Target user-oriented comprehensive energy technology matching degree intelligent recommendation system

Abstract

The invention discloses a target user-oriented comprehensive energy technology matching degree intelligent recommendation system, and particularly relates to the technical field of comprehensive energy technology matching. The system firstly constructs a double-clock data base with a unified time base and a fixed step length, governs late and additionally recorded data and generates an evidence pointer; consistency and traceability of measurement calibers of electricity, heat, cold, gas, steam and the like are ensured. And then performing same-caliber linkage with exogenous factors such as meteorology, carbon factors and policies on the quality-weighted user portraits to form a scene library which can be used for future deduction. Generating candidate schemes based on a technical library, executing coarse-to-fine system-level feasibility check of multi-energy coupling and engineering implementation constraints, and performing quasi-dynamic verification in a prediction window to obtain a feasible scheme set and binding a key state quantity interval of the feasible scheme set with evidence; and finally, calculating a comprehensive matching degree and a confidence interval by using a multi-objective function of scene probability and risk penalty, and outputting a scheme and a strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical matching field of comprehensive energy technology, more specifically, the present application relates to a target user-oriented comprehensive energy technology matching degree intelligent recommendation system. BACKGROUND

[0002] When the comprehensive energy system faces the park, building group and industrial scene, it often involves multiple energy forms such as electricity, heat, cold, gas and steam. The existing scheme is mostly centered on single energy or static design working condition for capacity configuration and technology comparison and selection, which is difficult to express the measurement caliber, time step and physical constraint relationship of multi-energy in a unified caliber, resulting in difficulty in direct comparison and reuse of cross-modal data. There are problems such as data delay, supplementary recording and inconsistent caliber in engineering field, and there is a lack of mechanism for constructing unified time base and time grid from event time / processing time double clock, and there is also a lack of explicit labeling and evidence-based tracing of unavailable / inferred data segments, which affects the reliability of subsequent evaluation.

[0003] In terms of demand side image and exogenous scene, the traditional method often separates the statistical profile of historical load and future meteorology, institutional factors for modeling, which is difficult to complete consistent caliber linkage deduction on the same time skeleton; and the data confidence is not transmitted to the prediction uncertainty. In the generation and checking level of candidate technical scheme, the existing technology lacks unified feasible domain expression and convergent checking process for the coordination constraints of electric / thermal / gas network, the operation feasible domain of multi-energy coupling device and engineering implementation constraints.

[0004] In the evaluation and recommendation link, the common practice focuses on a single target, lacks a multi-objective comprehensive matching degree model combining quality weighting, scene probability and tail risk penalty, and it is difficult to connect the network / engineering safety boundary and data evidence quality to the final score and recommendation result. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a target user-oriented comprehensive energy technology matching degree intelligent recommendation system to solve the problems proposed in the background art.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme: The target user-oriented comprehensive energy technology matching degree intelligent recommendation system comprises the following modules: The application boundary unification and data base construction module is used for constructing a unified time base and fixed step event time and processing time double clock data base, governing delayed and supplementary recording data and generating evidence pointers, and outputting a standardized multi-energy data view containing availability bitmap, inference bitmap and data alignment quality score; A user portrait construction and exogenous scenario modeling module is configured to perform quality-weighted feature extraction based on a standardized multi-energy data view, construct a target user portrait, and link meteorological, carbon factors, and policy systems under a unified standard to form a scenario library for future deduction; A candidate scheme generation and system-level feasibility checking module is configured to generate candidate schemes and perform multi-energy coupling and engineering implementation constraint from coarse to fine convergence system-level feasibility checking, perform quasi-dynamic verification within a prediction window, output a feasible scheme set and its key state quantity interval, and bind with evidence; A multi-objective matching degree evaluation and landing closed loop module is configured to perform multi-objective comprehensive evaluation on the feasible scheme set according to scenario occurrence probability and introduce risk penalty, calculate comprehensive matching degree and its confidence interval, and output a landable scheme and strategy.

[0007] In a preferred embodiment, the application boundary unification and data base construction module adopts a double clock data organization model with a unified time base and a fixed step length, sets a buffer window with a limited waiting time for late data, and generates an evidence pointer with uniqueness and traceability for overdue arrival of supplementary data, the evidence pointer including data source identification, observation time, processing time, and version identification.

[0008] In a preferred embodiment, the user portrait construction and exogenous scenario modeling module calculates sample weights based on data alignment quality score and importance score, and performs weighted aggregation on multi-modal measurement values of electricity, heat, cold, gas, and steam to generate a quality-weighted user energy consumption portrait.

