A multi-dimensional analysis and evaluation zero-carbon building method based on BIM technology

Abstract

The application relates to the technical field of zero-carbon building design and evaluation, and discloses a multi-dimensional analysis and evaluation method for zero-carbon buildings based on BIM technology, which comprises the following steps: S1, obtaining an original BIM model, a dynamic time sequence data source and a localized carbon emission factor library, expanding a multi-dimensional evaluation attribute set in the BIM model, and storing data in layers as a static geometry library, a dynamic time sequence library and a localized factor library to generate a BIM unified data center; wherein the dynamic time sequence data source is simulation data or typical meteorological year data in the design stage of a new building, and is IoT real-time monitoring data in the evaluation stage of an existing building. The application forms a complete, self-consistent and engineering-implementable technical link from data integration, model management, optimization decision to result verification, and significantly improves the comprehensiveness, accuracy, robustness and operability of zero-carbon building analysis and evaluation.

Description

Technical Field

[0001] This invention relates to the field of zero-carbon building design and evaluation technology, and in particular to a multi-dimensional analysis and evaluation method for zero-carbon buildings based on BIM technology. Background Technology

[0002] In the construction industry's journey towards carbon neutrality, the design and evaluation of zero-carbon buildings have become crucial technological directions. Currently, the industry typically employs Life Cycle Assessment (LCA) methods, green building evaluation systems (such as LEED and BREEAM), or BIM-based energy simulation tools to analyze building carbon emissions and environmental performance. However, existing technologies have the following shortcomings when applied to the multi-dimensional analysis and evaluation of zero-carbon buildings:

[0003] First, the evaluation dimensions are relatively singular. Most methods only focus on carbon emissions during the operation phase, or only use energy consumption as the core indicator, failing to incorporate multi-dimensional green indicators such as indoor environmental quality, water resource efficiency, material recycling rate, and renewable energy ratio into a unified evaluation framework. This makes it difficult for the analysis results to fully reflect the true zero-carbon performance of buildings. Second, the data silo phenomenon is serious. BIM models, IoT monitoring data, LCA databases, green certification rule tables, etc., belong to different systems, and data from different stages (design, construction, operation and maintenance) cannot be connected, making it difficult to implement full life cycle evaluation. Third, when attempting to integrate multi-source data, new technical contradictions arise: semantic and temporal granularity conflicts between different systems, BIM model volume explosion due to writing back historical data throughout the life cycle, design parameter oscillations caused by automatic optimization closed loops, and lack of universally accepted trade-off criteria for explicit conflicts between multi-dimensional indicators. In addition, existing methods generally lack a closed-loop feedback mechanism between evaluation results and optimization suggestions, and have not formed a systematic error control and result credibility assurance system.

[0004] Therefore, there is an urgent need for a method and system that can systematically solve the above problems and achieve multi-dimensional analysis and evaluation of zero-carbon buildings. Summary of the Invention

[0005] To address the technical problems mentioned in the background section, this invention provides a multi-dimensional analysis and evaluation method for zero-carbon buildings based on BIM technology, comprising the following steps:

[0006] S1. Obtain the original BIM model, dynamic time-series data source, and localized carbon emission factor library. Expand the multi-dimensional evaluation attribute set in the BIM model and store the data in layers as a static geometry library, a dynamic time-series library, and a localized factor library to generate a unified BIM data platform. The dynamic time-series data source is simulation data or typical meteorological year data in the new building design stage and IoT real-time monitoring data in the existing building evaluation stage.

[0007] S2. Based on the BIM unified data platform, the dynamic time series data in the dynamic time series library is aggregated according to the unified evaluation benchmark granularity to obtain aggregated time series data. The low-frequency BIM attributes in the static geometry library are sampled according to the time window to obtain sampled attribute data. The concepts of different systems are aligned through a preset mapping semantic bridge to obtain aligned semantic data. The gaps during the data missing period that do not exceed a preset threshold are interpolated and confidence weights are added to the interpolated data points to generate a multi-source heterogeneous fusion dataset.

[0008] S3. Based on the multi-source heterogeneous fusion dataset, the BIM geometric subject and static attributes are retained in the original model, and the additional historical data of the entire life cycle is stored in a time-series database or object storage. Only a lightweight pointer to the additional data is retained in the BIM model. During evaluation, the corresponding time period data is loaded as needed through the pointer, and the dynamic evaluation view is released after the evaluation is completed, generating a dynamically stripped lightweight BIM evaluation model.

[0009] S4. Based on the dynamically stripped lightweight BIM evaluation model, perform the following operations according to the application scenario:

[0010] If it is in the design phase of a new building, the top N design parameters with the greatest impact on the overall green performance can be identified directly by ranking the single-parameter sensitivity. In each iteration, only one parameter is recommended to be adjusted and at most two alternative adjustment directions are provided.

[0011] If it is in the existing building assessment stage, the lightweight BIM evaluation model is first calibrated using IoT real-time monitoring data, and then the top N renovation parameters with the greatest impact on the overall green performance are identified by single-parameter sensitivity ranking.

[0012] Each adjustment takes effect only after manual confirmation by the user and is locked after it takes effect. The automatic suggestion is terminated when the oscillation detector changes in the opposite direction during multiple consecutive iterations, and a zero-carbon building design scheme or zero-carbon renovation scheme is generated after manual access control iteration optimization.

[0013] S5. Based on the zero-carbon building design scheme or zero-carbon retrofit scheme, a pre-set index conflict matrix is ​​established and the index values ​​of each dimension are calculated. Instead of automatic weighted summation, a Pareto front diagram is generated for the user to manually select a compromise scheme. The algebraic sum of the total carbon emissions and carbon offset of the building throughout its entire life cycle is calculated according to the compromise scheme selected by the user. When the algebraic sum is ≤0, it is determined to be zero carbon. A comprehensive evaluation report containing the compliance degree of multiple dimensions of indicators and the zero carbon determination result is generated.

[0014] Furthermore, S1 expands the multi-dimensional evaluation attribute set in the BIM model, specifically including:

[0015] Obtain the type identification and spatial location information of each component in the original BIM model;

[0016] Based on the component type identifier, query the corresponding evaluation attribute type to be attached to the component from the predefined multi-dimensional attribute mapping table;

[0017] The multi-dimensional evaluation attribute types include at least carbon emission factor attributes, indoor environmental quality index attributes, water resource efficiency attributes, material recycling rate attributes, and renewable energy proportion attributes;

[0018] Create an attribute extension record for each component, and write the default value or calculation expression corresponding to the above evaluation attribute type into the attribute extension record;

[0019] The attribute extension record is associated with the components of the original BIM model and stored to generate the extended BIM model.

[0020] Furthermore, S1 stores data in layers: a static geometry library, a dynamic time-series library, and a localized factor library, specifically including:

[0021] Extract BIM geometric data and component static attribute data from the expanded BIM model, store them in the static geometry library, and save them in a lightweight compression format;

[0022] Time-series data is obtained from a dynamic time-series data source. In the design phase of new buildings, the time-series data consists of simulated energy consumption data, typical meteorological year data, or reference operating data retrieved from a localized factor library. In the assessment phase of existing buildings, the time-series data consists of real-time sensor monitoring data. The real-time monitoring data includes at least energy consumption time-series data, temperature and humidity time-series data, and equipment start-up and shutdown status data. The time-series data is then stored in a dynamic time-series library after being appended with a timestamp and a source identifier.

