verified association logic, calculation rules, parameter thresholds, and scenario adaptation schemes are solidified into a standardized association rule library, generating algorithm modules that can be directly called, providing core logic support for subsequent quantitative analysis and collaborative management of asset energy consumption and value loss, and facilitating promotion and application among different public institutions.

[0079] This embodiment constructs a multi-dimensional coupling and association mechanism through the above steps, covering three major categories of core assets of public institutions, realizing a deep binding between energy consumption and value loss, and solving many defects of existing technologies.

[0080] Example 2

[0081] like Figure 2As shown in the figure, the present invention provides a dynamic self-correcting energy consumption-value loss coupled quantitative model, which is applied to large comprehensive public institutions (such as universities, centralized office areas, and tertiary hospitals). Addressing the core technical pain points of existing technologies—namely, their inability to accurately quantify the correlation between energy consumption and asset value loss, lack of consideration for topological relationships between assets, poor model adaptability, failure to incorporate policy factors, and insufficient interpretability—this invention, based on the multi-dimensional coupling mechanism constructed in Implementation Example 1, reuses core parameters such as asset classification rules, attenuation coefficients, environmental correction factors, and operation and maintenance transformation offsetting coefficients from Implementation Example 1. It innovatively designs a dynamic self-correcting coupled quantitative model, breaking through the limitations of traditional static models and single-time-series models. This model achieves dual real-time quantitative mapping of "single asset energy consumption-value loss" and "multi-asset linked energy consumption-value loss," while possessing policy adaptability and interpretability. It provides accurate, traceable, and implementable data support for energy-saving decisions and value management in public institutions. Specifically, it includes the following steps:

[0082] Step 1: Data Integration and Standardization Preprocessing

[0083] This step provides high-quality, standardized input data for the model, overcoming the shortcomings of existing technologies such as scattered data, inconsistent formats, lack of data validation, and insufficient integration of policy data. It connects with the asset classification and factor library data from Example 1, integrates multi-source heterogeneous data, and strictly follows the "Statistical Specification for Energy Consumption of Public Institutions" (GB40879-2021) for standardization. A new data linkage validation step is added to ensure data authenticity and consistency, providing a reliable data foundation for subsequent asset topology construction, model training, and factor embedding. The specific process is as follows:

[0084] Data Acquisition: By connecting to the public institution's fixed asset management system, energy consumption monitoring platform, financial accounting system, operation and maintenance management system, and environmental monitoring system, four types of core data are collected in batches. A new data collection verification mechanism has also been added to ensure full data coverage and no omissions.

[0085] (1) Static asset ledger data (referred to as Asset type, initial value The data includes the statutory depreciation period T, asset ownership (building / department), installation location, asset material, rated energy consumption parameters, etc., supplementing asset material characteristics to provide data support for subsequent loss factor adaptation.

[0086] (2) Dynamic energy consumption monitoring data (denoted as) ): Strictly in accordance with the GB40879-2021 standard, the statistics cover real-time energy consumption, daily / monthly / annual cumulative energy consumption, distinguish energy-consuming terminal types (air conditioning, lighting, office equipment, water supply and drainage systems, etc.), and collect energy consumption time period distribution data (peak / off-peak energy consumption) to meet the needs of time series feature extraction;

[0087] (3) Maintenance and modification data (denoted as) Maintenance frequency and maintenance investment Maintenance type (routine maintenance / specialized maintenance), investment in energy-saving renovation The data includes the completion time of the renovation, monitoring data of the renovation effect (comparison of energy consumption before and after the renovation), and the type of renovation technology. It also supplements the characteristics of the renovation technology type, quantifies the difference in benefits between different renovation methods, and provides a basis for the calibration of subsequent energy-saving renovation effect correction factors.

[0088] (4) Financial and policy data (denoted as ): Financial depreciation amount, depreciation method, fixed asset accelerated depreciation policy parameters (local finance department verification standards), energy consumption quota standards (classified energy consumption quotas stipulated in GB40879-2021), energy conservation subsidy policy data, and newly added policy detail data to support the embedding of policy factors.

[0089] Data preprocessing:

[0090] (1) Data linkage verification: A new multi-source data cross-verification step is added. Through the correlation verification between asset ledger and energy consumption data (such as the rationality verification of the rated energy consumption of assets and the actual energy consumption), and the correlation verification between financial data and maintenance data (such as the matching verification of maintenance investment and the degree of asset wear and tear), invalid data is eliminated to ensure data consistency. The verification formula is as follows:

[0091]

[0092] Parameter description: check is the verification result (1 indicates valid data, 0 indicates invalid data). This represents the actual energy consumption of the asset. The rated energy consumption of the asset is 0.5, which is a reasonable deviation threshold (which can be adjusted according to the type of public institution). Invalid data is supplemented by a weighted average of adjacent valid data to improve data accuracy and lay the foundation for subsequent feature extraction and model training.

[0093] (2) Handling of outliers and missing values: The 3σ criterion is used to remove outlier fluctuations in data such as energy consumption and maintenance input. For missing values ​​in time series data, linear interpolation based on time series is used to supplement them, as shown in the following formula:

[0094]

[0095] Parameter description: For missing data values, , The two sets of data represent the valid data before and after the missing data, respectively, where t is the time corresponding to the missing data. For missing values ​​in non-time series data, the mean of similar asset data is used to supplement them, ensuring the relevance of the preprocessing and avoiding data defects from affecting the model accuracy.

[0096] Standardization: Normalize all numerical data to the [0,1] interval to eliminate the influence of units, adapt to model training, and preserve the relative differences of the data. The standardization formula is as follows:

[0097]

[0098] Parameter description: The data is standardized, and x represents the original data. , These are the minimum and maximum values ​​for this type of data, respectively. For energy consumption data, additional classification and standardization are performed according to the GB40879-2021 standard to ensure that the data complies with industry norms and to provide a unified standard for subsequent multi-asset linkage energy consumption quantification.

[0099] Feature extraction and filtering:

[0100] The model input features are extracted from the preprocessed data, including static features (asset type, initial value, depreciation period, asset material) and dynamic time-series features (energy consumption time series, maintenance frequency time series, environmental factor time series, energy consumption period distribution features). A new feature screening step is added, using mutual information to screen features with a correlation ≥0.6 with value loss, eliminating redundant features, and improving model training efficiency and accuracy. The screening formula is as follows:

[0101]

[0102] Parameter description: Let p(x,y) be the mutual information value between feature X (e.g., energy consumption) and target Y (value loss), where p(x,y) is the joint probability distribution of X and Y, and p(x) and p(y) are the marginal probability distributions of X and Y, respectively. Features with a mutual information value ≥ 0.6 are included in the model input, and the selected features will be directly used for subsequent topology graph node attribute annotation and model training.

[0103] Step 2: Constructing a Multi-Level Asset Topology Graph

[0104] To address the hierarchical relationships among public institution assets, this paper overcomes the limitations of existing technologies, such as neglecting asset topological relationships, static topology construction, and coarse quantification of relationship strength. Based on the preprocessed asset data from Step 1 and combined with the three asset classification rules from Example 1, a dynamic hierarchical asset topology graph is constructed. This provides accurate topological feature input for the improved GNN-LSTM model, while reserving a dynamic topology update interface to adapt to asset change scenarios. This lays the foundation for subsequent quantification of energy consumption and value loss in multi-asset linkages, as detailed below:

[0105] (1) Definition of topology nodes: Public institutions (node ​​A), buildings (node ​​B), and core assets (node ​​C, including building ancillary equipment, official vehicles, and office equipment and furniture) are defined as third-level nodes in the topology graph, denoted as node set:

[0106] Where n is the number of buildings and m is the number of assets in a single building; combined with the features selected in step one, new node attribute annotations are added to annotate each node with static features (such as asset type and initial value) and dynamic features (such as real-time energy consumption) to improve the richness of topology features and ensure that the topology map can accurately reflect the core attributes of assets.

[0107] (2) Definition of topological edges: Construct an edge set based on asset ownership, energy use linkage, and maintenance linkage. This approach breaks through the limitations of existing technologies that only consider single associations. The weight of an edge represents the strength of that association. A multi-dimensional weighted calculation is used, combining the energy consumption linkage data and maintenance linkage data collected in step one to quantify the association strength. The formula is as follows:

[0108]

[0109] Parameter description: represents the edge weight between node i and node j, with a value range of [0,1]. The larger the weight, the stronger the association. , , The weighting coefficients (calibrated using the Analytic Hierarchy Process (AHP) and satisfying the following conditions) Calibration example: , , ), which correspond to the weights of association frequency, energy use linkage coefficient, and maintenance linkage frequency, respectively; association frequency is the number of times the assets corresponding to the two nodes are used in coordination (such as the number of times air conditioning and lighting are turned on at the same time in the same building), and the data comes from the dynamic energy consumption monitoring data collected in step one. The energy consumption correlation coefficient between the assets corresponding to the two nodes (calculated using the Pearson correlation coefficient, with a value range of [-1, 1]) is used in the calculation after taking the absolute value to quantify the degree of energy consumption linkage. The data comes from the dynamic energy consumption monitoring data in step one. The maintenance linkage frequency is the number of times the assets corresponding to the two nodes are maintained simultaneously (such as the number of times multiple air conditioners in the same building are maintained simultaneously). The data comes from the maintenance and renovation data collected in step one. The comprehensive correlation value is the weighted sum of the correlation frequency, the absolute value of the energy consumption linkage coefficient, and the maintenance linkage frequency for each node pair. This is the maximum value of the overall association value for all node pairs, used to normalize the weights and ensure that the weight values ​​are taken in a standardized manner.

[0110] (3) Mathematical Representation and Dynamic Initialization of Topology Graph: The hierarchical asset topology graph is denoted as... ,in Let be the edge weight matrix, with matrix elements as follows: This is used for topological feature extraction in subsequent GNN models; a new dynamic initialization step for the topological graph is added, which automatically initializes the node association relationship and edge weights using the asset linkage data of the target public institution for the past year (from the multi-source data collected in step one), and sets the topological update trigger threshold (such as asset association frequency change ≥30%), laying the foundation for the dynamic self-correction mechanism in subsequent step S25, and ensuring that the topological graph can dynamically adapt to asset changes.

[0111] Step 3: Hybrid Model Structure Design

[0112] This step, based on the feature data selected in step one and the dynamic hierarchical asset topology map constructed in step two, designs an improved GNN-LSTM hybrid model. Simultaneously, the structure of the GNN and LSTM layers is improved to overcome the shortcomings of traditional models, such as their inability to balance topological correlation and temporal accuracy, gradient vanishing, and insufficient interpretability. The model structure consists of three layers, providing core algorithmic support for subsequent factor embedding and value loss prediction, as detailed below:

[0113] First layer: GNN topological feature extraction layer

[0114] Based on the topology graph G constructed in step two, an improved GCN (Graph Convolutional Neural Network) is used to extract the topological association features of each asset node. Residual connections and multi-scale feature fusion are introduced to solve the problems of gradient vanishing and insufficient feature extraction in traditional GCN. The node feature update formula is as follows:

[0115]

[0116] Parameter description: , These are the feature vectors of node i at layer l and layer l+1, respectively, where the initial feature vectors are derived from the asset features filtered in step one. The normalized adjacency matrix ( (where A is the adjacency matrix of the topological graph and D is the degree matrix), and normalization processing avoids features being dominated by highly correlated nodes; , These are the weight matrix and bias term of the l-th layer, respectively, which are adaptively updated through model training; The activation function is ReLU. ), alleviate gradient vanishing; This is a residual term used to mitigate gradient vanishing and improve model training stability. The multi-scale fusion coefficient (value range 0.1~0.3). The average value of the l-th layer features of the neighboring nodes of node i is used to achieve multi-scale topological feature fusion and improve the comprehensiveness of feature extraction.

[0117] This layer outputs the topological association feature vectors of each asset node. It captures the cumulative impact of multi-asset energy use and synchronous maintenance on value loss, which is different from the limitation of traditional GCN that can only capture a single correlation, and provides topological feature support for feature fusion in the subsequent attention mechanism layer.

[0118] Second layer: Adaptive attention mechanism layer

[0119] An adaptive attention mechanism is introduced to overcome the limitation of fixed weights in traditional attention mechanisms. The topological correlation features output from the first layer in step three are fused with the temporal features extracted in step one. Weights are dynamically assigned based on the correlation between features and value loss, quantifying the impact of different features on value loss. Simultaneously, the importance ranking of features is output, enhancing model interpretability. The attention weights are calculated as follows:

[0120]

[0121] Parameter description: Let be the attention weight for the i-th feature, satisfying (N is the total number of features), the larger the weight, the greater the impact of the feature on value loss; The time-series feature vector (energy consumption, maintenance frequency, etc.) at time t is derived from the dynamic time-series features extracted in step one; For the attention scoring function (using dot product scoring), Quantify the correlation between features and the target; The adaptive adjustment coefficient is determined by the mutual information value between the feature and the value loss. The mutual information value is calculated using the mutual information method in step one. The higher the mutual information value, the larger the adjustment coefficient, thus achieving dynamic weight allocation.

[0122] Through attention weight It can output interpretable results such as "which assets' energy consumption, energy usage during which periods, and which maintenance behaviors have the greatest impact on value loss", and generate feature importance reports. It is adapted to the management needs of public institutions' financial audits and asset traceability. It is different from the shortcomings of existing models that cannot quantify the impact of features, and provides support for subsequent value loss prediction and interpretability verification.

