Building demand response strategy evaluation and optimization method considering future climate scene
By constructing a building demand response strategy assessment method under future climate scenarios, dynamically generating meteorological data and using machine learning models to optimize building load forecasts, the problem of insufficient assessment of the impact of future climate change on building flexibility regulation is solved, and efficient future climate adaptability assessment and regulation strategy optimization are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTHEAST UNIV
- Filing Date
- 2026-01-29
- Publication Date
- 2026-05-01
AI Technical Summary
The existing building demand response potential assessment has failed to adequately consider future climate change, leading to a decline in the effectiveness or failure of energy system planning and regulation strategies in future operation.
By dynamically generating EPW meteorological data under future climate scenarios, a building performance calculation model is constructed. The hyperparameters are optimized using the XGBoost machine learning surrogate model and the NSGA-II algorithm to establish a high-precision load prediction model. Global sensitivity analysis is then performed to assess the building's potential for flexible regulation.
It enables rapid, accurate, and forward-looking assessment of a building's demand response potential, provides guidance for optimizing regulation strategies under future climate conditions, and ensures that buildings maintain superior responsiveness in the long term.
Smart Images
Figure CN121960191A_ABST
Abstract
Description
Evaluation and Optimization Methods for Building Demand Response Strategies Considering Future Climate Scenarios Technical Field
[0001] This invention belongs to the field of building demand response strategy design, specifically relating to a method for evaluating and optimizing building demand response strategies that consider future climate scenarios. Background Technology
[0002] Building demand response is a key technology for improving grid stability and promoting renewable energy consumption, and its core lies in the accurate assessment of the potential for flexible control of building loads. However, existing methods generally rely on historical meteorological data or static climate assumptions, neglecting the profound impact of future short-, medium-, and long-term climate dynamics on building heating and cooling loads. Against the backdrop of global climate change, frequent extreme weather events and altered temperature patterns have become the new normal. Assessments based on past data cannot reflect the responsiveness of buildings under real-world climate conditions, leading to a risk of reduced effectiveness or even failure in energy system planning and control strategies based on these data in future operations. Therefore, there is an urgent need for a systematic approach to assessing and optimizing building demand response potential in future climate scenarios that can address these issues. Summary of the Invention
[0003] Purpose of the invention: To address the shortcomings of existing building demand response potential assessments, such as insufficient consideration of future climate factors and inadequate adaptability of building flexibility control strategies and energy system design, this invention proposes a method for assessing and optimizing building demand response strategies that considers future climate scenarios. This method enables a scientific and accurate assessment of building demand response potential, providing crucial support for the planning, design, operation, and control of flexible buildings and new energy systems.
[0004] Technical Solution: The present invention provides a method for evaluating and optimizing building demand response strategies considering future climate scenarios, comprising the following steps:
[0005] (1) Dynamically generate EPW meteorological data for the assessment region under near-term, medium-term, and long-term future climate scenarios;
[0006] (2) Construct a building performance calculation model for a typical office building prototype and calculate building load data under recent and future climate scenarios;
[0007] (3) Set the relevant ranges of meteorological variables, design variables and operational variables related to the building demand response potential, and obtain the "variable-performance index" dataset by Latin hypercube sampling and batch simulation;
[0008] (4) Construct an XGBoost machine learning surrogate model for rapid prediction of building load, and use the NSGA-II algorithm to optimize the hyperparameters of XGBoost to obtain a high-precision surrogate model.
[0009] (5) The load adjustable capacity is assessed based on a high-precision proxy model, and the relative change rate under different climatic conditions is calculated to characterize the adaptability and stability of the demand response strategy under future climatic conditions.
[0010] (6) Establish a global sensitivity analysis method for the impact of building design and operation variables on demand response potential, and select the demand response strategy that still maintains high demand response potential stability under future climate scenarios as the building's early demand response design and later operation optimization scheme.
[0011] Furthermore, the implementation process of step (1) is as follows:
[0012] Based on typical meteorological year EPW data files of the assessment area, the CCWorldWeatherGen tool or climate data downscaling tool is used to import future climate prediction data and dynamically generate future meteorological parameters representing different life cycle stages of buildings, forming a future EPW file library corresponding to different climate scenarios.
[0013] Furthermore, the implementation process of step (2) is as follows:
[0014] Typical building prototypes, including small pure office buildings, large pure office buildings, small commercial office buildings, and large commercial office buildings, were selected. Using the building energy consumption simulation software DesignBuilder, their geometry, floor plan, and building envelope type were modeled and set, and the IDF building performance calculation model was output for subsequent parameter setting and batch simulation.