[0023] Obtain the climate zone and administrative division identifier of the target building, retrieve the carbon emission benchmark factor and green evaluation standard benchmark value that match the climate zone and administrative division from the external data source, and store them in the localized factor library;

[0024] Establish a logical association index between the static geometry library, the dynamic time series library, and the localized factor library. The logical association index shall at least contain the mapping relationship between the globally unique identifier of the component and the corresponding time range.

[0025] Furthermore, the semantic bridge of the preset mapping described in S2 is implemented using a rule engine, including:

[0026] The standard clauses to be transformed are extracted from the standard clause library, and the standard clauses are transformed into structured rules based on regular expressions and keyword matching. Standard traceability information is attached to each rule.

[0027] Construct a configurable mapping table, configure a standard source number for each mapping entry, and use the mapping table as the basis for concept alignment;

[0028] Reject any automatic completion of new concepts; when an unmapped concept is encountered, a dialog box will pop up asking the user to select from the existing library or create it manually.

[0029] Based on the structured rules and the mapping table, concept alignment between different systems is performed to generate traceable semantic alignment results.

[0030] Furthermore, S2 includes error budget management after adding confidence weights to the interpolated data points, specifically including:

[0031] The upper limit of the allowable error for each processing step is preset, and the processing steps include at least data interpolation, time series aggregation, and reduced-order model invocation;

[0032] The percentage of interpolated data in the current evaluation period relative to the total data volume is calculated.

[0033] When the percentage exceeds a preset threshold, the current evaluation result is downgraded to a pre-evaluation and marked as unusable for formal certification;

[0034] The error values ​​of each step are accumulated in real time, the ratio of the accumulated error to the total budget is calculated, and the ratio is displayed on the user interface in real time.

[0035] Furthermore, the aggregated data obtained after time-series aggregation in S2 adopts a physical model-first strategy when performing energy consumption simulation, specifically including:

[0036] The model calls a pre-built white-box physical model, inputs aggregated time-series data, and calculates the energy consumption value predicted by the model.

[0037] The actual energy consumption values ​​for the same time period are extracted from dynamic time-series data sources. The actual energy consumption values ​​are simulation input values ​​in the design stage of new buildings and IoT measured values ​​in the evaluation stage of existing buildings. The residual between the actual energy consumption values ​​and the model prediction values ​​is calculated.

[0038] The residual is compared with a preset residual threshold;

[0039] When the absolute value of the residual exceeds the threshold continuously, an abnormal operation status alarm is generated and the alarm information is pushed to the user terminal.

[0040] It does not automatically correct the physical model parameters, thus maintaining the deterministic calculation logic of the physical model.

[0041] Furthermore, S3 stores the entire lifecycle of attached historical data in a time-series database or object storage, and retains only lightweight pointers in the BIM model, specifically including:

[0042] Identify the components in the BIM model that require full lifecycle traceability and obtain the globally unique identifier of the component;

[0043] Collect additional historical data generated by the component during the design, construction, and existing building assessment phases. The additional historical data includes at least construction logs, equipment maintenance records, energy consumption history data, and material replacement records.

[0044] The additional historical data is written into a time-series database or object storage in chronological order, and a timestamp and data type tag are attached to each data record.

[0045] In the BIM model, a lightweight pointer record is created for the component. The lightweight pointer contains a globally unique identifier for the component, the storage location address of the additional historical data, and the time range covered by the data.

[0046] When the evaluation engine needs to access the full lifecycle data of the component, it retrieves the data for the corresponding time period from external storage on demand through the lightweight pointer. After the evaluation is completed, the evaluation view in memory is released, and the external data is not written back to the BIM model.

[0047] Furthermore, the single-parameter sensitivity ranking in S4 employs a reduced-order model to accelerate calculations, specifically including:

[0048] For specific combinations of building types and climate zones, a training sample set is generated based on a high-precision physical model;

[0049] The training sample set was trained offline using the intrinsic orthogonal decomposition method to obtain a reduced-order model.

[0050] During the online optimization phase, candidate parameter combinations are input into the reduced-order model to quickly evaluate the comprehensive green performance index of each candidate scheme.

[0051] Based on the evaluation results of the order reduction model, all candidate solutions are ranked, and the top three candidate solutions are extracted.

[0052] The top three candidate solutions are then precisely validated using the complete physical model to generate the final parameter recommendation results.

[0053] Furthermore, S5 employs a dual-mode evaluation weighting approach when generating the comprehensive evaluation report, specifically including:

[0054] In authentication mode, the fixed weight configuration table specified by the national standard is forcibly read, the fixed weights are applied to the calculation of indicators of each dimension, and the scoring results that meet the authentication requirements are output.

[0055] In the design decision mode, the user-defined weight configuration is read, the custom weights are applied to the calculation of indicators in each dimension, and the scoring results for internal decision-making are output.

[0056] The evaluation results corresponding to the custom weights are stored separately in the design decision database and are not mixed with the evaluation data in the certification mode;

[0057] When generating the evaluation report, clearly indicate the weighting mode currently in use and its corresponding weight values.

[0058] Furthermore, S5 also includes hash-based notarization of the comprehensive evaluation report, specifically including:

[0059] Read the complete data content of the comprehensive evaluation report generated by S5, and calculate the Merkle root hash value of the data content;

[0060] Submit the Merkle root hash value and the current timestamp together to the timestamp storage service of the National Time Service Center to obtain the storage certificate;

[0061] Only the hash value and the certificate of evidence are put on the blockchain or stored in the notary and notary system, while the original report data is kept in the central database;

[0062] When verification is required, the hash value of the original report is recalculated and compared with the hash value in the evidence storage certificate to generate a result indicating whether the verification passed or failed.

[0063] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0064] This invention constructs a unified BIM data platform for zero-carbon assessment. By extending multi-dimensional attribute sets into the BIM model and combining them with three separate databases (static geometry database, dynamic time-series database, and localized factor database), it fundamentally solves the problems of data silos and stage fragmentation, achieving unified management of full lifecycle data and integrated analysis of multi-dimensional green indicators. Based on this, through a semantic bridge combining rule engines and manual verification, controlled interpolation and confidence level addition mechanisms, and a physical model-prioritized energy consumption simulation strategy, it effectively resolves conflicts in semantics, temporal granularity, and data quality among multi-source heterogeneous data, ensuring the traceability and credibility of the fused data. Simultaneously, by employing dynamic stripping storage and on-demand loading technologies, historical data is decoupled from the BIM model, completely avoiding model size expansion and enabling efficient operation of the full lifecycle assessment under conventional hardware conditions.