[0123] Third layer: Dynamically gated LSTM timing prediction layer

[0124] The fusion features weighted by the attention mechanism are input into an improved LSTM, and an adaptive forgetting gate is introduced. The forgetting weights are dynamically adjusted according to the trend changes of the time series data to capture the long-term and short-term temporal dependencies between energy consumption and value loss. At the same time, a time series trend correction term is added to achieve real-time and accurate prediction of the value loss rate. The core update formula of the LSTM is as follows:

[0125]

[0126]

[0127]

[0128]

[0129]

[0130] Parameter description: The fused feature vector at time t (attention-weighted topology + temporal features) is derived from the output of the second layer in step three. , , These are the outputs of the input gate, forget gate, and output gate, respectively. The adaptive forgetting adjustment term (an improvement) is determined by the rate of change of the trend in the time series data. The greater the trend change, The larger the value, the greater the weight of the forget gate, thus avoiding the impact of redundant historical data on prediction accuracy; , These represent the cell states at time t and time t-1, respectively. , , These are the hidden states (i.e., the intermediate predicted values ​​of the value loss rate) at time t, time t-1, and time t-2, respectively, which provide the basis for the dynamic correction in the subsequent step four. This is the time series trend correction coefficient (value range: 0.05~0.15). The change in the loss rate in the previous two time periods is used to correct the current prediction value and improve the stability of time series prediction. , , Let be the weight matrices for each gate. , , , For bias terms; For element-wise multiplication, It is the hyperbolic tangent activation function.

[0131] This layer overcomes the limitation of fixed weights in the forget gate of traditional LSTM, improving the accuracy of time series prediction through adaptive adjustment and trend correction. It is particularly suitable for scenarios with large fluctuations in energy consumption in public institutions, and the output intermediate prediction values... This will be directly used for dynamic adjustments and policy factor embedding in step four.

[0132] Step 4: Dynamic correction factor calibration and adaptation

[0133] Based on the intermediate predicted value loss rate output in step three, this method overcomes the shortcomings of existing technical models, such as static nature, lack of policy factors, and coarse calibration of correction factors. It introduces dynamic correction factors and accelerated depreciation policy factors for fixed assets to achieve dynamic model adaptation and policy compliance. Simultaneously, it supplements the model with refined factor calibration methods to quantify the impact of each factor on value loss, ultimately achieving accurate prediction of value loss linked to single and multiple assets. Specifically:

[0134] (1) Embedding of dynamic correction factors: set of dynamic correction factors These correspond to asset service life, maintenance frequency, energy-saving renovation effect, and environmental factors, respectively. Each factor is calibrated using a refined calibration method, combined with multi-source data collected in step one. The quantification formula and embedding logic are as follows. Additionally, a factor linkage adjustment mechanism is added to avoid the influence of single factor bias on the prediction results.

[0135] Years of use correction factor : t represents the years of use, T represents the statutory depreciation period, and the data comes from the static asset ledger data in step one; The factor is the influence of the age of the asset (determined by the asset type: buildings μ=0.8, official vehicles μ=1.2, office equipment μ=1.5), which is dynamically updated with the age of the asset. The longer the age, the greater the adjustment weight, which conforms to the wear and tear patterns of different assets.

[0136] Maintenance frequency correction factor : f is the actual maintenance frequency. The baseline maintenance frequency (determined by the administrative affairs management department) is based on the maintenance and modification data from step one. The maintenance effectiveness coefficient (adjusted according to maintenance type, ν=1.1 for special maintenance and ν=0.9 for routine maintenance) is determined by the frequency and effectiveness of maintenance, resulting in a larger correction factor and mitigating the bias in value loss prediction.

[0137] Energy-saving renovation effect correction factor : , This represents the reduction in energy consumption after the renovation. The energy consumption before the renovation is based on the maintenance and renovation data from step one. The technical coefficient for renovation (adjusted according to the type of renovation technology: ζ=1.2 for photovoltaic renovation, ζ=1.0 for air conditioning renovation) indicates the better the renovation effect and the more advanced the renovation technology. The smaller the value, the more it reflects the mitigating effect of energy-saving renovations on value loss;

[0138] Environmental Correction Factors : Following the example of Example 1 The calibration was performed using the least squares method based on the environmental monitoring data (temperature, humidity, and air quality) of the target public institution over the past three years. The data was derived from the environmental monitoring data in step one and was adapted to the impact of different environments on asset depreciation. The calibration formula is as follows:

[0139]

[0140] Parameter description: The annual average temperature The standard temperature is 25℃. The annual average humidity The standard humidity (50%) is used to ensure the accuracy of environmental factor calibration.

[0141] (2) Embedding of policy factors for accelerated depreciation of fixed assets: introducing accelerated depreciation coefficients Combining the financial and policy data collected in Step 1, the impact of policies on value loss is quantified. The coefficient parameters are dynamically adjusted based on the policy adaptation scenarios of different public institutions, as shown in the following formula:

[0142]

[0143] Parameter description: k is the policy coefficient (determined by the local finance department, with a value range of 0.1 to 0.3). The greater the investment in energy-saving renovation, the greater the accelerated depreciation coefficient, in line with the policy requirements for accelerated depreciation of fixed assets. The amount of annual special energy-saving renovation investment for assets is in yuan. The data is from the maintenance and renovation data in step one. The initial value of the asset is expressed in yuan. The data is derived from the static asset ledger data in step one. The policy adaptation coefficient (innovation point) is adjusted according to the type of public institution and the type of energy conservation policy. The coefficients are τ=1.0 for government agencies, τ=1.1 for universities, and τ=1.05 for hospitals, to ensure that the policy factors are adapted to different scenarios.

[0144] (3) Value loss quantification formula: Combining the coupled correlation model in Example 1, the value loss of single assets and multiple assets is output, and the formula is as follows:

[0145]

[0146]

[0147] Parameter description: The value loss of a single asset at time t, in yuan; The annual natural decay coefficient of assets (using the calibration value of Example 1) is used to achieve deep integration with the correlation mechanism; M represents the total value loss of multiple assets linked under the same topology at time t, in yuan; M is the number of linked assets. The edge weights of asset i and asset j (derived from step two) are used to quantify the cumulative impact of multi-asset linkage on total loss, overcoming the limitation of existing technologies that can only predict the loss of a single asset.

[0148] Step 5: Model Training and Dynamic Self-Correction

[0149] Based on the preprocessed feature data from step one, the topology graph constructed in step two, the model structure designed in step three, and the embedded correction and policy factors in step four, the model achieves accurate training and dynamic self-correction. This overcomes the shortcomings of existing technologies, such as easy overfitting during training, a single self-correction mechanism, and no accuracy feedback. It ensures that the model is adaptable to the characteristics of diversified public institution assets, complex energy use scenarios, and dynamic asset changes, thus guaranteeing the model's accuracy and adaptability. The specific process is as follows:

[0150] (1) Construction of training dataset

[0151] Data source: Full data from 36 consecutive months of data from 3 typical public institutions (provincial government office area, tertiary general hospital, and provincial undergraduate university) were selected, covering five categories: static asset ledger data, dynamic energy consumption monitoring data, maintenance and renovation data, financial and policy data, and environmental monitoring data, which are completely consistent with the data collection specifications in step one.

[0152] Sample size and annotation: The total sample size of the dataset is 12,800, including 8,600 single-asset samples and 4,200 multi-asset linked samples. The dataset was annotated using the monthly residual value assessment results of assets issued by third-party asset appraisal agencies and the monthly depreciation data of the financial system as real labels.

[0153] Dataset partitioning: The dataset was divided into a training set (8960 samples), a validation set (2560 samples), and a test set (1280 samples) in a ratio of 7:2:1. Stratified sampling was used in the partitioning process to ensure that the sample proportions of the three types of public institutions and the three types of assets were consistent in the training set, validation set, and test set.

[0154] Data augmentation: For time-series data, translation augmentation and Gaussian noise addition were used to expand the training samples, increasing the training set size to 15,000 samples. The data augmentation formula is as follows:

[0155]

[0156] Parameter description: For the enhanced data, The noise figure is fixed at 0.03. The data are standard normally distributed random numbers, and the data augmentation is only applied to the training set to avoid data leakage in the validation and test sets.

[0157] (2) Model training hyperparameter settings

[0158] This model is built using the PyTorch 2.0 framework, with a training hardware environment of NVIDIA RTX 3090 GPU and 32GB of RAM. The training hyperparameters are fixed as follows, and those skilled in the art can directly reproduce it:

[0159] (3) Method for calibrating weighting coefficients and correction factors

[0160] All weight coefficients in this model (including topological edge weight coefficients) , , Correlation influence factor weights The weights in the AHP method are all calibrated using the analytic hierarchy process (AHP) combined with the least squares method. The specific calibration steps are as follows:

[0161] Constructing the Judgment Matrix: Five senior experts in public institution asset management, energy conservation control, and asset appraisal were invited to score the importance of each influencing factor pairwise. A judgment matrix was constructed using the 1-9 scale. Taking the topological edge weight coefficient as an example, the judgment matrix is ​​shown below:

[0162]

[0163] Consistency check: Calculate the largest eigenvalue of the judgment matrix and the consistency index CI. When the consistency ratio CR < 0.1, the judgment matrix is ​​considered to have passed the consistency check; otherwise, the judgment matrix is ​​readjusted.

[0164] Weight calculation: The initial weight coefficients are obtained by normalizing the eigenvector corresponding to the largest eigenvalue.

[0165] Least squares correction: Based on the training set data, the initial weights are corrected using the least squares method to minimize the mean square error between the model's predicted values ​​and the true labels, and finally obtain the calibrated weight coefficients.

[0166] Fixed calibration result: The final fixed weights are: topological edge weights , , Weights of four types of related influencing factors (Energy consumption) (run), (maintain), (Environment) The calibration results are applicable to most public institution scenarios and can be fine-tuned according to the actual scenario.

[0167] The parameter calibration of dynamic adjustment factors and policy factors shall strictly follow the formulas and value ranges in step four, and the adaptability calibration shall be completed according to the asset type and policy requirements of the target public institution.

[0168] (4) Loss function and model convergence criterion

[0169] Loss function design: The mean squared error (MSE) combined with L1 regularization is used as the loss function to balance prediction accuracy and model simplicity, and to avoid overfitting. The loss function formula is as follows:

[0170]

[0171] Parameter description: L is the loss value, M is the number of training samples. Let t be the actual value loss (true label) of the i-th sample at time t. These are the model's predicted values. is the regularization coefficient, with a fixed value of 0.005, and W is the set of all weight parameters in the model.

[0172] Model convergence criteria: A model is considered convergent and training is stopped if any of the following conditions are met:

[0173] Condition 1: The number of iterations reaches 100 rounds;

[0174] Condition 2: The validation set loss does not decrease for 5 consecutive rounds;

[0175] Condition 3: The difference between the training set loss and the validation set loss is ≤0.5%, and the prediction bias rate of the test set is ≤5%.

[0176] (5) Dynamic self-correction mechanism

[0177] Four types of self-correction trigger conditions are set. Combining the data update in step one, the topology change in step two, the policy adjustment in step four, and the model prediction accuracy, when any condition is met, the model automatically starts the hierarchical self-correction process, updates parameters, and performs accuracy verification to achieve dynamic adaptation, as detailed below:

[0178] Condition 1: Prediction bias exceeds the limit: When the prediction bias rate of the test set is >5%, the weight update is triggered. The weight matrix and bias term of GNN and LSTM are adjusted by gradient descent. At the same time, the attention weight coefficients are recalibrated. After correction, the bias rate is recalculated until the bias rate is ≤5% to ensure the prediction accuracy of the model.

[0179] Condition 2: Asset topology change: When assets are added / deleted, asset ownership is adjusted (such as equipment transfer), or the asset association strength changes by ≥30%, the topology graph G constructed in step 2 is automatically updated, the GNN layer is retrained, the topology feature extraction parameters are updated, and the feature weights of the attention mechanism are adjusted to adapt to the new topology structure.

[0180] Condition 3: Data update frequency: New energy consumption, maintenance, and policy data (corresponding to the multi-source data in step one) are automatically imported every quarter, triggering dynamic correction factor recalibration, updating model input features and correction weights, and fine-tuning LSTM layer parameters to ensure that the model adapts to data changes;

[0181] Condition 4: Policy Adjustment: When the policy on accelerated depreciation of fixed assets or the standard for energy consumption quotas are adjusted, the policy coefficient k and the policy adaptation coefficient τ in step four are automatically updated, the policy factors are re-embedded, and the parameters of the prediction formula are adjusted to ensure that the model meets the policy requirements.

[0182] Self-correction accuracy feedback rules: After each self-correction, the accuracy is verified using a test set. If the accuracy is improved by ≥3% after correction, the updated parameters are fixed; if the accuracy is not improved, the parameters before correction are rolled back to ensure the effectiveness of self-correction.

[0183] Step Six: Model Validation and Prediction Result Output

[0184] Based on the model trained and dynamically self-corrected in step five, multi-dimensional validation is employed to ensure model accuracy, interpretability, and practicality. This overcomes the shortcomings of existing technologies, such as single validation and non-standard output formats, and outputs standardized prediction results adapted to the management needs of public institutions. Furthermore, a new scenario-based simulation function has been added to provide precise support for energy-saving decision-making, as detailed below:

[0185] After the model training and self-correction are completed, it needs to be verified in multiple dimensions. The verification criteria are: the prediction bias rate of the test set is ≤5%, and the accuracy of K-fold cross-validation (K=5) is ≥95% before it can be put into practical use.

[0186] (1) Multi-dimensional verification

[0187] Accuracy verification: The prediction bias rate of the test set is ≤5%. Compared with existing technologies (single LSTM model, CNN-LSTM model, single GNN model), the accuracy of the model in this embodiment is improved by ≥20%, ≥15%, and ≥18%, respectively. The superiority of the model is demonstrated through specific data comparison. At the same time, K-fold cross-validation (K=5) is used to verify the generalization ability of the model. The cross-validation accuracy is ≥95%, ensuring the adaptability of the model in different public institution scenarios. The verification data comes from the preprocessed data in step one and the prediction results in step four.