[0009] In a preferred embodiment, the importance score is composed of state weight component, statistical significance weight component, and event-driven weight component; the state weight component is used to represent the relative importance of the target user's energy consumption behavior at a specific time in the system running state, its value is determined by querying the built-in business rule library according to the pre-set load period division rule and the real-time device start-stop and working condition switching signals obtained from the monitoring system; the statistical significance weight component reflects the deviation degree of the current observation data relative to the historical same period benchmark running mode, its value is obtained by comparing the real-time data with the historical data profile and performing nonlinear conversion on the calculated deviation degree; the event-driven weight component is triggered by external input of strategic or environmental events, the system maintains an external event list containing event type, effective time, and pre-set importance level, and obtains the component value according to the highest level of active events in the list at the current time.

[0010] In a preferred embodiment, the candidate scheme generation and system-level feasibility checking module adopts a coarse-to-fine convergent checking process, sequentially performing preliminary screening based on preset rules and engineering experience boundaries, system-level energy conservation and equipment capacity matching checking, and collaborative simulation checking of multi-energy coupling of electricity, heat, and gas.

[0011] In a preferred embodiment, when performing collaborative simulation checking of multi-energy coupling of electricity, heat, and gas pipe networks, the system introduces a fast proxy model calibrated offline through high-fidelity simulation data to approximately evaluate key network state quantities and their out-of-limit risks; for operation points whose evaluation results by the proxy model are close to or touch safety barriers, falling back to a detailed simulation model based on physical principles for secondary confirmation.

[0012] In a preferred embodiment, the candidate scheme generation and system-level feasibility checking module introduces a safety margin function to tighten physical constraints based on data alignment quality scores, converting data uncertainty into engineering safety margins.

[0013] In a preferred embodiment, the multi-objective matching degree evaluation and landing closed-loop module performs quality-weighted aggregation of the operation performance in the future prediction window on a unified time base, calculating the quality-weighted aggregated amount of each evaluation index under a specific exogenous scenario.

[0014] In a preferred embodiment, the multi-objective matching degree evaluation and landing closed-loop module combines scenario occurrence probability and risk penalty terms to calculate the comprehensive matching degree score of the candidate scheme; wherein the risk penalty term is used to quantify the weakening effect of adverse tail scenarios on the robustness of the scheme, and embeds a safety margin equivalent penalty determined by data alignment quality.

[0015] In a preferred embodiment, the multi-objective matching degree evaluation and landing closed-loop module performs intervalized calculation of the resource and environmental benefits of the candidate scheme throughout its life cycle, outputs interval estimation results of key environmental indicators such as total carbon emission reduction and primary energy saving amount, and includes them in the final risk summary report.

[0016] Technical effects and advantages of the present application: The present application constructs a unified time base and fixed step event time / processing time dual clock data base, introduces an unusable / inferred state bitmap and data alignment quality score, realizes unified and evidence-based traceability of multi-energy data such as electricity, heat, cold, gas, and steam, and forms a quality-weighted energy consumption portrait for target users. On this basis, meteorological, carbon factors, and policy systems are linked as an exogenous scenario library on the same time skeleton, from the source, inhibiting the conduction of data uncertainty on features and predictions, significantly improving the comparability and transferability of the portrait and the credibility and reproducibility of scenario deduction.

[0017] Further, the application adopts a convergent system-level feasibility check from coarse to fine, performs quasi-dynamic verification of physical constraints and engineering implementation boundaries of a multi-energy coupling network in a prediction window of a unified time base, and converts data quality into constraint tightening on the engineering level by a safety margin function; combines a hierarchical verification mechanism of a fast proxy model and physical simulation to improve scheme screening efficiency while ensuring safety. Finally, through a multi-objective comprehensive matching degree model coupled with scenario probability and risk penalty, the score with confidence interval and the environmental benefit interval result in the whole life cycle are output, and the version information of data, model and threshold is fixed with an evidence pointer full-link, so as to realize auditable, reproducible and landable scheme recommendation and closed-loop delivery. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to facilitate the understanding of those skilled in the art, the application will be further described below in conjunction with the drawings; Figure 1 The structure diagram of the comprehensive energy technology matching degree intelligent recommendation system for the target user of the application is shown in Figure 2 The user portrait construction and exogenous scenario modeling process diagram is shown in DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the application will be described clearly and completely below in conjunction with the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application. EMBODIMENT

[0020] The comprehensive energy technology matching degree intelligent recommendation system for the target user of the application, as shown in Figure 1 includes the following modules: Module 1: application boundary unification and data base construction module: The application boundary unification and data base construction module is used to establish a unified calculation boundary and construct a high-reliability multi-energy flow data base. First, the energy categories, measurement caliber and core evaluation indexes covered by the system are defined; the evaluation index system takes energy efficiency and carbon emission as the core, including power, energy flow rate, carbon emission intensity and energy grade. Then, the multi-energy flow data base is constructed to provide data basis for subsequent evaluation and recommendation.