[0065] At the optimization level, this invention replaces traditional fully automatic multi-objective optimization with single-parameter sensitivity ranking, manual access control iteration, parameter locking, and oscillation detectors. While ensuring controllable convergence, it empowers designers with final decision-making power, eliminating the parameter sensitivity and uninterpretable nature of optimization algorithms. Furthermore, the Pareto visualization trade-off and dual-mode weighting transparently present the contradictions between multi-dimensional indicators, balancing standardized certification with the personalized design needs of projects. Finally, through error budget management, independent verification datasets, and hash-based evidence storage, a complete credibility assurance system is constructed from three levels: data quality, model accuracy, and the legal validity of the results.

[0066] In summary, this invention forms a complete, self-consistent, and engineerable technical chain from data integration, model management, optimization decision-making to result verification, significantly improving the comprehensiveness, accuracy, robustness, and operability of zero-carbon building analysis and evaluation. Attached Figure Description

[0067] Figure 1 This is a flowchart of the multi-dimensional analysis and evaluation method for zero-carbon buildings based on BIM technology proposed in this invention. Detailed Implementation

[0068] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0069] Below reference Figure 1 This embodiment proposes a multi-dimensional analysis and evaluation method for zero-carbon buildings based on BIM technology, which is implemented as follows:

[0070] It should be noted that this method has two working modes depending on the application scenario:

[0071] New building design optimization mode: At this time, the dynamic time series data source is the simulation data generated by the energy consumption simulation software based on typical meteorological year data. Step S4 directly performs sensitivity analysis and iterative optimization on the original design parameters (such as insulation layer thickness, window-to-wall ratio, equipment selection) and outputs the optimized design scheme.

[0072] Existing building renovation and optimization mode: In this mode, the dynamic time-series data source is the actual operating data collected by IoT sensors deployed within the building. First, these measured data are used to calibrate the lightweight BIM evaluation model generated in step S3 (e.g., adjusting the actual operating parameters in the physical model, correcting the equipment efficiency attenuation coefficient, etc.) so that the deviation between the model's predicted output and the actual energy consumption is less than a preset threshold. Then, based on the calibrated model, sensitivity analysis and iterative optimization are performed on the parameters to be renovated (such as the thickness of the additional insulation layer on the external wall, the COP of replacing with high-efficiency units, increasing photovoltaic capacity, etc.) to output a zero-carbon renovation plan.

[0073] The data processing flow in steps S1-S3 and S5 remains consistent for both modes, with the aforementioned differences only existing in step S4.

[0074] Taking an office building located in a hot-summer, cold-winter region as an example, the building has a BIM model and an IoT monitoring system has been deployed.

[0075] Step S1 proposed in this embodiment includes: acquiring the original BIM model, IoT real-time monitoring data and localized carbon emission factor library, expanding the multi-dimensional evaluation attribute set in the BIM model, and storing the data in layers as a static geometry library, a dynamic time series library and a localized factor library to generate a unified BIM data platform.

[0076] In this embodiment, firstly, the original BIM model of the office building is acquired, using the IFC format… Simultaneously, a dynamic time-series data source is acquired. Taking the existing building assessment phase as an example, this embodiment uses an energy consumption monitoring system, temperature and humidity sensors, and equipment status collectors deployed within the building to acquire continuous real-time IoT monitoring data. This data records power consumption, indoor temperature, and air conditioning start / stop status in each area once per second. It should be noted that if the building is in the design phase, hourly load data generated by energy consumption simulation software (such as EnergyPlus) based on typical meteorological year data is used as the dynamic time-series data source. Furthermore, a pre-built localized carbon emission factor library is loaded. This library stores the average carbon emission factor of the local power grid, the transportation distance of major building materials and their corresponding carbon emission coefficients, and the benchmark thresholds for various green indicators specified in the national "Zero Carbon Building Evaluation Standard," categorized by climate zone (hot summer, cold winter) and the provincial administrative division of the project location.

[0077] After acquiring the above data, the original BIM model is expanded with multi-dimensional evaluation attributes. Specifically, each component in the BIM model is traversed, and its type identifier (e.g., "exterior wall - 200mm concrete") is read. Based on this type identifier, the required evaluation attribute types for the component are queried from a predefined multi-dimensional attribute mapping table. These include: carbon emission factor attributes (used for subsequent calculations of carbon emissions during building material production and transportation), indoor environmental quality index attributes (e.g., heat transfer coefficient, sound insulation level), water resource efficiency attributes (e.g., water-saving appliance flow rate level), material recycling rate attributes (e.g., recycled aggregate ratio), and renewable energy ratio attributes (e.g., photovoltaic panel coverage area). An attribute expansion record is created for each component, and the default values ​​or calculation expressions corresponding to the above attributes (e.g., "carbon emission factor = value of concrete in the local factor library") are written into the record. This record is then associated with the component in the original BIM model for storage, thereby generating the expanded BIM model.

[0078] Subsequently, the data is stored in three logical libraries in a hierarchical manner. Geometric data (mesh, patches) and component static attributes (such as material name and design dimensions) are extracted from the expanded BIM model and stored in the static geometry library, using a lightweight compression format (such as SVF2) to reduce file size. Time-series data is read from dynamic time-series data sources (IoT-collected data in the existing building assessment phase and simulation data in the design phase), and each data point is appended with a precise timestamp and source identifier (e.g., "Source=IoT sensor ELEC_01" or "Source=EnergyPlus simulation"), then written chronologically into the dynamic time-series library, which uses Apache IoTDB to support high compression ratios and fast range queries. Based on the climate zone (hot summer, cold winter) and administrative division (province and city) of the project location, matching carbon emission benchmark factors (e.g., the regional power grid emission factor is 0.58 kgCO2 / kWh) and green evaluation standard benchmark values ​​(e.g., indoor CO2 concentration limit of 1000 ppm) are retrieved from external data sources and stored in the localized factor library. Finally, a logical association index is established between the three libraries: in the static geometry library, a globally unique identifier (GUID) is recorded for each component, and mapping tables with the GUID and time range as keys are established in the dynamic time series library and the localization factor library, respectively, thus forming a unified BIM data platform for subsequent steps. In addition, the system also integrates building material carbon emission inventory data exported from LCA software (such as SimaPro) as supplementary input.

[0079] Through the above operations, a unified BIM data platform with a clear structure and complete data was constructed, laying the foundation for the subsequent integration and evaluation of multi-source heterogeneous data.

[0080] Step S2 proposed in this embodiment includes: based on the BIM unified data platform, high-frequency IoT data in the dynamic time series library is time-series aggregated according to a unified evaluation benchmark granularity to obtain aggregated time series data; low-frequency BIM attributes in the static geometry library are sampled according to a time window to obtain sampled attribute data; concepts of different systems are aligned through a preset mapping semantic bridge to obtain aligned semantic data; and gaps not exceeding a preset threshold during data loss are interpolated and confidence weights are added to the interpolated data points to generate a multi-source heterogeneous fusion dataset.

[0081] In this embodiment, we will continue to use an office building in a hot-summer, cold-winter region as an example for explanation.