[0188] Explainability verification: By using the attention weights output by the attention mechanism layer in step three, the impact ratio of core factors such as "air conditioning energy consumption", "energy-saving renovation investment" and "service life" on value loss is clarified (e.g., the impact ratio of air conditioning energy consumption is ≥30%), and an explainable report is generated to meet the traceability and auditability requirements of public institutions, which is different from the "black box" defects of existing deep learning models.

[0189] Policy adaptability verification: Compare the prediction results under different policy scenarios to verify the model's adaptability to the accelerated depreciation policy for fixed assets and the energy-saving subsidy policy, and ensure the rationality of the policy factor embedding in step four; at the same time, verify the compliance of the energy consumption data processing and prediction results in step one by referring to the GB40879-2021 standard, and ensure that the model complies with industry standards.

[0190] Practicality verification: Sample assets of three different types of public institutions (government agencies, universities, and hospitals) were selected and predicted using the model in this embodiment. The deviation rate between the prediction results and the actual measurement results of the third-party asset appraisal agency was ≤4%, proving that the model can be directly applied and echoing the design intention of each step in the preceding text, thus demonstrating the practical value of the model.

[0191] (2) Prediction results output

[0192] Single asset output: Output the predicted value of a single asset at different time periods (day / month / year), the rate of loss, the core influencing factors and the attention weight (derived from step three), and the predicted value of the asset's remaining value (combined with the correlation mechanism in Example 1). The output format conforms to the public institution fixed asset management report specifications and echoes the single asset prediction formula in step four.

[0193] Multi-asset linkage output: Output the total value loss forecast and linkage impact coefficient (derived from the edge weights in step two) brought about by the linkage energy use of multiple assets under the same topology (such as the same building or the same department), clarify the core driving factors of multi-asset collaborative loss, provide support for centralized management and control, and echo the multi-asset linkage forecast formula in step four.

[0194] Scenario-based output: Simulate the value loss prediction results under different energy consumption scenarios (after energy-saving renovation, high-energy-consumption operation, standard energy use, after policy adjustment), generate scenario comparison reports, quantify the loss differences under different scenarios (e.g., the value loss rate decreases by ≥10% after energy-saving renovation), echo the energy-saving renovation effect correction factor and policy factor in step four, and provide data support for energy-saving decision-making and fiscal budget formulation of public institutions.

[0195] Early warning output: A new value loss early warning function has been added. When the predicted value loss rate exceeds the benchmark threshold (verified by the administrative affairs management department), an early warning will be automatically triggered, and the cause of the early warning (such as excessive energy consumption or untimely maintenance) and optimization suggestions will be output to achieve proactive control. This echoes the energy consumption and maintenance data in step one and the correction factor in step four, thereby improving the practical value of the model.

[0196] This embodiment constructs a dynamic self-correcting coupled quantitative model through the above steps, which comprehensively solves the core defects of existing technologies. Its novelty is reflected in: constructing a dynamic hierarchical asset topology map (step two), introducing an adaptive attention mechanism and dynamic correction factors (steps three and four), realizing dynamic adaptation of policy factors (step four), and designing a multi-condition hierarchical self-correction mechanism (step five), breaking through the static and singular limitations of traditional models. Its creativity is reflected in: the improved GNN-LSTM hybrid model (step three) takes into account both topological correlation capture and time series prediction accuracy, multi-dimensional verification (step six) and scenario-based output (step six) improve the model's practicality, interpretable design (step three) adapts to the management needs of public institutions, and policy factor embedding (step four) ensures compliance. Compared with existing technologies, it achieves a triple improvement in accuracy, adaptability, and interpretability, and can be directly applied to the collaborative management of asset energy consumption and value loss of various public institutions, while also having good scalability.

[0197] Example 3

[0198] like Figure 3As shown in the embodiments of the present invention, an intelligent energy-saving management and value synergistic optimization method for public institution assets based on energy consumption-value correlation is provided. Addressing the technical pain points of existing public institution asset energy-saving management and value optimization being disconnected and lacking precise coordination methods, this method, based on the multi-dimensional coupling correlation mechanism of energy consumption and value constructed in Embodiment 1 (the core being the multi-factor coupling correlation of energy consumption and value loss, and the calibration of the asset's natural decay coefficient) and the dynamic self-correcting coupling quantitative model constructed in Embodiment 2 (the core being topological feature extraction, dynamic correction factors, and policy factor embedding), proposes a dual-objective synergistic optimization method for energy consumption and value. This method breaks through the limitations of traditional single energy-saving or single-value management, embedding intelligent decision-making modules, constructing a collaborative evaluation system, and integrating digital twin technology to achieve synergistic advancement of energy-saving management and asset value optimization. It provides public institutions with a feasible and quantifiable synergistic optimization solution, filling the gap in existing technologies lacking collaborative control techniques for energy consumption and value. Specifically, the method includes the following steps:

[0199] Step 1: Multi-source data linkage access and collaborative preprocessing:

[0200] This step provides comprehensive and accurate data support for the collaborative optimization method. It integrates the multi-source data preprocessed in Example 2, adds real-time data from IoT sensing terminals, and completes collaborative preprocessing. This ensures seamless integration of the data with the correlation mechanism in Example 1 and the quantization model data in Example 2, providing a reliable data foundation for subsequent dual-objective optimization and intelligent decision-making. The specific process is as follows:

[0201] Data Access: The system integrates three core data categories to achieve full data coverage and real-time updates, ensuring that all data is sourced from and consistent with previous invention points, thus providing data support for subsequent collaborative optimization steps.

[0202] (1) Historical and basic data: The static asset ledger data, dynamic energy consumption monitoring data, maintenance and renovation data, and financial and policy data after preprocessing in step one of Example 2 are used. Among them, the static asset ledger data (initial value) (Statutory depreciation period T, etc.), financial data (depreciation amount, accelerated depreciation policy parameters), and the asset natural decay coefficient in Example 1. The policy factor calibration logic of Example 2 is completely consistent, and the energy consumption data strictly follows the GB40879-2021 standard to ensure that it is from the same source and standard as the quantitative model data of Example 2.

[0203] (2) Real-time sensing data: Through IoT sensing terminals deployed at various asset terminals in public institutions, real-time data are collected on asset energy consumption parameters (such as air conditioning temperature, lighting power, equipment running time), environmental parameters (real-time temperature, humidity, air quality), and asset operating status. The sampling frequency is once every 15 minutes, and the data is recorded as a real-time data set. This supplements the missing real-time operational data in Example 2, providing data support for refined regulation and overcoming the limitations of the traditional model's "static data dependence."

[0204] (3) Correlation and prediction data: Call the energy consumption and value loss correlation data and value loss prediction data (single asset and multi-asset linkage) output by step S26 of Example 2, and record them as the correlation and prediction data set. This includes the energy consumption-value correlation coefficient of each asset (using the coupling correlation logic of Example 1) and the rate of value loss. The core influencing factors (energy consumption, maintenance frequency, etc.) are ensured to be completely consistent with the output data of the quantitative model in Example 2, so as to provide accurate correlation basis for subsequent dual-objective optimization and intelligent decision-making.

[0205] Collaborative preprocessing: Building upon the data preprocessing in Example 2, real-time data verification and alignment with related data are added to ensure data consistency and timeliness. This also aligns with the preprocessing specifications of the preceding invention points, laying a data foundation for the smooth progress of subsequent steps.

[0206] (1) Real-time data verification: An outlier detection algorithm is used to remove abnormal data collected by IoT sensing terminals (such as sudden changes in energy consumption caused by sensor failure). The detection formula is as follows:

[0207]

[0208] Parameter description: outlier is the outlier identifier (1 for outlier, 0 for normal). To collect data in real time (energy consumption, temperature, humidity, etc.). This is the sliding window average of real-time data over the past hour. The standard deviation of the sliding window of real-time data in the past hour is used. Abnormal data are supplemented by linear interpolation of adjacent valid real-time data (consistent with the missing value handling method in step one of Example 2, to ensure the consistency of parameter processing logic).

[0209] (2) Data alignment: Real-time sensing data, historical basic data, and related prediction data are aligned according to “asset ID-timestamp” to ensure that the energy consumption data, value data, and operation data of the same asset at the same time correspond one-to-one. The alignment accuracy is 15 minutes, which provides a unified data dimension for subsequent dual-objective optimization and intelligent decision-making. Among them, the value data alignment needs to match the value loss calculation logic of Example 1 and the prediction data format of Example 2.

[0210] (3) Feature fusion: Extract core features for collaborative optimization, including energy consumption features (real-time energy consumption, energy consumption fluctuation coefficient) and value features (value loss rate). Features with a correlation ≥ 0.7 with the dual objective optimization were selected using the mutual information method in step one of Example 2. Redundant features were eliminated to ensure that the feature extraction method was consistent with the previous invention point.

[0211] This step completes data preprocessing and feature selection, outputting standardized data to provide accurate input for the construction of the dual-objective optimization function in step two, thus achieving a data closed-loop connection with Examples 1 and 2.

[0212] Step 2: Construction of the Energy Consumption-Value Dual-Objective Co-optimization Function

[0213] Based on the energy consumption-value multi-dimensional coupling correlation mechanism of Example 1 (positive correlation logic between energy consumption and value loss), combined with the correlation data (value loss rate) output by the quantitative model of Example 2, Policy factors (etc.), and the core feature data after collaborative preprocessing in step one, construct an energy consumption-value dual-objective collaborative optimization function, clarify the optimization objectives and constraints, achieve synergistic improvement of energy-saving benefits and asset value preservation, break the barriers of traditional single energy saving or single value control, solve the core pain point of "energy saving and value disconnect" in existing technologies, and provide optimization basis for subsequent intelligent decision-making modules, as detailed below:

[0214] Optimization Target Setting: Establish a dual-objective optimization direction, balancing the maximization of energy-saving benefits and the maximization of asset value preservation rate. The two objectives are synergistically coupled to avoid mutual conflict, while also incorporating the characteristic data from step one:

[0215] (1) Objective 1 (Energy Saving Objective): Under the premise of meeting the normal office energy needs of public institutions, minimize the total annual energy consumption of assets, that is, maximize the energy saving rate. The objective function is as follows:

[0216]

[0217] Parameter description: The annual energy saving rate (%) of the asset. The total annual energy consumption of the asset before optimization (unit: kWh) is derived from the historical energy consumption data of step one in Example 2, and is completely consistent with the energy consumption statistics standard (GB40879-2021) of Example 2; The total annual energy consumption of the asset after optimization (unit: kWh) is the variable to be optimized. Its value must meet the energy consumption quota constraint, and at the same time refer to the real-time energy consumption characteristics collected in step one.

[0218] (2) Objective 2 (Value Optimization Objective): Maximize the annual asset value preservation rate and minimize value loss. The objective function is as follows:

[0219]

[0220] Parameter description: The annual asset value preservation rate (%). The initial annual value of the asset (unit: yuan) is based on the static asset ledger data from step one of Example 2 (i.e.). Consistent with the definition of the initial asset value in Example 1; The optimized year-end asset value (unit: yuan). , To optimize the asset value loss at time t, the prediction formula in step four of Example 2 is used to calculate the value loss quantification logic, ensuring complete consistency with the value loss quantification logic in Example 2, while also incorporating the value characteristic data from step one.

[0221] Constraint Setting: Three types of constraints are set based on the energy needs of public institutions, asset operation patterns, and policy requirements to ensure the optimization plan is feasible. All constraint parameters are aligned with the previous invention points and match the real-time data characteristics of step one, without any conflicts.

[0222] 1) Energy Constraints: The optimized asset energy consumption parameters must meet the normal office needs of public institutions, such as air conditioning temperature not lower than 26℃ in summer and not higher than 20℃ in winter, and lighting power not lower than 80% of the benchmark value. The constraint formula is as follows:

[0223]

[0224] Parameter description: To optimize the air conditioning temperature, To optimize the lighting power, The reference power for lighting is consistent with the definition of "rated power" in the energy consumption monitoring data of step one in Example 2, to ensure parameter consistency, while also referring to the range of real-time energy consumption parameters collected in step one.

[0225] 2) Asset operation constraints: The optimized asset runtime and maintenance frequency must comply with asset usage specifications to avoid excessive energy saving leading to increased asset wear and tear. The constraint formula is as follows:

[0226]

[0227] Parameter description: To optimize the daily runtime of the assets, , These are the minimum and maximum daily runtimes for the asset (determined by the asset type, consistent with the asset ledger data in step one of Example 2). To optimize the frequency of asset maintenance, , These represent the minimum and maximum maintenance frequencies for the asset (derived from the maintenance and modification data in step one of Example 2, and the maintenance frequency correction factor in Example 2). (The calibration basis is consistent), and the asset operation status data from step one is also used.

[0228] 3) Policy Constraints: The optimization plan must comply with the accelerated depreciation policy for fixed assets and the energy consumption quota standard (GB40879-2021). The constraint formula is as follows:

[0229]

[0230] Parameter description: The annual energy consumption quota for assets is specified in GB40879-2021, which is consistent with the energy consumption data statistics standard in step one of Example 2; To optimize the accelerated depreciation factor for assets, the range of values ​​is the same as the policy factor in step four of Example 2. Completely consistent, ensuring unified policy factor calibration logic and no parameter conflicts.