[0021] The system is designed and adopts a unified time base , fixed step The time organization model of double clock. All observation data from field meters, controllers and edge gateways are synchronized to record the event time generated by the data source end and the processing time received by the system platform when entering the system. The data is mapped to a unified time grid in subsequent pipeline processing. The event time is used for timing alignment and grid operation, and the processing time is used for identifying delayed data, data link tracing and system operation audit.

[0022] The system uses interval difference method for cumulative readings and maps them to equivalent average power or equivalent average flow to ensure the consistency of statistical indicators within the system. Taking the electric energy meter as an example, on the unified time base , the equivalent average power at the center time t of the time grid is calculated by the following formula: In the formula, represents the cumulative value of the electric energy meter reading up to time t; represents the cumulative value of the reading at the previous time interval point of time t; is the unified target sampling step set by the system, for example, 5 minutes or 15 minutes; is the equivalent average power calculated in the time interval .

[0023] The system configures a delay data buffer window with limited waiting time on the data access layer of the edge side and the center platform side. Data samples arriving within the preset window period will be backfilled to the corresponding historical time grid according to their own event time; samples arriving beyond the buffer window period will be introduced into a special data processing bypass. For each data entering the system through the data processing bypass, the system will generate an evidence pointer with uniqueness and traceability. The information contained in the evidence pointer at least includes: the unique identifier of the data source device, the accurate pairing record of the event time and the platform arrival time, the software version information of the data acquisition module and the forwarding link, and the data batch fingerprint generated by the platform. The preset window period corresponds to the waiting time of the delay data buffer window, and the preset window period is integer times, for example, 5 times .

[0024] After the time alignment and statistical caliber unification of multi-source data, the system explicitly labels the data availability state and data inference state for each time grid unit, so as to avoid confusion between data fragments that cannot be directly calculated or are only inferred by algorithms and real physical observation data in the case of missing upstream data collection segments, abnormal meter reading back jumps, device replacement or manual intervention supplement, etc. Specifically, the system maintains two key state bitmaps: the unavailable bitmap is used to mark the case that the corresponding index in the time grid cannot be directly calculated due to missing source data or logical verification conflicts; the inference bitmap is used to mark that the value in the time grid is a derived value obtained by a specific interpolation algorithm or state estimation model. The above two types of state labels are accompanied by a record of the cause summary and the path identification of the processing algorithm adopted, and are associated with the aforementioned evidence pointer, so that any advanced derived conclusion based on this data base can be traced back to the original data and all processing steps experienced.

[0025] At the same time, the system maintains a dynamically updated data alignment quality score for each time series data , which is used to comprehensively quantify the overall condition of the current sequence in terms of data completeness, statistical confidence and time series stability. The score is obtained by a weighted combination model in numerical value, specifically represented as: . Wherein, is the data availability at time t, and its calculation method is: in the time window centered at t, the ratio of the number of available data points to the total number of data points; is the data confidence at time t, which is determined by the device accuracy level of the data source and the reasonableness verification result of the current instantaneous value, and the calculation formula is , can be determined by comparing the current value with its historical quantile (such as 5% to 95% interval), if it falls within the interval, it is 1, otherwise it is linearly attenuated according to the degree of deviation; is the normalized time series drift measure, and its calculation method is: calculate the absolute value of the first order difference between the current value and the previous time step value , and divide it by the standard deviation of the time series in the past N time steps, that is ; is the quality weight coefficient fixed in the data view at system initialization, for example .

[0026] The module introduces the physical topological relationship of energy networks such as electricity, heat, gas and steam in the park into the data consistency calibration process in the parallel process of generating unified time series. Within the set time window scale, the system automatically checks the source-sink balance relationship of the energy system, the input-output feasible region of the key energy conversion equipment and the basic energy conservation constraint. When detecting that there is a non-negligible inconsistency between the cross-modal data, the system corrects the identified local abnormal data under the constraint or gives a weight processing in the subsequent calculation according to the principle of minimum change. Through the synergistic effect of the three core technologies of time caliber unification, delayed data governance and physical consistency calibration, the system finally outputs a standardized multi-energy data view that integrates energy efficiency and carbon emission information for efficient computing design. The data view takes a unified time base and a fixed step size as the skeleton, and completely contains the observed value set after standardization and robustness processing, the data availability and inferred state marker set generated by the system, the dynamically calculated data alignment quality score set, and the evidence pointer set that can be traced throughout the process. This standardized data view will serve as the only reliable data source for subsequent implementation of user portrait modeling, exogenous scenario generation, candidate technical solution feasibility review and multi-objective matching degree evaluation, so that the system described in the application has a solid foundation capability of auditability, reproducibility and quantifiability from the beginning of data access in the engineering site deployment.