[0015] Furthermore, the implementation process of step (3) is as follows:
[0016] A multi-dimensional demand response scenario parameter space is constructed, including meteorological variables, design variables, and operational variables. The meteorological variables include: meteorological period variables and different demand response meteorological days over 92 days; the meteorological period variables include the current climate scenario, the climate scenario for the next 30 years, and the climate scenario for the next 50 years; the design variables include: wall heat transfer coefficient, roof heat transfer coefficient, window heat transfer coefficient, and window solar heat gain coefficient; the operational variables include: adjustment temperature, adjustment start time, and adjustment duration.
[0017] The Latin hypercube sampling method is used to uniformly sample the above demand response scenarios to obtain demand response solutions;
[0018] Based on the obtained demand response plan, the IDF model file output by DesignBuilder is batch-tuned using the third-party Python library Eppy to generate demand response scenario IDF files, and parallel computation simulation is performed to obtain load reduction values under various demand response scenarios, thus forming a "variable-performance" dataset.
[0019] Furthermore, the implementation process of step (4) is as follows:
[0020] Data preprocessing involves dividing the dataset into training and testing sets, and using the root mean square error coefficient of variation (RMSE), standardized mean bias error (NMBE), and coefficient of determination (R²) as the core evaluation metrics for model accuracy. ; ; ;
[0021] in, These are simulated values. These are the predicted values from the XGBoost model. The average of the simulated values;
[0022] A basic XGBoost model is built using the Python third-party library Scikit-learn, and initial training is performed using a pre-divided dataset.
[0023] By setting reasonable ranges for each hyperparameter to be optimized, the hyperparameters of the XGBoost model are iteratively updated using the NSGA-II algorithm, continuously narrowing the gap between the model's predicted data and the simulated data in terms of cv(RMSE), NMBE, and R. 2 Error; when the error meets the preset requirements, determine the final model parameters, complete the establishment and verification of the XGBoost model, and ensure that the model can accurately reflect the thermal characteristics of the building.
[0024] Furthermore, the implementation process of step (5) is as follows:
[0025] Reduce load per unit area Characterizing the adjustable load per unit area of a building: ;
[0026] in Total building load under the reference temperature setting condition. The total building load after implementing the response strategy, For building air-conditioned area;
[0027] Define the relative rate of change of adjustable building load capacity from near term to long term. index: ;
[0028] like A value >0 indicates that the building or the regulation strategy has enhanced flexibility and superior regulation capabilities under future climate conditions; if =0 indicates that the greater the potential sustainability of the building or the regulation strategy, the more likely it is to maintain its existing regulation capacity under future climate conditions; if <0 indicates that the potential of the building or the control strategy is diminishing, and The smaller the value, the greater the degree of attenuation, and the worse the adaptability to future climate conditions.
[0029] Furthermore, the implementation process of step (6) is as follows:
[0030] Latin hypercube sampling was performed again, with a sample size of 5000. After obtaining the samples, the adjustable capacity of the load per unit area was calculated using the XGBoost high-precision model.
[0031] Based on a third-party Python library, a global sensitivity analysis method for DMIM was established using a resampled dataset to conduct sensitivity analysis on meteorological variables, design variables, and operational variables.
[0032] The first-order sensitivity index is distinguished from the sensitivity analysis results to reflect the direct contribution of input variables to output variables, and the meteorological variables, design variables, and operational variables that affect the building demand response index are ranked.
[0033] The sensitivity analysis results distinguish the overall sensitivity index, which reflects the importance of the interaction between variables to the output variable under the scenario of multiple input variable changes. The variables are ranked by this index, thus obtaining the ranking of the importance of the variables, which provides a reference for subsequent building design and maintenance.
[0034] The present invention also provides a storage medium storing a computer program, which, when executed by at least one processor, implements the steps of the building demand response strategy assessment and optimization method considering future climate scenarios as described above.
[0035] The present invention also provides an electronic device, including a memory and a processor, wherein:
[0036] Memory is used to store computer programs that can run on a processor;
[0037] The processor is configured to, while running the computer program, perform the steps of the building demand response strategy assessment and optimization method considering future climate scenarios as described above.