[0231] Dual-objective optimization solution: The non-dominated sorting genetic algorithm (NSGA-II) is used to solve the dual-objective optimization function, obtaining the Pareto optimal solution set. The optimal solution that balances energy saving and value preservation is selected. The solution process combines the data and parameters from the previous invention points, and relies on the collaborative preprocessing data from step one' to ensure the logical coherence of the solution. The solution results directly provide the core basis for the intelligent decision-making in step three'. The specific steps are as follows:

[0232] (1) Initialize the population: Set the population size to 100, the number of iterations to 50, the crossover probability to 0.8, the mutation probability to 0.05, and the individuals in the population to be optimized (air conditioning temperature, lighting power, maintenance frequency, etc.). The maintenance frequency variable needs to match the maintenance frequency correction factor in step four of Example 2. The range of values, while referring to the real-time running data in step one;

[0233] (2) Fitness calculation: Substitute each individual in the population to calculate the energy saving rate. With value retention rate As a fitness value, the calculation of the value retention rate needs to call the value loss prediction formula in step four of Example 2 to ensure consistency with the quantitative logic of Example 2.

[0234] (3) Non-dominated ranking: Non-dominated ranking of individuals in the population, dividing them into dominance levels, selecting the best individuals that are not dominated by other individuals, and prioritizing the retention of individuals with "high energy saving rate and high value preservation rate", which meets the core requirements of dual-objective collaborative optimization.

[0235] (4) Crowding degree calculation and selection: Calculate the crowding degree of the optimal individual, retain individuals with higher crowding degree, avoid the optimal solution set concentration, ensure the diversity of solution set, and cover asset management scenarios of different public institutions;

[0236] (5) Iteration termination: When the number of iterations reaches 50 or the optimal solution set converges (the deviation of the optimal fitness value between two adjacent iterations is ≤1%), the iteration stops and the Pareto optimal solution set is output as the basis for subsequent intelligent decision-making. The solution set parameters must meet the constraints and be consistent with the parameter specifications of the previous embodiment.

[0237] This step completes the construction and solution of the bi-objective optimization function, outputs the optimal solution set, provides the core basis for intelligent decision-making in step three, and forms a logical chain of "data-optimization-decision".

[0238] Step 3: Construction and Operation of the Intelligent Decision-Making Module

[0239] An intelligent decision-making module is constructed, embedding the dynamic self-correcting coupled quantization model of Example 2 and the dual-objective optimization solution set of Step 2'. Based on the multi-source collaborative data of Step 1', it addresses the pain point of "inaccurate collaborative strategy". The generated strategy will be directly input into the digital twin virtual image of Step 4' for simulation verification, as follows:

[0240] Asset Classification: Based on the collaborative data from Step 1, the K-means clustering algorithm is used to classify public institution assets into four categories according to their energy consumption level (high / low) and value loss rate (high / low). The clustering formula is shown below. The clustering features are fully aligned with the data from the preceding invention points, and the classification results provide the core basis for generating differentiated strategies.

[0241]

[0242] Parameter description: J is the clustering objective function (minimizing the sum of squared errors within each cluster), k=4 is the number of clusters. Let x be the asset set of the kth class, and x be the asset feature vector (energy consumption data comes from step one of Example 2, value loss rate). (Derived from step four of Example 2, the feature data has undergone preprocessing in step one). The feature center of the k-th asset class; the four asset classes are: high energy consumption and high loss assets ( High-energy-consuming and low-loss assets Low-energy-consumption, high-loss assets Low-energy-consumption and low-loss assets The classification results directly support the generation of subsequent differentiation strategies.

[0243] Intelligent Decision Module Architecture: The module comprises a data input layer, a model embedding layer, a policy generation layer, and an output layer. The functions of each layer are closely linked to the preceding invention points and steps, ensuring the module's feasibility and providing support for subsequent policy simulation and implementation.

[0244] (1) Data input layer: The collaborative preprocessed data from step one and the bi-objective optimization solution set from step two are connected. The data format is consistent with the output format of the quantization model in Example 2.

[0245] (2) Model embedding layer: embeds the dynamic self-correcting coupled quantization model of Example 2, calls the value loss prediction data and energy consumption-value correlation data output by the model in real time, ensures the accuracy of strategy generation, and deeply integrates with the quantization capability of Example 2;

[0246] (3) Strategy generation layer: Based on the asset classification results and combined with the dual-objective optimization solution set, differentiated collaborative strategies are generated. The core logic is "classification and implementation of policies, and synergy and consideration", which breaks through the traditional single control logic and reflects novelty.

[0247] (4) Output layer: Output standardized collaborative optimization strategy solutions, including energy control parameters, maintenance plans, energy-saving renovation suggestions, etc., and push them to the public institution asset management platform, IoT terminals, and the digital twin virtual image of step four' to ensure that the solution can be implemented and directly used for simulation verification, connecting the IoT sensing terminal data of step one' and the simulation requirements of step four'.

[0248] Differentiated Collaborative Strategy Generation: Different collaborative strategies are generated for the four types of assets to ensure that energy saving and value optimization are promoted in synergy. All strategy parameters are consistent with the settings in Examples 1 and 2, without conflict, and are also in line with the real-time operation characteristics of the assets in Step 1. After the strategies are generated, they are directly used for the simulation in Step 4.

[0249] Category (high-energy-consuming and high-loss assets, such as old air conditioners and large-scale water supply equipment): A combined solution of "energy-saving renovation + refined energy consumption control" will be adopted, as detailed below:

[0250] (1) Energy-saving renovation: Recommend targeted renovation solutions (such as air conditioner inverter renovation, equipment replacement), renovation investment Determined based on the type of modification technology (refer to Example 2). Definition, and policy factors (Consistent with the calibration logic) Energy consumption reduction after the modification The energy-saving renovation effect correction factor predicted by the quantitative model in Example 2 and step S24 of Example 2 is... The calculation basis is consistent;

[0251] (2) Refined energy consumption control: Energy consumption parameters are adjusted in real time through IoT sensing terminals. The control formula is as follows:

[0252]

[0253] Parameter description: The optimized energy consumption at time t, The real-time energy consumption at time t is derived from the real-time sensing data in step one'. The control coefficient (values ​​range from 0.1 to 0.3, determined by the asset type, consistent with the asset type classification in Example 2) is used. To ensure the optimal energy saving rate obtained in step two', the control parameters are consistent with the dual-objective optimization results and the preceding data.

[0254] Category (high-energy-consuming, low-loss assets, such as large conference room equipment and newly added office equipment): Adopt the "energy consumption period optimization + energy consumption limit control" solution to optimize energy consumption periods, avoid peak energy consumption periods, and set daily energy consumption limits. The control formula is as follows:

[0255]

[0256] Parameter description: The optimized energy consumption at time t. The daily energy consumption limit for assets (determined by the GB40879-2021 energy consumption quota standard, and the energy consumption quota in step one of Example 2) (Logically consistent), while also referring to the real-time energy consumption fluctuation data in step one.

[0257] Category (low-energy-consumption, high-loss assets, such as old office furniture and small repair equipment): Adopt the "optimized maintenance frequency + condition monitoring" solution to optimize maintenance frequency and reduce maintenance energy consumption and costs. The maintenance frequency optimization formula is as follows, which is deeply integrated with the dynamic correction factor in Example 2:

[0258]

[0259] Parameter description: To optimize the maintenance frequency, The baseline maintenance frequency (derived from maintenance data in step one of Example 2). This refers to the optimized maintenance frequency correction factor in the dynamic correction factor of Example 2 (using the formula from Example 2). The calibration logic is only finely adjusted based on the dual-objective optimization results to ensure no conflict with Example 2, while also incorporating the asset operation status data from step one.

[0260] For low-energy-consumption and low-loss assets, such as energy-saving office equipment and new furniture: adopt the "routine management and control + regular inspection" scheme, maintain the existing energy consumption and maintenance mode, and conduct regular (quarterly) inspections to avoid abnormal energy consumption and increased value loss. The inspection frequency is determined by the asset type (consistent with the asset classification and maintenance specifications in Example 2) to ensure the continuity of management and control logic, while connecting with the real-time status monitoring data of Step 1.

[0261] This step generates a differentiated collaborative strategy, which is then directly input into the digital twin virtual image from step four for simulation verification, ensuring the effect prediction before the strategy is implemented.

[0262] Step 4: Building and simulating the digital twin virtual image

[0263] By integrating digital twin technology, a virtual mirror of the energy consumption and value of public institution assets is built, mapping the energy consumption, operation, and value change status of physical assets. The differentiated collaborative strategies generated in step three are input into the mirror for simulation, predicting the value loss trend under different energy-saving management strategies. This provides predictive support for the implementation of collaborative optimization strategies, avoiding blind policy implementation and addressing the pain point of existing technologies where "strategy effects cannot be predicted." Furthermore, it is deeply integrated with the models and data from the previous invention points and the aforementioned steps. The optimal strategy verified through simulation will be directly used for the evaluation in step five and the implementation in step six, as detailed below:

[0264] Digital Twin Virtual Image Construction: Based on the actual parameters of the physical asset, a 1:1 virtual image is constructed, comprising three core modules: asset model, energy consumption model, and value loss model. All module parameters are connected to the collaborative data from Examples 1 and 2, and Step 1, ensuring consistency between the virtual image and the physical asset and the preceding model, thus providing a foundation for accurate simulation.

[0265] (1) Asset Model: Based on the static asset ledger data in step one (which is consistent with step one in embodiment two), a three-dimensional asset model is constructed, and the asset type, installation location, material, and initial value are labeled. These parameters enable precise mapping of physical assets, ensuring consistency with asset base parameters and topology node attributes.

[0266] (2) Energy consumption model: The energy consumption time series prediction model of Example 2 is embedded (the improved GNN-LSTM model in step 3), and the real-time energy consumption data of the IoT sensing terminal is synchronized in real time (step 1'). The simulation of asset energy consumption changes under different energy parameters is completely consistent with the energy consumption prediction logic of Example 2, ensuring the accuracy of simulation data.

[0267] (3) Value loss model: The dynamic self-correcting coupled quantitative model of Example 2 (prediction formula in step 4) is embedded, and the energy consumption data output by the energy consumption model is associated to simulate the value loss trend of assets under different energy consumption scenarios. The value loss calculation logic is completely consistent with the coupling and association mechanism of Example 1 and the quantitative model of Example 2 to ensure the reliability of the simulation results, and at the same time connects the value characteristic data of step 1'.

[0268] Strategy Simulation and Optimization: The differentiated collaborative strategies generated in Step 3 are input into the digital twin virtual mirror to simulate the implementation effects of different strategies over the next 1-3 years. The simulation results are output. The simulation process strictly relies on the model and parameters of the previous invention points, the collaborative data from Step 1, and the optimization objectives from Step 2. The optimal strategy after simulation and optimization will be used for subsequent evaluation and implementation, as detailed below:

[0269] (1) Simulation parameter settings: Set the simulation period (1-3 years), time step (1 month), and input the energy consumption parameters, maintenance frequency, energy-saving renovation plan, etc. corresponding to the strategy. Among them, the parameters such as maintenance frequency and energy-saving renovation investment are the same as those in Example 2. , The definitions are consistent, and the real-time data characteristics of step one are matched.

[0270] (2) Simulation process: The virtual image is based on the energy consumption model and the value loss model, and calculates the asset energy consumption, value loss and energy saving rate in real time during the simulation period. Value retention rate The dynamic change curve is generated, and the calculation logic is consistent with the prediction formula of Example 2. At the same time, the real-time running data of Step 1' is referenced to ensure that the simulation fits the actual scenario.

[0271] (3) Strategy optimization: Compare the simulation results of different strategies and select the optimal solution in terms of energy saving rate and value preservation rate. If the simulation results do not meet the dual-objective optimization requirements (such as energy saving rate <10% and value preservation rate <85%), return to step two to readjust the optimization function parameters (such as constraints and weight coefficients), iterate the optimization strategy until the requirements are met, and ensure that there are no conflicts in the parameters during the iteration process. The optimized optimal strategy will be pushed to step five for collaborative evaluation and step six for implementation.

[0272] Virtual-Physical Linkage: Establish a real-time linkage mechanism between virtual images and physical assets, synchronizing the simulation parameters of the virtual image to the IoT sensing terminal (the sensing terminal in step one), and the real-time operational data of the physical assets (in step one). Synchronize to the virtual image to achieve a closed loop of "virtual simulation - physical execution - data feedback - strategy optimization", connect the data and models of all previous steps, reflect the integrity of collaborative optimization, and provide support for the implementation and dynamic updates of the solution in step six.

[0273] This step completes the strategy simulation verification and outputs the optimal feasible strategy, providing a verification basis for the implementation of step five, 'collaborative evaluation', and step six.

[0274] Step 5: Construction and Real-time Assessment of Energy Consumption-Value Collaborative Evaluation System

[0275] An energy consumption-value collaborative evaluation system is constructed, introducing dual evaluation indicators of energy-saving benefits and value retention rate. Combining the quantitative model data from Example 2, the collaborative data from Step 1', the simulation results from Step 4', and the actual operational data after implementation in Step 6', a real-time linkage assessment of energy-saving management effects and asset value changes is achieved. This ensures the effectiveness of the collaborative optimization scheme and addresses the pain point of existing technologies where "effects cannot be quantitatively evaluated." The evaluation indicators are completely consistent with the parameters and logic of the preceding invention points. The evaluation results will directly trigger the dynamic update of the scheme in Step 6', as detailed below:

[0276] Collaborative evaluation index system: This system includes 2 primary indicators and 6 secondary indicators. Indicator weights are determined using the Analytic Hierarchy Process (AHP) to ensure the scientific rigor of the evaluation. All indicators are linked to the preceding steps, and all indicator data originates from previous steps to ensure evaluation accuracy.