[0027] Module two: user portrait construction and exogenous scenario modeling module The user portrait construction and exogenous scenario modeling module, as shown in Figure 2 , is used to construct a fine portrait around the energy behavior characteristics and equipment operation features of the target user after the data caliber unification, delayed data governance and physical consistency calibration described in module one are completed, and a standardized multi-energy data view under a unified time base and a target step size is formed. Subsequently, the system extends the constructed user portrait to the future on the time axis, and couples it with external driving factors such as meteorology, carbon emission factors and policies, to form a consistent and future-declinable exogenous scenario library.

[0028] To ensure that the constructed user portrait has comparability and migratability for subsequent technology recommendation, the system uses the availability bitmap, inferred bitmap and data alignment quality score output in the data base as the core confidence basis for weighted feature extraction of the multi-energy sequence. This mechanism ensures that high-quality observation data contribute more weight to portrait formation, while low-quality or inferred data samples are naturally down-weighted, thereby suppressing the disturbance of data uncertainty on portrait accuracy at the feature level. Specifically, the system uses a unified time base On a discrete grid, a feature vector x(t) containing multimodal measurements such as electricity, heat, cold, gas, and steam is constructed for each time t. Simultaneously, the system defines the sample weights for that time moment. The calculation formula is as follows: ,in ∈[0.1,0.3] is the quality rejection threshold, for example, it can be taken as... =0.2; λ is the trade-off between control multiplication coordination and linear harmonicity, for example, λ=0.5; δ is the adjustment of the preference for reliability and importance, for example, δ can be 0.6 to 0.8 during the debugging period, and δ can be 0.3 to 0.5 during the stabilization period; ≥1 is used to sharpen high-weight samples and suppress edge samples, for example It equals 1 or 1.5. When At that time, no matter how important the business is, it will not be included in the training / evaluation, ensuring that reliability comes first and importance comes later.

[0029] The importance score is calculated as follows: ; in, , , For example, weighting coefficients. = =0.4, =0.2. All components are robustly statistically normalized to [0,1] within the same window. State weight components. This component represents the relative importance of a target user's energy consumption behavior at a specific moment within the system's operational status. Its value is determined based on pre-defined load period division rules and real-time equipment start-up / shutdown and operating condition switching signals obtained from the monitoring system, and is comprehensively determined by querying the built-in business rule base. (Statistical significance weight component) This reflects the degree of deviation of current observed data from the historical baseline operating mode. Its value is obtained by comparing real-time data with historical data profiles and performing a nonlinear transformation on the calculated deviation, aiming to highlight operational inflection points and suppress the interference of extreme outliers. Event-driven weighted components. It is triggered by external strategic or environmental events. The system maintains a list of external events that includes event type, effective time and preset importance level, and directly obtains the component value by matching the highest level of the active event in the list at the current time.

[0030] Based on these sample weights, the system performs robust weighted aggregation of profile features. For a specific feature that needs to be aggregated, its weighted aggregation value is... Through formula The calculation yields a sum of weighted eigenvalues ​​in the numerator and a sum of weights in the denominator. The summation operation is performed in a unified time base. with a step size on a discrete grid.

[0031] In the process of image generation, the system does not completely replace the physical law with a pure statistical method, but within the framework of the energy network topology and the physical feasible region of the park, the multi-modal benchmark energy profile of electricity, heat, gas, steam, etc. and the static information from the production process side, building envelope side, etc. are physically consistent constraints. For users with obvious periodic working condition switching or beat operation, the system uses event annotations recorded under a unified time base (such as device start-stop signals) and device account information as anchor points to clearly distinguish the steady-state operation section and the dynamic transition section on the time axis, so as to ensure that the power spikes at the device start-stop moment, the planned maintenance window or the transient process caused by abnormal energy return will not be mistakenly confused as the inherent long-term energy consumption rule of the user. In the scene of cold start of the system or new user access with insufficient historical data, the system maps the prior knowledge of the user's industry classification, production scale, process temperature control demand and building thermal parameters into the structured label space of the image, and takes the typical prior profile of the same industry or the same type of image group as the initial reference template. With the continuous arrival of new data, the system will gradually update the initial template according to the sample weight of the foregoing.

[0032] After completing the image construction of the user's historical and current state, the image is extended forward on the time axis to construct a future exogenous scenario that exactly matches it. Gather meteorological reanalysis data and observation data from local micro meteorological stations to generate short-term and medium-term prediction sequences of key meteorological variables including temperature, humidity, solar irradiance, wind speed, etc. At the same time, the system will project the institutional factors that may affect the user's energy structure and carbon emissions, such as government regulations, carbon emission factor updates, technology standard evolution, etc. onto the time grid of the unified time base to form an institutional scenario time axis that is completely consistent with the data caliber of the image.