[0082] First, time-series aggregation and attribute sampling are performed. A unified evaluation benchmark granularity of 1 hour is set. Time-series data from the past 24 hours are read from the dynamic time-series library. If the building is in the existing building assessment phase, this data is real-time monitoring data collected by IoT sensors, with an original sampling frequency of once per minute; if the building is in the design phase, this data is hourly simulation data generated by energy consumption simulation software (such as EnergyPlus) based on typical meteorological year data. The read data is aggregated according to the unified evaluation benchmark granularity (1 hour): if the original data frequency is higher than the hourly granularity (e.g., minute-level), the arithmetic mean, maximum value, and cumulative start / stop count within that hour are calculated; if the original data is at the hourly granularity, the original values ​​are used directly. This yields the aggregated time-series data. Simultaneously, low-frequency attributes of BIM components, such as the thickness of the external wall insulation layer, window-to-wall ratio, and U-value, are read from the static geometry library. These attributes remain largely unchanged after construction; the latest valid values ​​(e.g., the current design values) are taken according to the time window to obtain the sampled attribute data.

[0083] Secondly, semantic alignment is performed. For data from different systems such as BIM, IoT, localized factor libraries (including carbon emission benchmarks and green evaluation standards), and LCA (Life Cycle Assessment) tools, a rule engine is used to achieve pre-defined mappings. Specifically, clauses to be transformed are extracted from national and industry standard clause libraries (such as the "Zero Carbon Building Evaluation Standard" GB / T5xxxx), for example, "Indoor CO2 concentration should not exceed 1000ppm". Based on regular expressions (such as matching the keyword "CO2 concentration") and keyword matching, this clause is transformed into a structured rule: {Indicator:"IAQ_CO2", Threshold:1000, Unit:"ppm", Comparison operator: "<="}, and each rule is appended with standard traceability information (standard number and clause number). Simultaneously, a configurable mapping table is constructed, configuring a standard source number for each mapping entry. For example, mapping "VAV-BOX-01" in BIM to "TerminalUnit_5" in the IoT system, and indicating that the mapping is based on the design drawing number. The system explicitly rejects any automatic replacement of entirely new concepts. When encountering an unmapped concept (such as a newly introduced sensor model), a dialog box pops up asking the user to select from the existing library or manually create a mapping. Based on the above structured rules and mapping table, concept alignment between different systems is completed, generating traceable semantic alignment results.

[0084] Next, data missing data handling and confidence level assignment are performed. For data segments in the dynamic time series database with consecutive missing durations not exceeding 2 hours, linear interpolation is used to fill in the missing data. For example, if energy consumption data for a certain hour is missing, the average of the values ​​before and after the missing hour is taken as the interpolation value, and a confidence level weight is assigned to this interpolation point. The weight calculation formula is: confidence level = 1 - (missing duration / 2 hours) × 0.2, meaning the shorter the missing time, the higher the confidence level. For windows with missing data exceeding 2 hours, they are directly marked as "data unreliable" and do not participate in subsequent weighted evaluation. Simultaneously, error budget management is implemented: preset upper limits for allowable errors in each processing step are set, such as interpolation error ≤ 2%, time series aggregation error ≤ 1%, and reduced-order model call error ≤ 5%. It should be noted that the "reduced-order model" described in this embodiment is a mathematical proxy model, used to quickly approximate the evaluation of building energy consumption or overall performance while ensuring a certain level of computational accuracy, replacing computationally intensive high-fidelity physical models (such as EnergyPlus). This reduced-order model is not a fixed model commonly used in this field, but rather a model pre-built offline for the building type (office building) and climate zone (hot summer and cold winter) of this embodiment. Its construction method and error control are as follows.

[0085] This embodiment uses a pre-built reduced-order model to accelerate the evaluation. This reduced-order model is built offline for the building type (office building) and climate zone (hot summer, cold winter) of this embodiment. The specific construction and verification process is as follows:

[0086] 1. Establishment of a high-fidelity physical model (complete model)

[0087] EnergyPlus V9.5, developed by the U.S. Department of Energy, was selected as the kernel for the high-fidelity physical model. Based on the building's BIM geometric information (area, floor height, orientation), thermal parameters of the building envelope (U-value of exterior walls, solar heat gain coefficient of windows, etc.), indoor thermal disturbances (hourly operation and shutdown of personnel, lighting, and equipment), and local typical annual meteorological data (TMY), a high-fidelity model of the building was established. The target variables output by the model are hourly comprehensive energy consumption (kWh) and indoor thermal comfort index (PMV).

[0088] 2. Generation and Sampling of Parameterized Sample Sets

[0089] The key design parameters affecting performance were determined to be: external wall insulation layer thickness (X1, range 100~200mm), window-to-wall ratio (X2, range 0.2~0.6), chiller unit COP (X3, range 3.5~6.0), rooftop photovoltaic area (X4, range 0~200m²), and glass curtain wall shading coefficient (X5, range 0.2~0.7). Within the above parameter space, 5000 parameter combinations were generated using the Latin hypercube sampling method. Where i = 1, ..., 5000. Input each set of parameters into the high-fidelity physical model from step 1, simulate for one year (8760 hours), and record the corresponding total annual carbon emissions (kgCO2) and indoor environmental dissatisfaction (based on PMV index) as the output response. This leads to the construction of a sample set. .

[0090] 3. Training of Reduced-Order Models Based on Eigenorthogonal Decomposition (POD)

[0091] Snapshot matrix construction: The time series output by the high-fidelity model for the i-th set of parameters (such as the energy consumption value over 8760 hours) is represented as a column vector. Arrange all 5000 responses into a snapshot matrix. .

[0092] Singular Value Decomposition (SVD): Perform SVD decomposition on matrix A, i.e., A = U·Σ·Vᵀ. Each column of U represents an orthogonal spatial basis (POD mode), and Σ is a singular value diagonal matrix.

[0093] Modal truncation: Calculating the cumulative energy percentage of singular values We choose the minimum number of modes r such that E(r) ≥ 99.9% (i.e., retain 99.9% of the energy). Calculations show that r = 24, meaning that 24 POD modes can accurately approximate the original 8760-dimensional response.

[0094] Constructing a reduced-order model: for any new combination of input parameters Its coefficient under the POD basis Quick prediction using a radial basis function (RBF) regression model: The RBF model uses 5000 sets of parameters. As input, the projection coefficients of the high-fidelity responses corresponding to each parameter onto the POD basis. The model is trained to produce the output. Ultimately, the predicted response of the reduced-order model is... .

[0095] 4. Error Verification and Budget Control of the Reduced-Order Model

[0096] 500 samples were randomly reserved from the 5000 samples as the validation set (not used in training). For each parameter in the validation set, the high-fidelity model output was calculated. Output of the reduced-order model The mean relative error (MRE) and maximum relative error (MAXRE) between them were calculated. The verification results were: MRE = 1.8%, MAXRE = 4.6%. Therefore, this embodiment sets the upper limit of the error for calling the reduced-order model to 5% (i.e., the error budget term mentioned in step S2). During online optimization, any evaluation result of calling the reduced-order model is included with a possible error range of ±5% by default.