[0038] Beneficial Effects: Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a multi-scale future climate file covering the near, medium, and long term of buildings, serving as the data foundation for forward-looking assessment. It considers the demand response potential of buildings in future climates, providing forward-looking guidance for early-stage building design to ensure that buildings have superior demand response capabilities in the long term. This invention constructs an XGBoost basic computational model and uses the NSGA-II multi-objective optimization algorithm to accurately verify the hyperparameters of the XGBoost machine learning model, ultimately obtaining a fast evaluation proxy model that combines high accuracy and high computational efficiency. This invention defines a potential change rate index to quantitatively analyze the sustainability and adaptability of building flexible regulation capabilities under future climate scenarios, providing clear guidance for optimizing regulation strategies and energy system planning. Sensitivity analysis yields a ranking of the importance of variables affecting demand response potential, providing a reference for future demand response regulation design. This invention achieves rapid, accurate, and forward-looking assessment of building demand response potential, providing key technical support and decision-making basis for the flexible participation of buildings in the regulation of new energy systems. Attached Figure Description
[0039] Figure 1 is a flowchart of the present invention. Detailed Implementation
[0040] The present invention will now be described in further detail with reference to the accompanying drawings:
[0041] As shown in Figure 1, this invention proposes a method for evaluating and optimizing building demand response strategies that consider future climate scenarios. The specific implementation process is as follows:
[0042] Step 1: Dynamically generate EPW meteorological data for the assessment region under near-term, medium-term, and long-term future climate scenarios.
[0043] Dynamic generation of multi-scale future climate data files. Specifically, based on typical meteorological year EPW data files of the assessment region, using the CCWorldWeatherGen tool or similar validated climate data downscaling tools, future climate prediction data under the shared socio-economic pathways (SSPs) or typical concentration pathways (RCPs) recommended by the Intergovernmental Panel on Climate Change (IPCC) are imported to dynamically generate future meteorological parameters (such as dry-bulb temperature, relative humidity, solar radiation, etc.) representing different stages of the building's life cycle (e.g., near-term operation stage, mid-term maintenance stage - next 30 years, long-term renewal stage - next 50 years), forming a future EPW file library corresponding to different climate scenarios.
[0044] Step 2: Construct an IDF building performance calculation model for a typical office building prototype to calculate building load data under near-term and future climate scenarios.
[0045] Select typical building prototypes (such as small pure office buildings, large pure office buildings, small commercial office buildings, and large commercial office buildings), and use the building energy consumption simulation software DesignBuilder to model and set their geometry, floor plan, and building envelope type, and output the IDF building performance calculation model for subsequent parameter setting and batch simulation.
[0046] Step 3: Define the relevant ranges for meteorological variables, design variables, and operational variables related to the building's demand response potential, and obtain a dataset of "variable-performance indicators" through Latin hypercube sampling and batch simulation.
[0047] Construct a multi-dimensional demand response scenario parameter space, including meteorological variables, design variables, and operational variables. Meteorological variables include: meteorological period variables (including current climate scenario, climate scenario for the next 30 years, and climate scenario for the next 50 years); and 92 different demand response meteorological days (July, August, and September). Design variables include: wall heat transfer coefficient (a continuous variable, such as 0.2~3 W / (m²)). 2 ·K)), roof heat transfer coefficient (a continuous variable, such as 0.1 ~2.5 (W / (m²)). 2 ·K)), window heat transfer coefficient (a continuous variable, such as 2.6 ~ 6.4 (W / (m²)). 2 •K) and window solar heat gain coefficient (a continuous variable, such as 0.4~0.69). Operating variables include: 3 set temperatures (1.0℃, 1.5℃, 2.0℃), 5 set start times (12:00, 13:00, 14:00, 15:00, 16:00), and 2 set durations (1h, 2h).
[0048] The Latin hypercube sampling method was used to efficiently and uniformly sample the above demand response scenarios, resulting in 5,000 demand response solutions.
[0049] Based on the obtained demand response scheme, the IDF model files output by DesignBuilder are batch-tuned using the Python third-party library Eppy to generate 5,000 demand response scenario IDF files. Parallel computation simulation is then performed to obtain load reduction values under various demand response scenarios. Combined with demand response variables, a dataset is formed.
[0050] Step 4: Construct an XGBoost machine learning model for load data prediction, and use the NSGA-II algorithm to optimize the hyperparameters of the machine learning model to obtain a high-precision machine learning model.