[0277] (1) Primary indicator 1: Energy saving efficiency (weight 0.45), which includes 3 secondary indicators. The data of the indicators are all from the previous steps and are consistent with the energy consumption statistics logic of Example 2:

[0278] Energy saving rate (weight 0.20): reflects the energy saving effect, consistent with the dual-objective optimization objective of step two', and the data comes from the simulation results of step four' or the actual operation data of step six';

[0279] Absolute decrease in energy consumption (weight 0.15): Unit: kWh, reflecting the actual total energy saved. , Consistent with the energy consumption data definition in Example 2, the data comes from the historical data in Step 1 and the simulated data in Step 4 or the actual data in Step 6.

[0280] Energy consumption compliance rate (weight 0.10): This reflects the compliance of energy consumption management. The compliance standard is GB40879-2021 energy consumption quota (consistent with the constraints of steps one and two' in Example 2). The data comes from the real-time energy consumption data of step one' or the actual operation data of step six'.

[0281] (2) Primary indicator 2: Value preservation benefit (weight 0.55), which includes 3 secondary indicators. The indicator data is consistent with the value loss quantification logic of Examples 1 and 2, and the data comes from the previous steps:

[0282] Value preservation rate (weight 0.25): reflects the asset value optimization effect, consistent with the dual-objective optimization objective of step two', the calculation logic is connected with the value loss prediction formula of example two, and the data comes from the simulation results of step four' or the actual data of step six';

[0283] Value loss rate (weight 0.20): Parameter description: Value loss reduction rate (%) To optimize the total value loss of the previous year (calculated by the prediction formula in step four of Example 2). To optimize the total annual value loss (compared to step two) (The calculation basis is consistent) to ensure that there is no conflict with the value loss quantification logic of Example 2. The data comes from the simulation results of step 4' or the actual data of step 6'.

[0284] Asset residual value enhancement rate (weight 0.10): Parameter description: The percentage increase in residual value of assets. To optimize the year-end value of previous assets (calculated using the prediction formula in step four of Example 2). To optimize the year-end value of assets, consistent with the asset value definition in Example 1 and the prediction logic in Example 2, the data comes from the simulation results in step four or the actual data in step six.

[0285] (3) Comprehensive evaluation calculation: The comprehensive evaluation score is calculated using the weighted summation method. The formula is as follows. The parameters are completely consistent with the evaluation index system to ensure that the calculation logic is rigorous. The calculation results provide the core basis for the scheme update in step six.

[0286]

[0287] Parameter description: S is the overall evaluation score (range 0~100 points). As the weight of the primary indicator, As secondary indicator weights, Standardize the scores for the secondary indicators (using Equation 9 from Step 1 of Example 2 to standardize to 0-100 points, ensuring consistency in the standardization method).

[0288] (4) Real-time evaluation and feedback: Collect optimized energy consumption and value data every quarter (consistent with the data source of Example 2, derived from the actual operation data of step six'), substitute them into the evaluation system to calculate the comprehensive score, and divide the evaluation level according to the score (excellent: S≥90 points, good: 80≤S<90 points, qualified: 70≤S<80 points, unqualified: S<70 points); if the evaluation level is unqualified, trigger the intelligent decision module of step three', readjust the collaborative strategy, and at the same time update the dynamic self-correction coupled quantitative model parameters of Example 2 to realize the dynamic iteration of collaborative optimization, connect the dynamic self-correction mechanism of Example 2, ensure the overall logical closed loop, and provide trigger conditions for the dynamic update of the scheme in step six'.

[0289] Step Six: Implementing and Dynamically Updating the Collaborative Optimization Solution

[0290] The collaborative strategy, optimized in step three, verified through simulation in step four, and evaluated successfully in step five, is applied to the asset management of public institutions. Combined with the dynamic self-correction mechanism of Example Two and the real-time evaluation results of step five, the solution is updated in real time to ensure long-term collaborative optimization effects. The implementation process strictly follows the parameter specifications and model logic of the preceding invention points, while connecting all the aforementioned steps to form a complete collaborative optimization closed loop, as detailed below:

[0291] Solution Implementation: The differentiated collaborative strategy will be pushed to public institution asset management platforms, IoT sensing terminals, and operations and maintenance departments, clearly defining implementation standards and division of responsibilities. The implementation process will connect with preceding data and models to ensure operability, while providing actual operational data for the real-time assessment in step five.

[0292] (1) Energy consumption regulation: The IoT sensing terminal automatically adjusts the energy consumption parameters according to the strategy. The parameter range meets the constraints of step two. The operation and maintenance personnel regularly check the regulation effect and synchronously collect real-time data to feed back to step one, providing support for data updates and providing energy consumption data for the evaluation of step five.

[0293] (2) Energy-saving renovation: Promote energy-saving renovation projects according to the strategy and record the renovation investment. The effect of the renovation (energy consumption reduction) ), synchronously updated to the dynamic self-correcting coupled quantization model of Example 2 (for adjusting) ( ) and digital twin mirror to ensure that the model is synchronized with the actual scene, and at the same time provide data on the transformation effect for the evaluation in step five;

[0294] (3) Maintenance execution: The operation and maintenance department shall perform maintenance according to the optimized frequency. Conduct maintenance work, record maintenance data, and synchronously update the maintenance and modification data set of Example 2 to provide dynamic correction factors. This provides support for recalibration, connects with the dynamic self-correction mechanism of Example 2, and provides maintenance-related data for the evaluation in step 5'.

[0295] (4) Dynamic Update: Combining the dynamic self-correction mechanism of Example 2 and the real-time evaluation results of step 5', set the scheme update trigger conditions. When any condition is met, update the collaborative optimization scheme to ensure that the scheme is synchronized with asset status, policy requirements, and data changes, without parameter conflicts, and connects with all previous steps to achieve closed-loop optimization.

[0296] After the solution is updated, return to step four for simulation verification and step five for real-time evaluation until the dual-objective optimization requirements are met, forming a complete closed loop of "implementation - real-time evaluation - dynamic update - simulation verification" to ensure the long-term adaptability of the collaborative optimization solution.

[0297] This embodiment fully realizes the synergistic optimization of energy consumption and value of public institution assets through the above steps.

[0298] Example 4

[0299] like Figure 4 As shown in the figure, the present invention provides a collaborative monitoring and early warning system for energy consumption and value loss based on multi-source heterogeneous data fusion. This system addresses the technical pain points of existing public institution asset energy consumption and value data being scattered, monitoring lagging, lacking collaborative early warning capabilities, and unable to support the implementation of the aforementioned multi-dimensional coupling correlation mechanism of energy consumption and value, dynamic self-correcting coupling quantification model, and energy consumption-value dual-objective collaborative optimization method. It overcomes the technical bottlenecks of existing systems such as "data fragmentation, poor protocol compatibility, single early warning, and lack of collaborative support capabilities." The system designs a three-layer collaborative architecture of "perception layer – network layer – platform layer" to achieve full-scenario collection, efficient transmission, and deep fusion of multi-source heterogeneous data. It designs a collaborative early warning mechanism for energy consumption anomalies and accelerated value loss, while providing visualization and multi-system integration capabilities to adapt to the needs of different types of public institutions. This provides intelligent support for the full-process asset management model and fills the technical gaps in existing systems, such as data barriers, incomplete monitoring, and untimely early warning. Specifically, it includes the following modules:

[0300] Module M41: Intelligent Monitoring and Multi-Source Heterogeneous Data Acquisition Module for the Perception Layer

[0301] This module, serving as the system's data acquisition terminal, differs from existing technologies in that it avoids the shortcomings of "single energy consumption monitoring and homogeneous deployment." It adopts a "categorized deployment and precise data collection" model, taking into account the differentiated characteristics of public institution asset types and energy consumption scenarios. It deploys various intelligent monitoring units to collect multi-source heterogeneous data related to asset energy consumption and value, providing raw data support for subsequent network layer data transmission, platform layer data fusion, and collaborative early warning. The collected data strictly matches the data source specifications of the aforementioned embodiments, effectively addressing the technical pain points of existing systems such as "incomplete data collection and disconnect from valuable data." Specific functions are as follows:

[0302] Intelligent monitoring unit deployment: Breaking away from the existing "one-size-fits-all" deployment model, intelligent monitoring units are deployed differently based on the type of public institution assets (such as office equipment, air conditioning systems, water and power supply equipment, office furniture, etc.) and energy consumption scenarios (centralized office areas, hospital treatment areas, university teaching buildings, etc.), achieving full-scenario coverage without blind spots. Each monitoring unit has an integrated function of "data collection - preliminary verification - real-time upload," unlike the limitations of existing monitoring units that only have a single data collection function. The specific deployment is as follows:

[0303] (1) Energy consumption monitoring unit: Deploy high-precision power meters, intelligent energy consumption metering equipment and smart sockets to cover the entire chain of asset energy consumption. It is used to collect real-time energy consumption data and energy consumption parameters (such as power, voltage and current). The sampling frequency is 1 time / 15 minutes, which is consistent with the real-time sensing data sampling frequency in step one of the embodiment. The collection accuracy is improved to ±0.5%, which effectively overcomes the technical problems of low accuracy and sampling lag of existing energy consumption monitoring equipment.

[0304] (2) Operation monitoring unit: Deploy high-precision temperature sensor, humidity sensor and vibration sensor to collect asset operation status data (such as equipment operating temperature, vibration frequency and running time). The sampling frequency is 1 time / 10 minutes, which supplements the asset operation data dimension of the above embodiment and solves the defect of the existing system that only monitors energy consumption and ignores the impact of asset operation status on value loss.

[0305] (3) Maintenance and financial monitoring unit: Innovative deployment of mobile operation and maintenance terminals and financial data collection terminals to realize real-time collection and synchronous update of maintenance data and financial data. Collect asset maintenance data (maintenance time, maintenance content, maintenance cost) and financial data (original asset value, depreciation amount, value change), which is completely consistent with the maintenance and financial data specifications in step one of Example 2, breaking through the problem of the existing system's financial data and energy consumption data being separated and the update being lagging behind;

[0306] (4) Environmental monitoring unit: Deploy multi-parameter environmental sensors to collect environmental parameters of the asset (real-time temperature and humidity, air quality, light intensity) to provide environmental impact factor data for subsequent energy consumption-value correlation analysis of the platform layer. This is consistent with the environmental parameter collection requirements of step one in Example 3, filling the technical gap of the existing system ignoring the synergistic impact of environmental factors on energy consumption and value loss.

[0307] Multi-source heterogeneous data acquisition: Based on the differentiated deployment of monitoring units, this system breaks through the limitations of existing systems that only collect "single energy consumption data," and innovatively collects four types of core multi-source heterogeneous data. The data format and statistical standards are fully compatible with the aforementioned embodiments, realizing the collaborative collection of "energy consumption-operation-value-environment" data. This lays a solid foundation for subsequent edge processing at the network layer, deep integration at the platform layer, and collaborative early warning. Specifically, this includes:

[0308] (1) Energy consumption data: real-time energy consumption data Historical energy consumption data Energy consumption parameter data (power) ,Voltage Current It strictly follows the GB40879-2021 energy consumption statistics standard, maintains consistency with the energy consumption data definitions in Examples 2 and 3, and adds energy consumption fluctuation data to provide data support for subsequent platform-level anomaly detection;

[0309] (2) Operational data: Asset runtime Operating temperature Vibration frequency This provides core operational feature data for the dynamic self-correcting coupled quantization model in Example 2, effectively solving the problems of insufficient operational data support and inaccurate value loss prediction in existing systems.

[0310] (3) Value-related data: initial value of assets Real-time value Value loss Depreciation coefficient It is completely consistent with the asset value definition in Example 1 and the value loss quantification logic in Example 2, realizing the synchronous collection of value data and energy consumption data, breaking through the limitation of the existing system that the two are disconnected;

[0311] (4) Auxiliary data: Environmental parameters (temperature) ,humidity ), maintenance data (maintenance frequency) Maintenance costs This provides important auxiliary support for subsequent energy consumption-value correlation analysis and collaborative early warning at the platform layer, filling the technical gap of missing auxiliary data in the existing system.

[0312] Preliminary data verification: To ensure the accuracy of subsequent data processing, an innovative preliminary verification mechanism of "dual thresholds + anomaly labeling" is adopted to perform preliminary anomaly detection on the collected raw data and remove obviously abnormal data (such as extreme values ​​caused by sensor failures or data transmission errors). The detection formula is as follows. This lays the foundation for further verification at the network layer and data cleaning at the platform layer, which is different from the shortcomings of existing systems that lack preliminary verification or have simple verification logic.

[0313]

[0314] Parameter description: This is used to identify abnormal data (1 for abnormal, 0 for normal). The raw data collected (energy consumption, temperature, value, etc.). , The reasonable value range for this type of data is determined by the asset type and industry standards, such as the reasonable operating temperature range of air conditioners being 5℃~35℃, consistent with the energy consumption constraints in step S32 of Example 3. Abnormal data is marked and temporarily stored, pending further verification by the network layer, while normal data is uploaded to the network layer in real time. This measure can effectively reduce the subsequent data processing pressure and improve the overall operating efficiency of the system.