[0097] 5. Co-validation mechanism between the reduced-order model and the complete model (to prevent recommendation bias)

[0098] In each iteration of optimization, the system uses the reduced-order model and the full model in tandem according to the following strategy:

[0099] Quick screening: The reduced-order model evaluates all candidate parameter combinations (e.g., 500 combinations generated by sensitivity ranking), sorts them in descending order of their overall predictive performance, and retains only the top three schemes to proceed to the next stage.

[0100] Precise Validation: For the top three schemes mentioned above, re-run the full high-fidelity physical model (EnergyPlus) from step 1 for precise simulation (approximately 2 hours per scheme). Compare the precise results of the full model with the prediction results of the reduced-order model.

[0101] Conflict Handling: If, after validation of the full model, the top-ranked solution differs from the top-ranked solution recommended by the reduced-order model (i.e., their rankings are reversed), the system will use the result of the full model and automatically record this discrepancy. If this reversal occurs three times consecutively in the same optimization problem, a "Reduced-order Model Failure Alarm" will be triggered. The system will pause automatic recommendations and prompt the user that "the current parameter space exceeds the effective range of the reduced-order model; it is recommended to expand the training sample set or use the full model directly for optimization."

[0102] Through the explicit construction, error verification, and collaborative mechanism of the aforementioned order reduction model, this embodiment achieves a speedup of over 100 times (reducing from 2 hours to less than 2 seconds) while ensuring the physical credibility of the final recommended solution.

[0103] The system calculates the percentage of interpolated data within the current evaluation period relative to the total data volume. If this percentage exceeds 5%, the current evaluation result is downgraded to "pre-evaluation" and marked as unsuitable for formal certification. It also accumulates the error values ​​for each step in real time, calculates the ratio of cumulative error to the total budget, and displays this ratio as a progress bar on the user interface.

[0104] Finally, energy consumption simulation is performed on the time-series aggregated data using a physical model-first strategy. A pre-built white-box physical model (e.g., a simplified thermal balance model based on the EnergyPlus kernel, which, according to building thermodynamics principles, takes as input outdoor meteorological parameters, indoor set temperature, and building envelope thermal performance, and outputs hourly predicted energy consumption values; EnergyPlus is an open-source building energy consumption simulation software developed by the U.S. Department of Energy, and its calculation principles are well-known in the field) is invoked, along with the aggregated time-series data (outdoor temperature, solar radiation, indoor set temperature, etc.), to calculate the hourly predicted energy consumption values. Actual energy consumption values ​​for the same time period are extracted from the dynamic time-series data source (IoT measured values ​​for existing buildings during the assessment phase, and simulation input values ​​or measured values ​​for the design phase), and the residual (measured value minus predicted value) is calculated. The residual is compared with a preset residual threshold (e.g., ±15%). When the absolute value of the residual continuously exceeds the threshold (e.g., a deviation exceeding 15% for 3 consecutive hours), an abnormal operating status alarm is generated and pushed to the user terminal, prompting "The air conditioning system's operating mode deviates significantly from the design; it is recommended to check the equipment or user habits." However, it is explicitly stated that the physical model parameters will not be automatically corrected, maintaining the deterministic computational logic of the physical model. At this point, all processed data is integrated into a multi-source heterogeneous fusion dataset for further use.

[0105] Through the above operations, the transformation from raw multi-source data to a highly reliable, aligned, and error-controlled fusion dataset was completed, providing reliable input for subsequent model loading and optimization.

[0106] Step S3 proposed in this embodiment includes: based on the multi-source heterogeneous fusion dataset, retaining the BIM geometric subject and static attributes in the original BIM model, storing the full life cycle of additional historical data in a time-series database or object storage, retaining only lightweight pointers to the additional data in the BIM model, loading the corresponding time period data as needed through the pointers during evaluation, and releasing the dynamic evaluation view after the evaluation is completed, thereby generating a dynamically stripped lightweight BIM evaluation model.

[0107] In this embodiment, the office building will continue to be used as an example for illustration.

[0108] First, identify the components in the BIM model that require full lifecycle traceability, such as exterior walls, floor slabs, and air conditioning units. Obtain a globally unique identifier (GUID) for each component, which is unique within the BIM model. For these components, collect additional historical data from the design, construction, and existing building assessment phases: the design phase includes material selection reports and carbon emission factors provided by manufacturers; the construction phase includes construction logs, material arrival batches, and concrete pouring records; and the existing building assessment phase includes equipment maintenance records, energy consumption ledgers, and component replacement records. This data is scattered across project management systems, Excel spreadsheets, or databases in various formats.

[0109] Then, the aforementioned additional historical data is written to a time-series database (such as TimescaleDB) or object storage (such as MinIO) in chronological order. Each data record is appended with a timestamp (accurate to the second) and a data type label (e.g., "Construction Log" or "Maintenance Record"). For example, a maintenance record for an air conditioning unit is stored as: {"Component GUID":"wall_001","Timestamp":"2025-03-15 14:30:00","Type":"Maintenance Record","Content":"Replace Sealing Ring"}. For large amounts of structured time-series data (such as daily energy consumption), it is directly stored in the time-series database.

[0110] In this embodiment, the historical data is not stored in the original BIM model. Instead, a lightweight pointer record is created for each component. This pointer contains three elements: a globally unique identifier for the component, a storage location address (such as a database table name or the URL path of the object storage), and the time range covered by the data (e.g., "2024-01-01 to 2025-03-31"). For example, the pointer for an exterior wall component can be represented as: {"GUID":"wall_001","storage":"timescaledb: / / history / wall_001","time_range":["2024-01-01","2025-03-31"]}. This pointer is attached to the component of the BIM model as an extended attribute, without increasing the geometric complexity of the model.

[0111] When the evaluation engine needs to assess the carbon emissions of a component throughout its entire lifecycle, it first reads the lightweight pointer from the BIM model. Based on the storage address and time range in the pointer, it retrieves data for the corresponding time period from a time-series database or object storage as needed (e.g., retrieving only the construction phase and maintenance data for the past year). An evaluation view is dynamically built in memory, and carbon emissions are calculated by combining this view with the component's static attributes. Once the evaluation is complete, the evaluation view in memory is immediately released, without writing any external data back to the BIM model.

[0112] Understandably, through the aforementioned dynamic stripping storage and on-demand loading, the BIM model remains lightweight (the original geometry file size remains unchanged), while the entire lifecycle of historical data is stored independently in an external system, allowing for unlimited expansion. This ultimately generates a dynamically stripped, lightweight BIM evaluation model, providing an efficient access foundation for subsequent iterative optimizations.

[0113] Step S4 in this embodiment includes: based on the dynamically stripped lightweight BIM evaluation model, performing new building design optimization or existing building renovation optimization according to the application scenario, identifying the top N parameters with the greatest impact on comprehensive green performance through single-parameter sensitivity ranking, recommending to adjust only one parameter in each iteration and providing at most two alternative adjustment directions, each adjustment only takes effect after manual confirmation by the user and the parameter is locked after taking effect, configuring an oscillation detector to terminate automatic suggestions when reverse changes occur in consecutive iterations, and generating a zero-carbon building design scheme or zero-carbon renovation scheme after manual access control iterative optimization.