[0033] All exogenous variables need to be finally checked for consistency of dimension and unit with the energy profile of the user image before entering the scenario library.

[0034] In the specific implementation of exogenous scenario generation, the system introduces intervalization and probability distribution characterization methods to address the inherent uncertainty of meteorological conditions and institutional factors. The system generates a limited number of representative scenario branches to cover normal operation, conservative boundary conditions, and extreme disturbance scenarios that may occur in the future. Each scenario branch records its data source, generation algorithm, confidence range of key variables, and applicable future time period, and establishes a one-to-one mapping relationship with the specific version of the user profile based on it. For industries or specific production processes that are particularly sensitive to changes in exogenous variables, the system introduces calibrated elasticity parameters at the linkage level of profiles and scenarios. These elasticity parameters enable the predicted changes in temperature, humidity, irradiance, and institutional factors to be mapped to the expected deformation of the user's future load profile and carbon emission profile through a quantitative relationship constrained by historical data regression analysis and expert experience, thereby preserving the user's rational adjustment space under different external environments. The parameter values of this elasticity mapping relationship are derived from the comprehensive calibration of historical data regression, physical constraints, and technical curves provided by equipment manufacturers, and their version information is also fixed in the generated scenario objects through evidence pointers.

[0035] When the user profile library and the exogenous scenario library are both completed, the system forms a complete profile and scenario combination view on their Cartesian product. Based on this combination view, and according to data quality weights and external uncertainties, the system generates multiple energy load trajectories and their corresponding confidence interval bands within the future prediction window for each profile and scenario combination. The confidence interval band here is consistent in semantics and dimension with the data alignment quality score in module one and the sample weight in module two , achieving continuous information transmission from bottom-layer data reliability to upper-layer prediction uncertainty. On this combination view, the system further calculates key operating boundary indicators related to equipment rated capacity, typical operating period, allowed start-stop frequency, system thermal inertia constraints, etc., to provide direct and consumable standardized inputs for candidate technology solutions in module three and system-level engineering feasibility checks. By then, this module outputs a user profile and exogenous scenario system that is tightly coupled with the bottom-layer data foundation, consistent with physical operating constraints, and fully aligned with external climate and institutional boundaries. This system can be quickly constructed and deployed in a cold start scenario, and can be smoothly and controlledly updated as data evidence quality accumulates and external environments change during the long-term operation of the system, thereby providing a stable, transparent, and auditable prior basis for subsequent technology configuration screening, network feasibility verification, and multi-objective matching degree evaluation.

[0036] Module three: candidate solution generation and system-level feasibility check module: The candidate scheme generation and system-level feasibility checking module is used to complete the user portrait and exogenous scenario construction described in module two, and obtain the portrait and scenario combination view based on the unified time base with the target step size After the portrait and scenario combination view, the system enters the candidate technology scheme generation and system-level feasibility checking phase. The core of this module is to integrate the user-side energy use profile, device asset account, and site boundary conditions with the future climate, carbon emission factors, and institutional rules on the time axis, and perform feasibility verification in the dimensions of device configuration, network access, and engineering implementation.

[0037] The system first solidifies the performance parameters of candidate technologies such as photovoltaic power generation systems, electrochemical energy storage systems, heat pumps, combined heat and power systems, gas boilers, and waste heat recovery devices into a calculable technology library. Each device or technical unit in the technology library is abstracted as a feasible operating interval under given external working conditions and internal load distribution, which is jointly defined by its rated capacity, partial load efficiency characteristics, start-stop ramp rate, device life cycle, maintenance cycle, noise and vibration limits, etc. At the same time, the system unifies the engineering constraints such as building structural capacity, roof and site available area, ventilation and heat dissipation conditions, safety retreat distance, equipment hoisting channel, grid-connected point electrical location, secondary system interface, and approval process milestones into three types of boundary conditions: space, time, and institution, and calls them engineering feasible regions. Based on the technology library and engineering feasible regions, the system forms the decision vector θ of the configuration scheme, which covers device selection, installation capacity and quantity, energy coupling and access system topology, key degrees of freedom of operation strategy, and output upper and lower limits on the unified time base The system ensures that all static parameters and dynamic decision variables have consistent physical dimensions and measurement relationships, thereby ensuring that the generated configuration scheme has mapping and verification capabilities at the engineering and network levels.