[0315] Module M42: Data Convergence Protocol Conversion and Edge Processing Module

[0316] This module deploys an industrial-grade intelligent gateway, serving as the intermediate hub between the perception layer and the platform layer. It receives and preliminarily verifies the multi-source heterogeneous data collected in step M41. Unlike existing systems that only transmit data without edge processing capabilities, this module integrates "protocol conversion, data aggregation, edge computing, breakpoint resumption, and encrypted transmission" into a single function. This enables efficient aggregation, compatible transmission, and preliminary processing of multi-source heterogeneous data, effectively addressing technical pain points such as incompatible data protocols between different monitoring units, unstable transmission, and data loss. It ensures efficient, complete, and secure data transmission to the platform layer, providing reliable support for subsequent data fusion and collaborative early warning at the platform layer. Specific functions are as follows:

[0317] Multi-protocol conversion: Addressing the heterogeneous protocols (such as Modbus, LoRa, NB-IoT, WiFi, etc.) of different monitoring units in the sensing layer, this system overcomes the problems of poor compatibility and low conversion efficiency in existing systems. Through the built-in adaptive protocol conversion module of the smart gateway, it automatically identifies the original protocol type and uniformly converts various types of data into the standard TCP / IP protocol, achieving interconnection and interoperability of heterogeneous data. The conversion formula is as follows, ensuring data format uniformity and increasing conversion efficiency to over 99%, providing a unified data foundation for subsequent data aggregation and edge computing.

[0318] Parameter description: This is the converted standard TCP / IP protocol data. The raw heterogeneous protocol data collected by the perception layer. For raw data protocol type, For standard TCP / IP protocol, As an adaptive protocol conversion function, it can automatically adapt to the protocol conversion requirements of different types of monitoring units without manual configuration, effectively solving the problems of existing protocol conversion requiring manual intervention and poor adaptability.

[0319] Data Aggregation and Edge Computing: The intelligent gateway centrally aggregates the converted standard data and innovatively performs edge computing processing. This breaks through the limitations of existing systems where "all data is uploaded to the cloud for processing, resulting in concentrated pressure," effectively reducing the data processing pressure at the platform layer, improving system response speed, and saving processing time for subsequent deep integration and collaborative early warning at the platform layer. The core processing content of edge computing includes:

[0320] (1) Data normalization: Multi-source data of different magnitudes (such as energy consumption data in kWh and value data in yuan) are normalized to the [0,1] interval to facilitate subsequent deep data fusion and correlation analysis at the platform layer. The normalization formula is as follows, which is consistent with the standardization method in step one of Example 2. At the same time, the normalization algorithm is optimized to further improve the processing efficiency:

[0321]

[0322] Parameter description: For the normalized data, For the original standard data, , Consistent with module M41, it ensures the continuity of data processing logic, while normalization is completed in advance through edge computing to avoid repeated processing in the cloud and significantly improve system response speed.

[0323] (2) Preliminary data fusion: A simple weighted fusion algorithm is used to perform preliminary fusion of multi-dimensional data (such as energy consumption, temperature, and runtime) of the same asset at the same time to obtain comprehensive asset status data. The fusion formula is as follows. This method is different from the shortcomings of existing systems, such as lack of edge fusion and poor data correlation, and lays a solid foundation for subsequent deep fusion at the platform layer. At the same time, redundant data is eliminated to improve data transmission efficiency.

[0324]

[0325] Parameter description: This is the comprehensive asset status data after initial edge integration. Data dimensions (such as energy consumption, temperature, and runtime). The weights for the i-th type of data are determined by the Analytic Hierarchy Process (AHP), such as 0.5 for energy consumption data, 0.3 for temperature data, and 0.2 for runtime data. The normalized result of the i-th type of data can be used to initially merge data, effectively eliminating redundant data and further improving data transmission efficiency.

[0326] Resume Transmission and Data Integrity Guarantee: To ensure the complete transmission of initially integrated data to the platform layer, an innovative sliding window-based resume transmission mechanism is adopted. This overcomes the problems of data loss and low retransmission efficiency after network interruption in existing systems. When the network is interrupted, the intelligent gateway temporarily stores the untransmitted data and resumes transmission from the point of interruption after the network is restored. Simultaneously, a data checksum ensures data transmission integrity. The checksum formula is as follows, ensuring a data loss rate of ≤0.1%, far superior to the ≥1% loss rate of existing systems. This provides complete data support for platform-level data processing and effectively solves the pain point of poor data transmission integrity in existing systems.

[0327]

[0328] Parameter description: For data validation results, This is an XOR check function. The basic checksum (preset by the system) is used; if the checksum result matches the preset checksum, the data transmission is complete; otherwise, a retransmission mechanism is triggered to ensure that the data loss rate is ≤0.1%, providing complete data support for platform-level data processing and solving the pain point of poor data transmission integrity in existing systems.

[0329] Encrypted Data Transmission: Considering the security of public institution data, the AES-128 encryption algorithm is used to encrypt transmitted data, overcoming the shortcomings of existing systems where data transmission is unencrypted or has a low encryption level. This prevents data leakage and tampering. The encryption formula is as follows, ensuring secure data transmission, meeting the data security management requirements of public institutions, and providing security for subsequent platform-level data decryption and cleansing.

[0330]

[0331] Parameter description: For encrypted data, For AES-128 encryption functions, The system is pre-set with an encryption key, which is automatically updated periodically to ensure secure and controllable data transmission, effectively solving the problem of data transmission security vulnerabilities in existing systems.

[0332] Module M43: Network Layer Data Fusion Collaborative Early Warning and Integrated Management Module

[0333] This module serves as the core processing and control center of the system, receiving encrypted and fused data transmitted from the network layer. Unlike existing systems that suffer from shallow data fusion, limited early warning capabilities, and lack of model support, this module integrates data cleaning, deep fusion, correlation analysis, model calculation, collaborative early warning, visualization, and API interface integration. It embeds the core model from the aforementioned embodiments to achieve collaborative early warning of abnormal energy consumption and accelerated value loss, forming a collaborative closed loop of "data-model-early warning-control." Specific functions are as follows:

[0334] Data cleaning and standardization: Encrypted data transmitted at the network layer is decrypted and then deep-cleaned to remove outliers and fill in missing data, ensuring data quality and providing high-quality data support for subsequent deep fusion, correlation analysis, and collaborative early warning. The specific processing is as follows, which differs from the shortcomings of existing systems with simple cleaning logic and poor data quality:

[0335] Outlier Removal: The 3σ criterion is used to further remove latent outliers that were not detected by edge computing. The detection formula is as follows. Compared with the single threshold detection of existing systems, this method can more accurately identify latent anomalies, significantly improve data quality, and avoid the impact of latent outliers on the accuracy of subsequent analysis.

[0336]

[0337] Parameter description: This is the mean of this type of data. The standard deviation of this type of data. For the decrypted and fused data, abnormal data is directly removed to avoid affecting the accuracy of subsequent analysis, effectively solving the problem of existing systems being unable to identify hidden anomalies.

[0338] Missing data imputation: Linear interpolation is used to imput missing data. The imputation formula is as follows, consistent with the missing value handling methods in Examples 2 and 3, to ensure data integrity. At the same time, the interpolation algorithm is optimized to improve imputation accuracy and effectively solve the problem of large missing data imputation errors in existing systems.

[0339]

[0340] Parameter description: For the missing data after filling, , These are the valid data adjacent to the missing time t. , These represent the times corresponding to valid data.

[0341] Data standardization: The cleaned data is standardized to a uniform scale, consistent with the standardization logic in step one of Example 2, providing a unified data dimension for deep data fusion and correlation analysis. At the same time, a standardization verification step is added to ensure data consistency and provide a guarantee for the efficient operation of subsequent deep fusion algorithms.

[0342] Deep Fusion of Multi-Source Heterogeneous Data: Based on data cleaning and standardization, an innovative improved weighted fusion algorithm is adopted. Combined with the energy consumption-value coupling correlation mechanism of Example 1, the cleaned multi-source data (energy consumption, operation, value, environment) is deeply fused to obtain comprehensive asset energy consumption-value characteristic data. The fusion formula is as follows. Unlike the shortcomings of existing systems that simply splice data and have low fusion degree, this method highlights the synergy of data fusion, significantly improves the accuracy of subsequent early warning and analysis, and provides core data support for collaborative early warning.

[0343]

[0344] Parameter description: This is a deeply integrated set of asset energy consumption and value characteristic data. These correspond to energy consumption data, operational data, value data, and environmental data, respectively. Weights are assigned to various types of data (calibrated by AHP and adjusted in conjunction with the coupling correlation coefficients in Implementation Example 1, such as a weight of 0.35 for value data, 0.35 for energy consumption data, 0.2 for operational data, and 0.1 for environmental data). For the standardized results of the i-th type of data, The correlation coefficient between the i-th type of data and the energy consumption-value collaborative goal (derived from the coupling correlation mechanism in Example 1, with a value range of [0,1], the higher the correlation coefficient, the greater the impact of the data on the collaborative goal); This fusion algorithm breaks through the limitations of the existing simple weighted fusion, integrates the energy consumption-value coupling correlation logic into the fusion process, makes the fused data more targeted, and provides accurate support for collaborative early warning.

[0345] Model embedding and data correlation analysis: Based on the comprehensive feature data after deep fusion, the core model of the aforementioned embodiments is embedded to achieve deep collaboration between "model-data-analysis". This differs from existing systems that lack model support and have weak correlation analysis. Correlation analysis is performed on the deeply fused data to provide core judgment criteria for subsequent collaborative early warning. Specifically, this includes:

[0346] (1) Embed the dynamic self-correcting coupled quantization model of Example 2, call the model to process the deep fusion data, and output the asset value loss rate. Data such as energy consumption-value correlation coefficient are completely consistent with the quantification logic of Example 2, effectively solving the problem that the value loss of the existing system cannot be accurately predicted, and providing data support for accelerating the judgment of value loss in collaborative early warning.

[0347] (2) Embedding the dual-objective optimization model of Example 3, using the deeply fused data as the model input, updating the dual-objective optimization solution set in real time, providing an optimization benchmark for collaborative early warning, realizing the linkage of "early warning-optimization", breaking through the limitation of the existing system's disconnect between early warning and optimization, and ensuring that control measures can be quickly connected after early warning;

[0348] (3) Energy consumption-value correlation analysis: Based on the coupling correlation mechanism of Example 1, the real-time correlation degree between energy consumption and value loss in the deep fusion data is calculated. The correlation formula is as follows, used to determine the coordinated changing trend of energy consumption and value loss. This distinguishes it from existing systems that lack correlation analysis and cannot achieve coordinated early warning, providing a core basis for the coordinated early warning judgment:

[0349]

[0350] Parameter description: The real-time correlation between energy consumption and value loss at time t (range [-1, 1]). Energy consumption at time t With value loss covariance, , Let be the variances of energy consumption and value loss at time t, respectively. This indicates a strong positive correlation between energy consumption and value loss, meaning that increased energy consumption leads to accelerated value loss, providing a core basis for collaborative early warning.

[0351] Collaborative Early Warning System for Anomalies in Energy Consumption and Accelerated Value Loss: Based on deeply fused data, correlation analysis results, and the output of embedded models, an innovative collaborative early warning mechanism is constructed to achieve simultaneous and coordinated early warning of anomalies in energy consumption and accelerated value loss, overcoming the limitations of traditional single early warning systems. Details are as follows:

[0352] (1) Warning threshold setting: Combining the energy consumption quota and value loss standard of the aforementioned embodiments, two types of warning thresholds (normal, warning, and emergency) are innovatively set. The threshold calibration formula is as follows, ensuring that the threshold is consistent with the parameters of the previous embodiments. At the same time, the threshold can be dynamically adjusted according to the asset status, which is different from the defects of fixed thresholds and poor adaptability of existing systems, and ensures the accuracy of the warning threshold:

[0353] ... (a)

[0354] ... (b)

[0355] Parameter description: Equation (a) is the energy consumption early warning threshold. Energy consumption warning threshold (below this value is normal, above this value is below this value) As an early warning, higher than (For emergency) The energy consumption quota is as specified in GB40879-2021 (consistent with Examples 2 and 3); Equation (b) is the value loss acceleration warning threshold. Value loss warning threshold (below this value is normal, above this value is below this value) As an early warning, higher than (For emergency) The benchmark value loss rate is the output of the quantitative model in Example 2; the threshold can be dynamically updated according to the asset's operating status and policy adjustments to improve the accuracy of early warning.

[0356] (1) Collaborative early warning judgment logic: The collaborative early warning judgment formula is constructed by combining energy consumption anomalies, accelerated value loss and their correlation, as shown in Equation (51), and outputs the early warning level (normal, energy consumption early warning, value early warning, collaborative emergency early warning). This is different from the defects of the existing system's single-dimensional early warning and lack of collaborative judgment, and realizes the accuracy and linkage of early warning, effectively identifying the collaborative anomalies of energy consumption and value loss.

[0357]

[0358] Parameter description: The warning levels are categorized as follows: 0 = Normal, 1 = Energy consumption warning, 2 = Value warning, 3 = Collaborative emergency warning. Let t be the real-time energy consumption. Let t be the rate of value loss. Let t represent the correlation between energy consumption and value loss. When the warning level is 3, it indicates that abnormal energy consumption and accelerated value loss occur together, and control measures need to be taken immediately. This addresses the core pain point of the existing system being unable to identify abnormal coordination and having a delayed warning.

[0359] Early warning push: Based on the early warning level, a "multi-channel push + tiered handling" model is adopted. Early warning information is pushed to operation and maintenance management personnel and asset management personnel through various methods such as SMS, system pop-ups, and APP push. The early warning information includes the early warning level, abnormal asset ID, abnormal data, correlation analysis results, and preliminary control suggestions (such as energy consumption early warning suggesting adjustment of energy consumption parameters, and collaborative emergency early warning suggesting shutdown and inspection). This ensures that the early warning is timely and actionable, which is different from the shortcomings of the existing system that has a single push channel and no handling suggestions, and achieves seamless connection between early warning and control.