[0114] Scenario 1: New Building Design Optimization Mode

[0115] Taking the new design of this office building as an example, five design parameters were optimized: the thickness of the exterior wall insulation layer, the window-to-wall ratio, the COP of the refrigeration unit, the area of ​​the roof photovoltaic system, and the shading coefficient of the glass curtain wall.

[0116] First, a baseline evaluation was run for the five parameters mentioned above, with each parameter individually changed by ±10%, and the change in overall green performance (a weighted comprehensive score of carbon emissions, indoor environmental quality, water resource efficiency, etc.) was calculated. The calculations showed that the insulation layer thickness had the greatest impact on carbon emissions (change rate 18%), followed by the window-to-wall ratio (12%), COP (9%), photovoltaic area (7%), and shading coefficient (5%). Therefore, the top N=3 parameters—insulation layer thickness, window-to-wall ratio, and COP—were identified as the sensitive parameters for this round of optimization.

[0117] To improve computational efficiency, a reduced-order model is employed to accelerate the evaluation. Beforehand, a sample set containing 5000 parameter combinations and corresponding comprehensive performance indicators is generated using a high-precision physical model (e.g., EnergyPlus) for the building type (office) and climate zone (hot summer, cold winter). The reduced-order model is then trained offline using the intrinsic orthogonal decomposition (POD) method, reducing the simulation time from 2 hours to less than 2 seconds. During online optimization, candidate parameter combinations are first input into the reduced-order model to quickly evaluate the comprehensive performance of each scheme. All candidate schemes are ranked, the top three are extracted, and then the complete physical model is used for precise verification, thereby generating the final parameter recommendation results.

[0118] During the iterative optimization phase, the system processes data sequentially in descending order of sensitivity. The first iteration targets the insulation layer thickness, currently 150mm. The system provides two alternatives: increase by 10% (165mm) or decrease by 10% (135mm), displaying the predicted carbon emission changes for each adjustment on the interface. The user (designer) compares the options and selects "increase by 10%" and clicks confirm. The system writes this adjustment into the current design scheme (i.e., the parameter attributes of the corresponding component in the BIM model) and locks the insulation layer thickness parameter, preventing automatic modification in subsequent iterations. The second iteration targets the window-to-wall ratio, currently 0.35. The system again provides two alternatives (0.385 or 0.315), and the user confirms the adoption of 0.315. The third iteration targets the COP, currently 4.5, and the user confirms keeping the original value unchanged.

[0119] Throughout the iteration process, an oscillation detector monitors in real time: if the same parameter (e.g., insulation layer thickness) is recommended in the opposite direction (increase, then decrease, then increase again) in three consecutive iterations, the system automatically terminates the automatic suggestion for that parameter and pops up a prompt box stating "This parameter exhibits oscillation; manual decision-making is recommended." No oscillation occurred in this embodiment. The iteration automatically terminates after all three sensitive parameters are confirmed, generating a zero-carbon building design scheme optimized through manual access control iteration.

[0120] Scenario 2: Existing Building Renovation and Optimization Model

[0121] Taking the same office building as an example, this building has been in operation for 3 years and is now planned to undergo zero-carbon renovation. First, the lightweight BIM evaluation model generated in step S3 is calibrated using real-time IoT monitoring data (hourly energy consumption data, indoor and outdoor temperature and humidity, and equipment start-up and shutdown records for at least one year). The calibration method is as follows: using measured outdoor meteorological parameters, indoor temperature, and equipment start-up and shutdown status as inputs, the actual operating parameters in the model (such as equipment efficiency decay coefficient, actual utilization rate, and staff work and rest) are adjusted so that the average relative error between the model's hourly energy consumption prediction value and the measured value is controlled within 5%. After calibration, the parameters to be modified (e.g., adding extra insulation layer thickness to the exterior walls: options 10mm, 20mm, 30mm; replacing the existing COP=3.2 chiller unit with a high-efficiency unit with COP=5.8; adding photovoltaic capacity to the roof: 50kWp, 100kWp, 150kWp) are used as variables to be optimized, and iterative optimization is performed using the same single-parameter sensitivity ranking method as the design mode. Each adjustment also requires manual confirmation and locking by the user before the final zero-carbon retrofit plan is output.

[0122] Through the above operations, the computation time is kept within an acceptable range while ensuring the optimization quality. At the same time, the manual confirmation mechanism avoids the potential loss of control and oscillation problems that may occur in fully automatic algorithms.

[0123] Step S5 in this embodiment includes: based on the zero-carbon building design scheme or zero-carbon retrofit scheme, a pre-set index conflict matrix is ​​set and the index values ​​of each dimension are calculated. Instead of automatic weighted summation, a Pareto front diagram is generated for the user to manually select a compromise scheme. The algebraic sum of the total carbon emissions and carbon offset of the building throughout its entire life cycle is calculated according to the compromise scheme selected by the user. When the algebraic sum is ≤0, it is determined to be zero carbon. A comprehensive evaluation report containing the compliance degree of multi-dimensional indicators and the zero carbon determination result is generated.

[0124] This step follows the optimized design scheme (new building scenario) or zero-carbon retrofit scheme (existing building scenario) generated in step S4, performs Pareto visualization trade-offs on conflicts of multiple dimensions of indicators, and outputs zero-carbon evaluation results.

[0125] Specifically, firstly, a pre-defined indicator conflict matrix is ​​used, which identifies typical constraints between various green indicators. For example, there is a contradiction between "reducing operational carbon emissions" and "improving natural lighting" (affecting the window-to-wall ratio): increasing the window-to-wall ratio can reduce lighting energy consumption but may increase air conditioning load; while "increasing insulation layer thickness" can reduce operational carbon emissions but will consume more building materials, thus increasing occult carbon emissions. The matrix is ​​stored in the system in tabular form for later reference when making trade-offs.

[0126] For the three candidate optimization schemes obtained in step S4 (i.e., the three superior performance schemes mentioned above, corresponding to different combinations of insulation layer thickness and window-to-wall ratio), the system calculates multi-dimensional index values ​​for each scheme: operational carbon emissions (kgCO2 / m²·year), implicit carbon emissions (kgCO2 / m²), comprehensive indoor environmental quality score (0-100 points), water resource efficiency score (0-100 points), and material recycling rate (%). Then, instead of performing automatic weighted summation, the system plots a Pareto front in a two-dimensional coordinate system: the horizontal axis represents operational carbon emissions, and the vertical axis represents implicit carbon emissions. Each point is represented by a different color to indicate the indoor environmental quality score. Users can see three candidate points on the graphical interface, with two points forming the Pareto boundary. Users can manually select a compromise scheme based on project priorities (e.g., prioritizing operational carbon emissions)—for example, choosing a scheme with lower operational carbon emissions but slightly higher implicit carbon emissions.