[0038] The generation of candidate solutions employs a coarse-to-fine convergent process. First, based on pre-defined rules and engineering experience boundaries, the system quickly eliminates obviously infeasible or technically inadequate combinations, completing the initial screening. Next, the system selects representative time periods from the profile and scenario combination views to perform rapid system-level energy conservation and equipment capacity matching checks, excluding solutions that cannot meet users' basic energy supply needs and comfort constraints under peak load, extreme minimum ambient temperature, or the most stringent regulatory constraints. Solutions that pass the above checks will proceed to a more refined simulation stage of the multi-energy coupled system of electricity, heat, and gas. In this stage, the system integrates the power flow distribution, voltage exceedance risk, line losses, reactive power balance, and power factor constraints of the power distribution network with the supply and return water pressure difference, temperature parameters, circulating pump head, and building thermal inertia response characteristics of the heating network, as well as the pipe segment pressure, medium flow rate, and valve group regulation capacity of the gas pipeline network, into a unified system-level model for collaborative verification. For application scenarios requiring rapid assessment of the feasibility of numerous solutions, the system introduces a fast surrogate model obtained through offline calibration using high-fidelity simulation data, while strictly preserving the first-order physical principles. This surrogate model can provide approximate evaluation results of key network state variables and their risk of exceeding limits within milliseconds. Each surrogate model is equipped with clearly defined prediction error boundaries and safety barriers within its applicable operating conditions. Any surrogate model evaluation result that approaches or touches the operating point of the safety barrier must be reverted to a detailed simulation model based on physical principles for secondary verification before proceeding to the next stage. This ensures a balance between speed and safety in the solution selection process. The safety barrier is the combination of prediction error boundaries allowed by the surrogate model within the calibration conditions; exceeding these boundaries forces a forced reversion to physical simulation.

[0039] To prevent the data quality uncertainties described in Module 1 from being indiscriminately transmitted to the solution verification stage and leading to misjudgments of optimistic feasibility, this module combines the usability bitmap, inference bitmap, and data alignment quality score output by the data base. A constraint tightening mechanism is introduced. Specifically, the physical constraints of the network and devices are expressed as inequalities in a general form: Where x(t) is the system state vector at time t, containing physical quantities such as node voltage, branch power, pipeline pressure, and temperature; θ is the decision vector for candidate configurations, representing equipment capacity, quantity, etc. The input is the exogenous scenario at time t, such as ambient temperature and carbon emission factor. The data supporting this constraint calculation is then used to determine the alignment quality score. When the energy balance residual is low, or large, the system operates with a safety margin function. The original constraints are conservatively modified, and the criterion is replaced with: In this criterion, This indicates the data alignment quality score; (·) is a monotonically non-decreasing non-negative function, which specifically means that the lower the given data alignment quality score, the larger the safety margin imposed, and vice versa. Thus, the uncertainty at the data level is explicitly translated into tighter feasible region at the engineering level. This constraint tightening process is consistent across all network and equipment constraints, so that the causality between data reliability and engineering safety is uniformly reflected at the model level, and can be restored through evidence pointers at the time of audit (·) Function version and parameter source.

[0040] Safety margin function is used to map the data alignment quality score into a non-negative safety margin value. The specific implementation of this function includes but is not limited to the following embodiments: Linear mapping embodiment: = a*(1 - Q), where a is a margin reference coefficient greater than zero. This coefficient is related to the specific network safety operation requirements, for example, for power distribution networks sensitive to voltage fluctuations, a can take a value of 5% to 10% of the voltage limit; for thermal networks with large thermal inertia, a can take a value of 2% to 5% of the temperature limit. The lower the Q value, the larger the safety margin imposed. .

[0041] The system-level check not only covers static design conditions, but also performs quasi-dynamic operation simulation within the prediction window W(t) on a unified time base: for energy storage devices and thermal systems with significant thermal inertia, the consistency of energy balance and state evolution is checked by time step, ensuring that the minimum start-stop time, maximum climbing ability, cycle depth and indoor thermal comfort boundary are met simultaneously within the entire window; any violation at any time is determined to be infeasible. At the same time, the progress and approval rhythm related to engineering implementation is structured into time constraints and injected into the decision space θ, including key equipment delivery cycle, construction allowed downtime window, and inherent rhythm of grid connection and approval; if an uncoordinated time conflict is formed on the critical path, it is marked as infeasible and removed. To ensure traceability and reproducibility, the simulation model version, boundary conditions, input data snapshot and exogenous scenario number are fixed at each time of discrimination, and evidence pointers are archived.

[0042] When the above coarse-to-fine convergence process is completed, the system will obtain a set of candidate solutions that are judged to be feasible in the three dimensions of network physical constraints, equipment operability and engineering implementation conditions. Each feasible solution in the set is bound to its applicable specific profile and scenario combination, data alignment quality score and safety margin function based on which the constraint is tightened, discrete selection results of key equipment and corresponding performance curve range, and the interval of key state variables obtained through joint verification of proxy model and physical simulation.