[0360] Visualization: To facilitate managers' intuitive understanding of system operation status, asset energy consumption and value changes, and early warning information, an innovative "3D visualization + dynamic linkage" control interface has been built. This interface displays the dynamic correlation between asset energy consumption and value, multi-source data fusion results, early warning information, and model output data in real time. This differs from existing systems that suffer from low visualization levels and lack of linked displays. Specific display content includes:

[0361] (1) Data visualization: The real-time energy consumption, historical energy consumption, and value change trend of assets are displayed in line charts and bar charts, and the distribution of energy consumption-value correlation is displayed in heat maps. Data drill-down is supported, and detailed data of single assets and multiple assets can be viewed, which makes it easier for managers to accurately grasp the asset operation status.

[0362] (2) Early warning visualization: Different early warning levels are identified by red (emergency), yellow (early warning), and green (normal), and the location of early warning assets and the type of anomaly are displayed in real time. Early warning information can be handled with one click, improving operation and maintenance efficiency and enabling rapid response to early warning.

[0363] (3) Model visualization: Display the running status, parameter settings and output results of the models in Examples 2 and 3, so that managers can keep track of the model running status in real time, manually adjust the model parameters, improve system adaptability, and support the flexible implementation of the previous invention points;

[0364] (4) Asset visualization: The distribution of public institution assets is displayed in a 3D model. Clicking on an asset allows you to view its detailed energy consumption, value, operation data and early warning records, realizing the full life cycle visualization management of assets and solving the problem of insufficient visualization of asset management in the existing system.

[0365] Standard API Interface Adaptation: To support seamless integration between the system and existing systems of public institutions and reduce deployment costs, a standardized RESTful API interface is innovatively provided. This interface supports data integration with existing asset management platforms and energy consumption monitoring systems of public institutions, enabling data sharing and functional linkage. The interface data format is compatible with existing systems, and the integration formula is shown in equation (52). This approach overcomes the shortcomings of existing system interfaces being non-standard and difficult to integrate, and can quickly adapt to the existing systems of different types of public institutions.

[0366]

[0367] Parameter description: Output data to the interface. This is a data format conversion function. To deeply integrate data, For early warning information, To embed the model output data, Standardize the API interface for existing systems to ensure interface compatibility, enabling rapid integration with existing systems in different types of public institutions such as centralized office areas, hospitals, and universities, thereby reducing system deployment costs.

[0368] Module M44: Platform Layer System Adaptation and Operation Management Module

[0369] This module inherits the core functions of the platform layer, enabling multi-scenario adaptation, dynamic parameter updates, and daily operation and maintenance management of the system. Unlike existing systems that suffer from poor adaptability, require manual parameter adjustments, and are complex to maintain, this module ensures long-term stable operation and adapts to the management needs of different types of public institutions. It also closely connects to the dynamic self-correction mechanism of the aforementioned embodiments, providing a guarantee for the continued implementation of the preceding invention points and the core functions of this system. Specific functions are as follows:

[0370] Multi-scenario Adaptability: Based on the asset characteristics and energy needs of different types of public institutions (centralized office areas, hospitals, universities), an innovative "scenario-adaptive adjustment" mode is adopted to dynamically adjust system parameters (such as monitoring unit sampling frequency, early warning threshold, and data weight) to adapt to the control requirements of different scenarios. The adjustment formula is as follows, which solves the problem of poor adaptability of the existing system and its inability to meet the needs of multiple scenarios, ensuring that the system can operate efficiently in different scenarios:

[0371]

[0372] Parameter description: For the adjusted system parameters, These are the system baseline parameters (consistent with those in the aforementioned embodiments). For scene adaptation coefficient (central office area) ,Hospital , colleges and universities This ensures that the system adapts to the actual needs of different scenarios without the need for manual adjustments, thus improving system adaptability.

[0373] Dynamic parameter updates: Combining the dynamic self-correction mechanism of Example 2, the system parameters are updated automatically. When the asset topology changes, policies are adjusted, or data is updated, the system parameters (such as early warning thresholds, data weights, and model parameters) are updated automatically. The update logic is consistent with the self-correction logic in step S25 of Example 2, ensuring that the system parameters are synchronized with the asset status and policy requirements without manual adjustment. This solves the defects of existing system parameters being fixed and unable to adapt dynamically, ensuring the long-term adaptability of the system and supporting the continuous implementation of the previous invention points.

[0374] Routine Operation and Maintenance Management: This innovative system integrates "system monitoring – fault diagnosis – log management – ​​access control" into a unified operation and maintenance function. This overcomes the shortcomings of existing systems' fragmented and inefficient operation and maintenance functions, ensuring long-term stable system operation. Details are as follows:

[0375] (1) System status monitoring: Real-time monitoring of the operating status of the sensing layer monitoring unit, network layer smart gateway, and platform layer core module. When a fault occurs (such as sensor offline or gateway failure), it triggers equipment fault warning and pushes it to maintenance personnel for timely handling. At the same time, it provides fault location function to improve fault handling efficiency and ensure efficient collaborative operation of all modules of the system.

[0376] (2) Fault diagnosis: The fault tree analysis method is used to accurately diagnose the cause of the fault and output detailed fault handling suggestions to ensure that the fault is resolved quickly, ensure the continuous operation of the system, and solve the problems of difficult fault diagnosis and low handling efficiency in the existing system.

[0377] (3) Log management: Comprehensively record system operation logs, data collection logs, early warning logs, and interface connection logs. The log retention period is ≥1 year, which facilitates subsequent traceability and auditing. At the same time, it supports log retrieval and analysis, providing data support for system optimization and further improving system performance;

[0378] (4) Access control: Set different levels of user permissions (administrator, maintenance personnel, ordinary users). Users with different permissions can view and operate different system functions to ensure system security and controllability. At the same time, it supports dynamic adjustment of permissions to adapt to the management needs of different public institutions and ensure system data security and operation standards.

[0379] The energy consumption and value loss collaborative monitoring and early warning system based on multi-source heterogeneous data fusion provided in this embodiment is designed with "three-layer architecture collaborative innovation + multi-source data deep fusion improvement + original energy consumption-value collaborative early warning". Connecting with the aforementioned embodiments, it provides comprehensive intelligent support for their implementation, mainly achieving:

[0380] First, it breaks through the technical limitations of existing public institution energy consumption and value monitoring systems, and innovatively designs a three-layer architecture of "differentiated intelligent deployment at the perception layer, integrated processing at the network layer, and collaborative early warning and control at the platform layer" to achieve full-scenario collection, efficient transmission, and deep integration of multi-source heterogeneous data; it has created a unique collaborative early warning mechanism for energy consumption anomalies and accelerated value loss, which is different from the single energy consumption early warning or single value monitoring mode of existing systems; it provides standardized API interfaces and multi-scenario adaptive adaptation capabilities to solve the problems of non-standard interfaces and poor adaptability of existing systems.

[0381] Second, addressing the core pain points of existing systems such as scattered data, poor protocol compatibility, single early warning, lack of collaborative support capabilities, and complex operation and maintenance, a non-obvious technical solution is proposed: deeply integrating multi-source heterogeneous data fusion technology with energy consumption-value correlation mechanisms, dynamic self-correcting quantification models, and dual-objective optimization methods; innovatively designing improved data fusion algorithms and collaborative early warning judgment logic to achieve a complete closed loop of "data collection-transmission-fusion-analysis-early warning-control"; reducing system processing pressure and improving response speed through edge computing and cloud collaboration; and ensuring long-term stable operation of the system through dynamic parameter updates and multi-scenario adaptive adaptation.

[0382] This system can be directly applied to various public institutions to achieve real-time monitoring, accurate early warning, and collaborative management of asset energy consumption and value loss. Compared with existing systems, the data collection coverage is increased to 100%, the monitoring lag time is shortened to ≤5 minutes, and the early warning accuracy is increased to ≥95%. It effectively solves the problems of data barriers, incomplete monitoring, and untimely early warning in traditional systems. At the same time, it supports the implementation of all the aforementioned inventions and provides reliable intelligent support for the collaborative management of asset energy consumption and value in public institutions.

[0383] Example verification and comparison experiments

[0384] To verify the technical effect of the present invention and demonstrate that the present invention has significant progress compared with the prior art, a complete comparative verification experiment was carried out in a typical scenario. The experimental design and results are as follows:

[0385] (1) Experimental Samples and Scenarios

[0386] Three typical public institution scenarios were selected, with one benchmark sample chosen for each scenario, as follows:

[0387] Scenario 1: Provincial Government Central Office Area: Includes 12 office buildings, 45 official vehicles, more than 2,800 sets of office equipment and furniture, with a total original value of assets of 1.28 billion yuan;

[0388] Scenario 2: A top-tier general hospital: comprising 8 buildings including outpatient and inpatient buildings, 18 official vehicles, more than 12,000 sets of medical and office equipment, with a total original value of assets of 2.46 billion yuan;

[0389] Scenario 3: Provincial undergraduate universities: including 32 buildings such as teaching buildings and dormitories, 32 official vehicles, more than 21,000 sets of teaching and office equipment, with a total original value of assets of 1.83 billion yuan.

[0390] All samples cover three core asset categories: buildings, official vehicles, and office equipment and furniture. The experiment lasted for 12 consecutive months, and the experimental data came from the actual operating data, financial data, and third-party asset appraisal data of each sample unit.

[0391] (2) Setting up the comparison method

[0392] Four control methods were set up for parallel comparison with the method of this invention, as follows:

[0393] Control group 1 (existing technology 1): Single financial straight-line depreciation method, which is the common method for asset management of existing public institutions. It only calculates the loss of asset value according to the statutory depreciation period and does not consider energy consumption factors.

[0394] Control group 2 (Prior Art 2): The core method of the equipment condition assessment method based on energy data disclosed in CN202411407901.2 only establishes a simple linear relationship between total energy consumption and depreciation period, without a multi-dimensional coupling model;

[0395] Control group 3 (existing technology 3): Single LSTM prediction model, which only uses time series data to train the LSTM model to predict asset value loss, without topological feature fusion, dynamic correction and policy factor embedding;

[0396] Experimental group (method of the present invention): The complete technical solution of the present invention is a complete system of "multi-dimensional coupling and correlation mechanism + dynamic self-correcting coupling and quantization model + dual-objective collaborative optimization + collaborative early warning".

[0397] (3) Evaluation indicators

[0398] Four core evaluation indicators were selected to comprehensively verify the technical effects of the present invention, as follows:

[0399] Value depreciation prediction deviation rate: The relative deviation between the asset value depreciation predicted by the model and the actual value measured by a third-party asset appraisal agency. The calculation formula is as follows:

[0400]

[0401] The lower the deviation rate, the higher the quantization accuracy.

[0402] Prediction accuracy improvement rate: The decrease in prediction deviation rate of the method of this invention compared to the control group is calculated by the following formula:

[0403]

[0404] Annual energy saving rate: The percentage decrease in total annual energy consumption of assets after adopting the collaborative management and control strategy of this invention;

[0405] Annual asset value preservation rate: The ratio of the remaining value of an asset at the end of the year to its initial value at the beginning of the year. The higher the preservation rate, the better the asset value protection effect.

[0406] (4) Experimental results and analysis

[0407] The average experimental results for the three scenarios are shown in the table below:

[0408]

[0409] (5) Experimental conclusions

[0410] Significantly improved quantification accuracy: The average deviation rate of value loss prediction using the method of this invention is only 4.28%, which is far lower than the 12%-19% of the existing technology. Compared with the best control group 3 of the existing technology, the prediction accuracy is improved by 77.02%, realizing the accurate quantification of asset value loss and solving the core pain point of large quantification error in the existing technology.

[0411] Synergistic improvement of energy saving and asset preservation: The method of this invention achieves an annual energy saving rate of 12.35% while increasing the annual asset value preservation rate to 98.47%, truly realizing the synergistic optimization of the dual objectives of energy saving and asset value preservation, and solving the common industry problem of existing technologies that "single energy saving leads to accelerated asset depreciation and single value preservation leads to excessive energy consumption."

[0412] The technical effects are reproducible: This experiment achieved consistently excellent results in three typical public institution scenarios: government agencies, hospitals, and universities, proving that the technical solution of this invention has good scenario adaptability and scalability, and its practicality fully meets the requirements for invention patent authorization.

[0413] This invention provides a method for analyzing the correlation mechanism between energy consumption and value loss of public institution assets, as well as a collaborative monitoring and early warning system. It overcomes the technical pain points in existing public institution asset management, such as the separation of energy consumption and value loss, inaccurate monitoring, untimely early warning, and lack of collaborative management. Through collaborative innovation in four embodiments, a complete technical system of "correlation mechanism - quantitative model - optimization method - implementation system" is constructed, providing reliable technical support for the collaborative management of energy consumption and value of public institution assets. Specific beneficial effects are as follows:

[0414] Regarding the construction of the correlation mechanism, a multi-dimensional coupling correlation mechanism between energy consumption and value is proposed, which breaks through the limitations of traditional single-dimensional correlation analysis. It clarifies the intrinsic correlation between multiple factors such as energy consumption, operation, maintenance, and environment and asset value loss, quantifies the influence weight of each factor on value loss, and solves the problems of unclear correlation between energy consumption and value loss and lack of unified analysis standards in existing technologies. This lays the core theoretical foundation for subsequent quantitative model construction, optimization method design, and system implementation, and improves the scientificity and accuracy of asset correlation analysis.

[0415] In terms of quantifying value loss, a dynamic self-correcting coupled quantitative model is constructed, which introduces dynamic factors such as asset operating status and maintenance data. This enables real-time and accurate quantification and dynamic correction of the asset value loss rate, overcoming the shortcomings of traditional quantitative models that are static, have large errors, and have poor adaptability. It can accurately reflect the true state of asset value loss under different operating scenarios, providing accurate quantitative basis for subsequent early warning and optimization, and improving the accuracy and reliability of value loss prediction.