[0127] The evaluation process employs a dual-weighting mode. In certification mode, the system forcibly reads the fixed weighting configuration table specified in the national standard "Evaluation Standard for Zero-Carbon Buildings" (e.g., operational carbon weight 0.5, implicit carbon weight 0.2, indoor environment weight 0.2, water resource weight 0.05, material recycling weight 0.05), calculates the comprehensive score, and outputs a score result that meets certification requirements. In design decision mode, users can customize weights (e.g., increasing the operational carbon weight to 0.6). The system calculates an internal decision score based on the customized weights and stores the results separately in the design decision database, without mixing them with certification data. The final evaluation report clearly indicates the weighting mode used and its corresponding weight values.

[0128] Based on the compromise chosen by the user, the system calculates the total carbon emissions over the building's entire lifecycle. For new buildings, the absolute emissions of the design scheme are calculated directly; for renovation schemes of existing buildings, the difference between the emissions after renovation and the baseline emissions before renovation (i.e., emission reduction) is calculated, and then the algebraic sum is calculated with the carbon offset. When the algebraic sum is ≤0, it is determined to be zero carbon (new building) or zero carbon renovation compliant (existing building).

[0129] Subsequently, a comprehensive evaluation report is generated. The report includes: compliance rates for each dimension of indicators (e.g., a 35% reduction in operating carbon emissions compared to the benchmark, and indoor environmental quality meeting the green two-star standard), the zero-carbon assessment result ("zero-carbon compliant" or "non-compliant"), and a diagram illustrating the Pareto compromise point selected by the user. Furthermore, the report undergoes hash-based notarization: the complete report data is read, a Merkle root hash value is calculated, and the hash value, along with the current timestamp, is submitted to the timestamp notarization service of the National Time Service Center to obtain a notarization certificate. Only the hash value and the notarization certificate are uploaded to the blockchain or stored in the notarized notarization system; the original report remains in the central database. In the future, the report can be verified against tampering by recalculating the hash value and comparing it with the certificate.

[0130] Through the above operations, a transparent balance of multi-dimensional indicators was achieved, taking into account both certification and actual design decision-making needs, and the credibility of the evaluation report was ensured through evidence preservation. Thus, the entire methodology completes the entire process from data construction, fusion, lightweight model optimization to multi-dimensional evaluation.

[0131] Furthermore, in this embodiment, after generating the comprehensive evaluation report, in order to prevent the report from being tampered with during transmission or storage, this embodiment performs hash-based evidence storage on the report.

[0132] Specifically, the complete data content of the comprehensive evaluation report generated in step S5 is read. This report can be a PDF file or structured JSON data. The Merkle tree root hash value of this data content is calculated using the SHA-256 hash algorithm. If the report data is large, it is first divided into multiple data blocks, and the hash value of each block is calculated separately. Then, the final root hash is calculated by combining the data layer by layer, thus obtaining a unique fixed-length hash value (e.g., a 64-bit hexadecimal string). Subsequently, this hash value, along with the current timestamp (accurate to milliseconds), is submitted to the timestamp notarization service of the National Time Service Center (or a legally established third-party trusted timestamp service institution). This service returns a notarization certificate containing the original hash value, timestamp, and the service institution's digital signature. For example, the returned certificate content is: {"hash":"3f4b8e9a2c...","timestamp":"2025-03-2010:30:45.123","signature":"......"}.

[0133] Only the hash value and the certificate of authenticity are recorded on the blockchain or stored in the database of the notary and notary system. The entire data content of the original report is kept in a central database, either locally or in the cloud, and is not recorded on the blockchain or transmitted externally. This ensures the verifiability of the report while avoiding the risks of blockchain size expansion and privacy leaks caused by storing complete data.

[0134] When the authenticity of the report needs to be verified in the future, the original report data is reread from the central database, and its Merkle root hash value is recalculated using the same algorithm. Then, the corresponding evidence certificate for the report is retrieved from the evidence storage system, and the recalculated hash value is compared with the hash value in the certificate to see if they match. At the same time, the validity of the timestamp signature is verified. If they match and the signature is valid, the report is determined not to have been tampered with; otherwise, the verification is deemed to have failed. This verification process can be performed independently by anyone without relying on the complete historical record of the original report file.

[0135] Through the above operations, a lightweight, publicly verifiable, and tamper-proof record of the evaluation report is achieved, ensuring the credibility and legal validity of the zero-carbon building evaluation results.

[0136] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0137] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0138] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0139] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0141] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0142] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-dimensional analysis and evaluation method for zero-carbon buildings based on BIM technology, characterized in that, Includes the following steps: S1. Obtain the original BIM model, dynamic time-series data source, and localized carbon emission factor library. Expand the multi-dimensional evaluation attribute set in the BIM model and store the data in layers as a static geometry library, a dynamic time-series library, and a localized factor library to generate a unified BIM data platform. The dynamic time-series data source is simulation data or typical meteorological year data in the new building design stage and IoT real-time monitoring data in the existing building evaluation stage. S2. Based on the BIM unified data platform, the dynamic time series data in the dynamic time series library is aggregated according to the unified evaluation benchmark granularity to obtain aggregated time series data. The low-frequency BIM attributes in the static geometry library are sampled according to the time window to obtain sampled attribute data. The concepts of different systems are aligned through a preset mapping semantic bridge to obtain aligned semantic data. The gaps during the data missing period that do not exceed a preset threshold are interpolated and confidence weights are added to the interpolated data points to generate a multi-source heterogeneous fusion dataset. S3. Based on the multi-source heterogeneous fusion dataset, the BIM geometric subject and static attributes are retained in the original model, and the additional historical data of the entire life cycle is stored in a time-series database or object storage. Only a lightweight pointer to the additional data is retained in the BIM model. During evaluation, the corresponding time period data is loaded as needed through the pointer, and the dynamic evaluation view is released after the evaluation is completed, generating a dynamically stripped lightweight BIM evaluation model. S4. Based on the dynamically stripped lightweight BIM evaluation model, perform the following operations according to the application scenario: If it is in the design phase of a new building, the top N design parameters with the greatest impact on the overall green performance can be identified directly by ranking the single-parameter sensitivity. In each iteration, only one parameter is recommended to be adjusted and at most two alternative adjustment directions are provided. If it is in the existing building assessment stage, the lightweight BIM evaluation model is first calibrated using IoT real-time monitoring data, and then the top N renovation parameters with the greatest impact on the overall green performance are identified by single-parameter sensitivity ranking. Each adjustment takes effect only after manual confirmation by the user and is locked after it takes effect. The automatic suggestion is terminated when the oscillation detector changes in the opposite direction during multiple consecutive iterations, and a zero-carbon building design scheme or zero-carbon renovation scheme is generated after manual access control iteration optimization. S5. Based on the zero-carbon building design scheme or zero-carbon retrofit scheme, a pre-set index conflict matrix is ​​established and the index values ​​of each dimension are calculated. Instead of automatic weighted summation, a Pareto front diagram is generated for the user to manually select a compromise scheme. The algebraic sum of the total carbon emissions and carbon offset of the building throughout its entire life cycle is calculated according to the compromise scheme selected by the user. When the algebraic sum is ≤0, it is determined to be zero carbon. A comprehensive evaluation report containing the compliance degree of multiple dimensions of indicators and the zero carbon determination result is generated.