[0043] Module four: multi-objective matching degree evaluation and landing closed loop module: The multi-objective matching degree evaluation and landing closed loop module is used after the candidate scheme set is obtained through system-level feasibility checking in module three. The system enters the multi-objective matching degree evaluation and landing closed loop stage. This module combines the image generated in module two and the scene combination view on the time skeleton of the unified time base and the target step , and performs multi-dimensional and quantitative evaluation on the running performance of each candidate scheme in the future prediction window, and explicitly solidifies the influence of data quality uncertainty and engineering risk in the evaluation results, so that the comprehensive matching degree score output finally reflects the engineering feasibility, information credibility and risk exposure level of the scheme. To ensure the consistency of evaluation range and traceability of whole-chain calculation, the system binds the availability bitmap, inference bitmap, data alignment quality score and evidence pointer from the data base with the configuration vector θ of the candidate scheme before entering this module, so that the formation of any evaluation index can be traced back to the original data, calculation model version and threshold setting.

[0044] In the specific index calculation process, the system establishes quantitative indexes that can be normalized to the same evaluation scale around the four core objectives of carbon intensity, primary energy utilization rate, comfort maintenance ability and operation reliability. To avoid the disturbance of low credibility data samples in the future prediction window to the final conclusion, the system performs quality weighted aggregation on the time series performance according to time. Let represent the instantaneous contribution value of candidate scheme θ to the kth evaluation index (such as instantaneous carbon emission intensity or instantaneous energy efficiency) at time t under exogenous scene s. The system uses the same time weight function w(t) defined in module two. Thus, the quality weighted aggregation quantity of the kth index under scene s is , and the calculation formula is: The numerator of this formula is the weighted instantaneous index sum, and the denominator is the weight sum. The summation operation is performed on the unified time base of the evaluation window on the discrete grid, and the calculation result is consistent with the physical dimension of the original index . Based on this aggregation quantity, the system maps the indexes of different dimensions to the dimensionless standard domain that can be compared and weighted according to the predetermined rules.

[0045] To give a robust comprehensive judgment under the condition of exogenous scene uncertainty and tail risk, the system introduces probability weight and risk penalty at the scene level, and integrates the indexes under each scene into a single comprehensive matching degree score. Let be the preference weight of the kth normalized index, which reflects the relative importance of each core objective to the decision maker; is the occurrence probability or representative weight of scenario s, used to depict the likelihood of different exogenous scenarios; is the normalized value of the kth indicator under scenario s after weighted aggregation; is the confidence level is the conditional risk value penalty term for performance deficiency or constraint violation of scheme θ under the condition The comprehensive matching degree score of candidate scheme θ can be expressed as: In this scoring function, the first term represents the weighted expectation of scheme multi-objective performance under different exogenous scenarios, reflecting the baseline performance of the scheme; the second term is a penalty term for quantifying the weakening effect of adverse tail scenarios on the robustness of the scheme. When the quality of historical data supporting the evaluation is low or the uncertainty of future prediction is high, the system automatically adjusts the value of risk aversion coefficient η within the preset range. To avoid the overestimation of the running points near the safety fence in module three in the evaluation, the system uses the safety margin function with tightened constraints to embed the calculation of in an equivalent penalty manner, so that the network physical constraints and the safety margins at the engineering implementation level can be directly reflected in the comprehensive matching degree score . The final output of is archived together with its confidence interval calculated based on scenario fluctuations, and a bidirectional reference relationship is established with all the corresponding evidence pointers, ensuring that the scoring result has the legal and engineering properties of auditable and reproducible.

[0046] While completing the multi-objective scoring, the system also performs intervalized calculation of the resource and environmental benefits of the candidate scheme throughout the life cycle. The implicit carbon emissions in the device manufacturing and recycling stages, energy consumption and direct carbon emissions in the operation stage, as well as the carbon emission reduction brought by energy efficiency improvement, are projected onto the time axis of the project evaluation period to form a carbon footprint and resource consumption trajectory unique to each scheme. Under the constraints of the device output and utilization hour sequence derived from the portrait and scenario views, the system calculates the interval estimation results of key environmental indicators such as total carbon emission reduction and primary energy saving of the scheme. To ensure consistency between the environmental benefit conclusion and external uncertainty, the system applies single-factor and multi-factor linked sensitivity perturbation to sensitive parameters such as carbon emission factors and key device performance degradation rates, and includes the changes in environmental indicators after perturbation and the changes in comprehensive matching degree score in the final risk summary report, ensuring that the final system recommendation reflects the consistency of excellent multi-objective evaluation and robust life cycle environmental benefits.