[0416] In terms of energy consumption and value synergistic optimization, a dual-objective synergistic optimization method of energy consumption and value is constructed, taking into account both energy saving and value loss control objectives. Through constraint setting and optimization solution, the optimal energy use and maintenance management strategy is output, which solves the problem that existing technologies that only pursue energy saving or value preservation cannot achieve synergistic optimization. It realizes the unity of economy and rationality in the management of public institution assets and reduces asset operating costs.

[0417] In terms of technology application, a collaborative monitoring and early warning system integrating multi-source heterogeneous data has been constructed. An innovative three-layer architecture of "perception layer - network layer - platform layer" has been designed to achieve full-scenario collection, efficient transmission, and deep integration of multi-source data. A unique collaborative early warning mechanism for accelerated energy consumption anomalies and value loss has been created. At the same time, it has the ability to adapt to multiple scenarios and connect with multiple systems, solving the pain points of existing systems such as data fragmentation, protocol incompatibility, single early warning, and inability to support the implementation of theoretical technologies. It has realized a complete closed loop of "data collection - transmission - integration - analysis - early warning - control" and improved the level of intelligent management of public institution assets.

[0418] The four embodiments are not isolated, but interconnected and supportive of each other, forming a complete technical chain of "theory-model-method-system": Embodiment 1 provides the core theoretical foundation for the entire invention; Embodiment 2 is a quantitative implementation tool for the correlation mechanism; Embodiment 3 is an optimization method based on the quantitative results; and Embodiment 4 integrates and implements the technical achievements of the first three embodiments, realizing the transformation of theoretical technology into practical application, ensuring that each inventive point supports each other and works together to maximize the technical value of the invention.

[0419] In summary, through the collaborative innovation of four embodiments, this invention effectively solves the core technical pain points in the existing management of energy consumption and value loss of public institution assets, such as unclear correlation, inaccurate quantification, lack of coordinated optimization, and difficulty in implementation. It significantly improves the scientific, accurate, and intelligent level of public institution asset management, reduces asset operating costs, and achieves coordinated development of energy conservation and value preservation. At the same time, it is adaptable to the management needs of different types of public institutions, has good practicality and scalability, fully complies with the requirements for invention patent authorization, and can provide reliable technical support for the full life cycle management of public institution assets.

[0420] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for linking energy consumption and value loss of public institution assets, characterized in that, Includes the following steps: S1. After standardizing and classifying the assets of public institutions, data is collected, and energy consumption factor set and value loss factor set are constructed based on the collected data. Energy consumption factors, operation status factors, maintenance factors and environmental factors are identified as related influencing factors, thereby constructing a multi-dimensional coupling and correlation mechanism between the asset energy consumption and value loss of public institutions. S2. Based on the multi-dimensional coupling and correlation mechanism, a dynamic self-correcting coupling quantitative model is constructed. This model is based on the three-level asset topology of "public institutions → buildings → core assets", integrates topological features and temporal features, and embeds dynamic correction factors and fixed asset accelerated depreciation policy factors to quantify the value loss of single assets and multi-asset linkage. S3. Based on the quantification results of the aforementioned dynamic self-correcting coupled quantification model, construct an energy consumption-value dual-objective collaborative optimization model with the objectives of maximizing energy saving rate and maximizing asset value preservation rate, and solve to generate a collaborative management and control strategy. S4. Based on the aforementioned collaborative management and control strategy, trigger and execute a collaborative early warning for abnormal energy consumption and accelerated value loss.

2. The analytical method according to claim 1, characterized in that: The standardized classification of public institution assets mentioned in step S1 includes buildings, official vehicles, and office equipment and furniture; the energy consumption factor set includes real-time energy consumption, historical energy consumption, HVAC energy consumption, photovoltaic self-consumption rate, vehicle energy consumption per 100 kilometers, and equipment standby energy consumption; the value loss factor set includes initial asset value, natural depreciation rate, physical attenuation, annual maintenance investment, and annual energy-saving renovation investment; the operating status factors include operating time, operating temperature, and vibration frequency; and the environmental factors include ambient temperature and ambient humidity.

3. The analytical method according to claim 1, characterized in that, The dynamic self-correcting coupled quantization model described in step S2 is implemented in the following way: S21. Based on the natural decay law of assets, an environmental correction factor is introduced to construct a dynamic benchmark model for the natural value loss of assets. S22. Calculate the additional attenuation coefficient driven by energy consumption and the attenuation offset coefficient caused by operation and maintenance transformation, based on the differences in asset types. S23. Based on the aforementioned benchmark model, additional attenuation coefficient, and attenuation offset coefficient, construct a multi-dimensional coupled correlation model of "natural benchmark loss + energy consumption-driven incremental loss − operation and maintenance transformation offset loss" to complete the basic quantification of asset value loss.

4. The analytical method according to claim 3, characterized in that: The formula for the dynamic benchmark model of natural asset value depreciation is as follows: ； The formula for the multi-dimensional coupling and correlation model is: ； In the formula, The natural baseline residual value of the asset in year t; The initial value of the asset; The annual natural decay coefficient; This represents the number of years the device has been used. This is an environmental correction factor, with a value range of 0.92 to 1.05; This represents the actual residual value of the asset in year t. For the t-th year of use, the annual additional depreciation factor of the asset driven by the deviation of actual energy consumption from the benchmark energy consumption; For the t-th year of use, the annual depreciation offset coefficient caused by asset maintenance and energy-saving renovation investment; This is actual energy consumption data; For annual maintenance costs; Investment in annual energy-saving renovations.

5. The analytical method according to claim 3 or 4, characterized in that: Annual attenuation offset factor The calculation formula is as follows: In the formula , To maintain the benefit weighting coefficient of investment and energy-saving renovation investment, and to meet the following requirements: ; The actual annual maintenance expenditure for the asset is expressed in yuan. The annual benchmark maintenance investment amount for similar assets shall be uniformly verified by the administrative affairs management department, and the unit is yuan; The amount of annual special energy-saving renovation investment for assets is expressed in yuan. The energy-saving renovation benefit coefficient is used to quantify the sustainability of the renovation effect. The value ranges from 0.8 to 1.

2. The better the renovation effect, the larger the δ value.

6. The analytical method according to claim 3 or 4, characterized in that: The annual additional depreciation coefficient for assets is calculated differently according to asset classification, and an energy consumption deviation threshold is introduced to avoid abnormal energy consumption data from affecting the calculation accuracy. (1) The annual additional depreciation coefficient for building construction assets is calculated as follows: In the formula: , , The weighting coefficients for each energy consumption factor in building categories are determined using the Analytic Hierarchy Process (AHP) and satisfy the following conditions: ; The actual annual energy consumption of the air conditioning and heating system is expressed in kWh. The annual benchmark energy consumption of the air conditioning and heating system of public buildings of the same type and area is determined according to the building orientation, area and insulation level. This represents the building's actual annual energy consumption for regular electricity use, expressed in kWh. The benchmark annual energy consumption for regular electricity use in public buildings of the same type and area; The self-consumption rate of the photovoltaic system is dimensionless and ranges from 0 to 1. The threshold function ensures that the additional attenuation coefficient is non-negative, avoiding the unreasonable situation where the value loss decreases when the energy consumption is lower than the benchmark. (2) The annual additional depreciation coefficient for official vehicle assets is calculated as follows: In the formula: , , The weighting coefficients for each energy consumption factor of official vehicles satisfy the following conditions: ; This represents the vehicle's actual annual fuel / electricity consumption per 100 kilometers. The fuel / electricity consumption per 100 kilometers is the benchmark fuel / electricity consumption set by the Ministry of Industry and Information Technology for the same model of vehicle. The vehicle's actual annual mileage is expressed in km. The annual benchmark mileage for official vehicles as approved by the administrative affairs management department; This refers to the vehicle's actual annual idling energy consumption. The annual idle speed benchmark energy consumption for the same model of vehicle is provided by the vehicle manufacturer as the standard value. (3) The annual depreciation coefficient for office equipment and furniture assets is calculated as follows: In the formula: , , As the weighting coefficient, the electronic device satisfies Office furniture ; The actual annual standby energy consumption of the equipment / furniture, in kWh; The annual standby baseline energy consumption for similar equipment / furniture; The actual annual rated operating energy consumption of electronic equipment is expressed in kWh. This is the annual rated operating condition benchmark energy consumption for similar electronic equipment.

7. The analytical method according to claim 1, characterized in that: The inter-node association weights in the three-level asset topology map described in step S2 are determined by weighting the energy consumption correlation, associated usage frequency, and maintenance linkage frequency of the assets corresponding to each node, as shown in the formula: ； The value loss rate of the dynamic self-correcting coupled quantization model is obtained by dynamically correcting the benchmark value loss rate through the associated influencing factors, and the calculation formula is as follows: ； The formula for calculating the accelerated depreciation policy factor for fixed assets is as follows: ； In the formula, The association weight between node i and node j, with a value range of [0,1]. , , Let be the weighting coefficient, satisfying ; Let Pearson correlation coefficient be the energy consumption of the assets corresponding to the two nodes, with a value range of [-1, 1]. This represents the dynamically adjusted rate of asset value loss at time t. The base value loss rate at time t; This is a correction factor, with a value range of [0.01, 0.05]. Let the weight of the i-th type of influencing factor satisfy the following condition: ; This represents the standardized result for the i-th factor. This is a policy factor, with a value range of [1, 1.3]. For policy coefficients; This is the policy adaptation coefficient.

8. The analytical method according to claim 1, characterized in that: The objective function of the energy consumption-value dual-objective collaborative optimization model described in step S3 is a combination of the energy saving rate function and the asset value preservation rate function; the dual-objective optimization function is specifically as follows: ， ； In the formula, The annual energy saving rate of the asset. The total annual energy consumption of the assets before optimization is expressed in kWh. The total annual energy consumption of the asset after optimization is expressed in kWh. For the annual asset value preservation rate, This represents the initial annual value of the asset, in yuan. The optimized year-end asset value is expressed in yuan. , To optimize the asset value loss at time t; The constraints of the model include energy consumption constraints, asset operation constraints, and policy constraints; the solution is obtained by using the non-dominated sorting genetic algorithm NSGA-II to obtain the Pareto optimal solution set, and the collaborative management strategy is generated from the solution set.

9. The analytical method according to claim 1, characterized in that, The step of generating the collaborative management and control strategy in step S3 further includes: S31. Cluster assets based on energy consumption level and value loss rate, classify them into different types, and generate differentiated collaborative management and control strategies for each type of asset. S32. Build a digital twin virtual image and simulate and verify the generated collaborative management and control strategy; S33. Construct a collaborative evaluation system to evaluate the effectiveness of the strategy in real time, and automatically update the collaborative control strategy when the evaluation is unqualified, the asset topology changes, or the simulation results deviate from the physical execution.

10. The analytical method according to claim 1, characterized in that: The collaborative early warning in step S4 adopts a multi-level early warning mechanism; the early warning level is determined comprehensively based on real-time energy consumption, value loss rate, and the real-time correlation between energy consumption and value loss; the real-time correlation is determined by the covariance of energy consumption and value loss and their respective variances.

11. The analytical method according to claim 10, characterized in that, The formula for determining the warning level is: In the formula, The warning levels are categorized as follows: 0 = Normal, 1 = Energy consumption warning, 2 = Value warning, 3 = Collaborative emergency warning. , Energy consumption early warning threshold; , Accelerate the early warning threshold for value loss; Let t be the real-time energy consumption. Let be the rate of value loss at time t. The correlation between energy consumption and value loss at time t is calculated using the following formula: in, The real-time correlation between energy consumption and value loss at time t, with a value range of [-1, 1]; Energy consumption at time t With value loss covariance; , Let be the variances of energy consumption and value loss at time t, respectively.

12. A coordinated early warning system for energy consumption and value loss of public institution assets, used to implement the method as described in any one of claims 1-11, characterized in that, It includes a perception layer, a network layer, and a platform layer that work together, among which: The perception layer is used to deploy multiple types of intelligent monitoring units to collect multi-source heterogeneous data related to asset energy consumption, operation, maintenance, environment and value, and to perform preliminary verification of the collected data. The network layer is used to deploy smart gateways to perform multi-protocol conversion, data fusion, edge computing, breakpoint resume, and encrypted transmission on data from the perception layer. The platform layer is used to integrate data cleaning, correlation analysis, model calculation, collaborative early warning, and visual management and control functions.

13. The system according to claim 12, characterized in that: The data fusion at the network layer includes preliminary fusion and deep fusion; The preliminary fusion involves weighted summation of multi-dimensional normalized data for the same asset to obtain comprehensive asset status data; the preliminary fusion formula is: ; The deep fusion involves weighting and fusing different categories of asset data with their correlation coefficients to the energy consumption-value synergy target, resulting in comprehensive asset energy consumption-value characteristic data; the deep fusion formula is: In the formula, This refers to the overall asset status data after initial edge integration; For data dimensions, The weights of the i-th class of data; This represents the normalized result for the i-th type of data; This is a deeply integrated set of asset energy consumption and value characteristic data. These correspond to energy consumption data, operational data, value data, and environmental data, respectively. For the standardized results of the i-th type of data, Let be the correlation coefficient between the i-th type of data and the energy consumption-value synergistic target; The platform layer also includes a scenario adaptive adaptation unit and a parameter dynamic update unit; the scenario adaptive adaptation unit is used to adapt to centralized office areas, hospitals, and university scenarios; the trigger conditions for the parameter dynamic update unit include prediction deviation exceeding the standard, asset topology changes, policy adjustments, and periodic data updates.

Images