2. The method according to claim 1, characterized in that, S1 expands the multi-dimensional evaluation attribute set in the BIM model, specifically including: Obtain the type identification and spatial location information of each component in the original BIM model; Based on the component type identifier, query the corresponding evaluation attribute type to be attached to the component from the predefined multi-dimensional attribute mapping table; The multi-dimensional evaluation attribute types include at least carbon emission factor attributes, indoor environmental quality index attributes, water resource efficiency attributes, material recycling rate attributes, and renewable energy proportion attributes; Create an attribute extension record for each component, and write the default value or calculation expression corresponding to the above evaluation attribute type into the attribute extension record; The attribute extension record is associated with the components of the original BIM model and stored to generate the extended BIM model.

3. The method according to claim 1, characterized in that, S1 stores data in layers: a static geometry library, a dynamic time-series library, and a localized factor library. Specifically, this includes: Extract BIM geometric data and component static attribute data from the expanded BIM model, store them in the static geometry library, and save them in a lightweight compression format; Time-series data is obtained from dynamic time-series data sources. In the design phase of new buildings, the time-series data consists of simulated energy consumption data, typical meteorological year data, or reference operating data retrieved from a localized factor library. In the assessment phase of existing buildings, the time-series data consists of real-time sensor monitoring data. The real-time monitoring data includes at least energy consumption time-series data, temperature and humidity time-series data, and equipment start-up and shutdown status data. The time-series data is then stored in a dynamic time-series library after being appended with timestamps and source identifiers. Obtain the climate zone and administrative division identifier of the target building, retrieve the carbon emission benchmark factor and green evaluation standard benchmark value that match the climate zone and administrative division from the external data source, and store them in the localized factor library; Establish a logical association index between the static geometry library, the dynamic time series library, and the localized factor library. The logical association index shall at least contain the mapping relationship between the globally unique identifier of the component and the corresponding time range.

4. The method according to claim 1, characterized in that, The semantic bridge of the preset mapping described in S2 is implemented using a rule engine, including: The standard clauses to be transformed are extracted from the standard clause library, and the standard clauses are transformed into structured rules based on regular expressions and keyword matching. Standard traceability information is attached to each rule. Construct a configurable mapping table, configure a standard source number for each mapping entry, and use the mapping table as the basis for concept alignment; Reject any automatic completion of new concepts; when an unmapped concept is encountered, a dialog box will pop up asking the user to select from the existing library or create it manually. Based on the structured rules and the mapping table, concept alignment between different systems is performed to generate traceable semantic alignment results.

5. The method according to claim 1, characterized in that, S2 includes error budget management after adding confidence weights to the interpolated data points, specifically: The upper limit of the allowable error for each processing step is preset, and the processing steps include at least data interpolation, time series aggregation, and reduced-order model invocation; The percentage of interpolated data in the current evaluation period relative to the total data volume is calculated. When the percentage exceeds a preset threshold, the current evaluation result is downgraded to a pre-evaluation and marked as unusable for formal certification; The error values ​​of each step are accumulated in real time, the ratio of the accumulated error to the total budget is calculated, and the ratio is displayed on the user interface in real time.

6. The method according to claim 1, characterized in that, In S2, the aggregated data obtained after time-series aggregation is used in energy consumption simulation with a physical model-first strategy, specifically including: The model calls a pre-built white-box physical model, inputs aggregated time-series data, and calculates the energy consumption value predicted by the model. The actual energy consumption values ​​for the same time period are extracted from dynamic time-series data sources. The actual energy consumption values ​​are simulation input values ​​in the design stage of new buildings and IoT measured values ​​in the evaluation stage of existing buildings. The residual between the actual energy consumption values ​​and the model prediction values ​​is calculated. The residual is compared with a preset residual threshold; When the absolute value of the residual exceeds the threshold continuously, an abnormal operation status alarm is generated and the alarm information is pushed to the user terminal. It does not automatically correct the physical model parameters, thus maintaining the deterministic calculation logic of the physical model.

7. The method according to claim 1, characterized in that, S3 stores the entire lifecycle of attached historical data in a time-series database or object storage, and only retains lightweight pointers in the BIM model, specifically including: Identify the components in the BIM model that require full lifecycle traceability and obtain the globally unique identifier of the component; Collect additional historical data generated by the component during the design, construction, and existing building assessment phases. The additional historical data includes at least construction logs, equipment maintenance records, energy consumption history data, and material replacement records. The additional historical data is written into a time-series database or object storage in chronological order, and a timestamp and data type tag are attached to each data record. In the BIM model, a lightweight pointer record is created for the component. The lightweight pointer contains a globally unique identifier for the component, the storage location address of the additional historical data, and the time range covered by the data. When the evaluation engine needs to access the full lifecycle data of the component, it retrieves the data for the corresponding time period from external storage on demand through the lightweight pointer. After the evaluation is completed, the evaluation view in memory is released, and the external data is not written back to the BIM model.

8. The method according to claim 1, characterized in that, In S4, the single-parameter sensitivity ranking uses a reduced-order model to accelerate the calculation, specifically including: For specific combinations of building types and climate zones, a training sample set is generated based on a high-precision physical model; The training sample set was trained offline using the intrinsic orthogonal decomposition method to obtain a reduced-order model. During the online optimization phase, candidate parameter combinations are input into the reduced-order model to quickly evaluate the comprehensive green performance index of each candidate scheme. Based on the evaluation results of the order reduction model, all candidate solutions are ranked, and the top three candidate solutions are extracted. The top three candidate solutions are then precisely validated using the complete physical model to generate the final parameter recommendation results.

9. The method according to claim 1, characterized in that, When generating a comprehensive evaluation report in S5, a dual-mode evaluation weighting is used, specifically including: In authentication mode, the fixed weight configuration table specified by the national standard is forcibly read, the fixed weights are applied to the calculation of indicators of each dimension, and the scoring results that meet the authentication requirements are output. In the design decision mode, the user-defined weight configuration is read, the custom weights are applied to the calculation of indicators in each dimension, and the scoring results for internal decision-making are output. The evaluation results corresponding to the custom weights are stored separately in the design decision database and are not mixed with the evaluation data in the certification mode; When generating the evaluation report, clearly indicate the weighting mode currently in use and its corresponding weight values.

10. The method according to claim 1, characterized in that, Following S5, the process also includes hash-based notarization of the comprehensive evaluation report, specifically including: Read the complete data content of the comprehensive evaluation report generated by S5, and calculate the Merkle root hash value of the data content; Submit the Merkle root hash value and the current timestamp together to the timestamp storage service of the National Time Service Center to obtain the storage certificate; Only the hash value and the certificate of evidence are put on the blockchain or stored in the notary and notary system, while the original report data is kept in the central database; When verification is required, the hash value of the original report is recalculated and compared with the hash value in the evidence storage certificate to generate a result indicating whether the verification passed or failed.

Images