[0047] When the comprehensive matching degree score and the environment benefit estimation form a closed evidence chain, the system parameterizes the recommended scheme to meet all the hard engineering constraints, the comprehensive matching degree score The optimal scheme with the environment benefit index at the leading level is the criterion to determine the final priority recommended scheme. The system parameterizes the recommended scheme in the energy management system or the microgrid control platform to generate specific device control set values and operation strategies. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0048] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be realized in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0049] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place or distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0050] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0051] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A comprehensive energy technology matching intelligent recommendation system for target users, characterized in that: Includes the following modules: The application boundary unification and data foundation construction module is used to build a dual-clock data foundation with a unified time base and fixed step size for event time and processing time, to manage late and supplementary data and generate evidence pointers, and to output a standardized multi-functional data view that includes availability bitmap, inference bitmap and data alignment quality score; The user profile building and exogenous scenario modeling module is used to extract quality-weighted features based on standardized multi-energy data views, build target user profiles, and link meteorological, carbon factors and policies and systems under a unified standard to form a scenario library for future extrapolation. The candidate solution generation and system-level feasibility verification module is used to generate candidate solutions and perform a coarse-to-fine convergent system-level feasibility verification of multi-energy coupling and engineering implementation constraints. It performs quasi-dynamic verification within the prediction window, outputs a set of feasible solutions and their key state variable intervals, and binds them with evidence. The multi-objective matching degree assessment and implementation closed-loop module is used to conduct a multi-objective comprehensive evaluation of the set of feasible solutions based on the probability of scenario occurrence and the introduction of risk penalties, calculate the comprehensive matching degree and its confidence interval, and output feasible solutions and strategies.

2. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 1, characterized in that: The application boundary unification and data foundation construction module adopts a dual-clock data organization model with a unified time base and fixed step size. It sets a buffer window with a limited waiting time for late data and generates unique and traceable evidence pointers for supplementary data that arrives late. The evidence pointers include data source identifier, observation time, processing time and version identifier.

3. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 1, characterized in that: The user profile building and exogenous scenario modeling module calculates sample weights based on data alignment quality scores and importance scores, and then uses these weights to aggregate multimodal measurements of electricity, heat, cold, gas and steam to generate quality-weighted user energy consumption profiles.

4. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 3, characterized in that: The importance score is composed of a weighted combination of state weight components, statistical significance weight components, and event-driven weight components. The state weight component is used to characterize the relative importance of the target user's energy consumption behavior at a specific moment in the system's operating state. Its value is determined by querying the built-in business rule base based on the pre-set load period division rules and the equipment start-up, shutdown, and operating condition switching signals obtained in real time from the monitoring system. The statistical significance weight component reflects the degree of deviation of the current observed data from the historical benchmark operating mode. Its value is obtained by comparing the real-time data with the historical data profile and performing a nonlinear transformation on the calculated deviation. Event-driven weighted components are triggered by externally input strategic or environmental events. The system maintains a list of external events that include event type, effective time, and preset importance level, and obtains the component value by matching the highest level of the active event in the list at the current time.

5. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 1, characterized in that: The candidate scheme generation and system-level feasibility verification module adopts a coarse-to-fine convergent verification process, which sequentially performs preliminary screening based on preset rules and engineering experience boundaries, system-level energy conservation and equipment capacity matching verification, and collaborative simulation verification of multi-energy coupling of electricity, heat and gas.

6. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 5, characterized in that: When performing collaborative simulation verification of multi-energy coupling of electric, thermal and gas pipeline networks, the system introduces a fast proxy model obtained through offline calibration of high-fidelity simulation data to approximate the evaluation of key network state variables and their risk of exceeding limits. For operating points where the evaluation results of the proxy model are close to or touch the safety barrier, the system falls back to the detailed simulation model based on physical principles for secondary confirmation.

7. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 1, characterized in that: The candidate solution generation and system-level feasibility verification module introduces a safety margin function, which tightens physical constraints based on data alignment quality scores, transforming data uncertainty into engineering safety margin.

8. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 1, characterized in that: The multi-objective matching degree evaluation and implementation closed-loop module performs quality-weighted aggregation of the operational performance within the future prediction window on a unified time base, and calculates the quality-weighted aggregation amount of each evaluation index under specific exogenous scenarios.

9. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 8, characterized in that: The multi-objective matching degree evaluation and implementation closed-loop module combines the probability of scenario occurrence and risk penalty term to calculate the comprehensive matching degree score of candidate solutions. Among them, the risk penalty term is used to quantify the weakening effect of adverse tail scenarios on the robustness of the solution, and embeds the equivalent penalty of safety margin determined by data alignment quality.

10. The intelligent recommendation system for comprehensive energy technology matching degree for target users according to claim 1, characterized in that: The multi-objective matching degree assessment and implementation closed-loop module performs interval-based calculations of the resource and environmental benefits of candidate solutions throughout their entire life cycle, outputs interval estimates of key environmental indicators such as total carbon emission reduction and primary energy savings, and incorporates them into the final risk summary report.

Images