[0051] Data preprocessing involves dividing the 5000 sample dataset generated in step 3 into training and testing sets at an 80% to 20% ratio. The root mean square error (RMSE), normalized mean bias error (NMBE), and coefficient of determination (R²) are used as the core evaluation metrics for model accuracy. ; ; ;
[0052] in, These are simulated values. These are the predicted values from the XGBoost model. This represents the average of the simulated values.
[0053] A basic XGBoost model is built using the Python third-party library Scikit-learn, and initial training is performed using a pre-divided dataset.
[0054] Set reasonable ranges for each hyperparameter to be optimized (Learning_rate, N_estimators, Max_depth, Colsample_Bytree, and Subample), and iteratively update the hyperparameters of the XGBoost model using the NSGA-II algorithm, continuously narrowing the gap between the model's predicted data and the simulated data in terms of cv(RMSE), NMBE, and R. 2 Error. When the error meets the preset requirements, the final model parameters are determined, and the XGBoost model is established and verified to ensure that the model can accurately reflect the thermal characteristics of the building.
[0055] Step 5: Conduct a comprehensive assessment of the potential for flexible regulation of building demand based on a high-precision machine learning model, calculate the relative rate of change, and clarify the future climate adaptability of the building load flexible regulation strategy.
[0056] Based on the high-precision XGBoost algorithm model obtained after training in step 4, input the corresponding variables and perform a building demand response potential analysis to evaluate its adjustable load per unit area. Specifically:
[0057] Reduce load per unit area Characterizing the adjustable load per unit area of a building: ;
[0058] in Total building load under the reference temperature setting condition. The total building load after implementing the response strategy, This refers to the building's air-conditioned area.
[0059] Define the relative rate of change of adjustable building load capacity from near term to long term. index: ;
[0060] like A value >0 indicates that the building or the regulation strategy has enhanced flexibility and superior regulation capabilities under future climate conditions; if =0 indicates that the greater the potential sustainability of the building or the regulation strategy, the more likely it is to maintain its existing regulation capacity under future climate conditions; if <0 indicates that the potential of the building or the control strategy is diminishing, and The smaller the value, the greater the attenuation and the worse the adaptability under future climate conditions. This method can clarify the potential for flexible control of building loads and the adaptability of optimization measures under different scenarios.
[0061] Step 6: Establish a sensitivity analysis method for building demand response potential and obtain the importance ranking of variables affecting demand response potential.
[0062] Latin hypercube sampling was performed again, with a sample size of 5000. After obtaining the sample, the adjustable load capacity per unit area of the building was calculated using the XGBoost high-precision model.
[0063] Based on a third-party Python library, a global sensitivity analysis method for DMIM was established using a resampled dataset to conduct sensitivity analyses on meteorological variables, design variables, and operational variables.
[0064] The sensitivity analysis results distinguish the first-order sensitivity index, which reflects the direct contribution of input variables to output variables. The meteorological variables, design variables, and operational variables that affect the building demand response index are ranked.
[0065] The sensitivity analysis results distinguish the total order sensitivity index, which reflects the importance of the interaction between variables to the output variable in scenarios with multiple input variables. The variables are ranked by this index, thus obtaining the ranking of variable importance, which provides a reference for future building design and operation and maintenance.
[0066] The present invention also provides a storage medium storing a computer program, which, when executed by at least one processor, implements the steps of the building demand response strategy assessment and optimization method considering future climate scenarios as described above.
[0067] The present invention also provides an electronic device, including a memory and a processor, wherein: the memory is used to store a computer program that can run on the processor; the processor is used to execute, when running the computer program, the steps of the building demand response strategy assessment and optimization method considering future climate scenarios as described above.
[0068] It should be understood that the embodiments and descriptions above are only the principles, main features and advantages of the present invention. Various changes and modifications can be made to the present invention without departing from the spirit and scope of the invention, and all such changes and modifications fall within the protection scope of the present invention.
Claims
1. A method for evaluating and optimizing building demand response strategies considering future climate scenarios, characterized in that, Includes the following steps: (1) Dynamically generate EPW meteorological data for the near, medium and long term future climate scenarios of the assessment area; (2) Construct a building performance calculation model for a typical office building prototype and calculate building load data under near and future climate scenarios; (3) Set the relevant ranges of meteorological variables, design variables and operation variables related to building demand response potential, and obtain a dataset of "variable-performance indicators" through Latin hypercube sampling and batch simulation; (4) Construct an XGBoost machine learning surrogate model for rapid prediction of building load, and use the NSGA-II algorithm to optimize the hyperparameters of XGBoost to obtain a high-precision surrogate model; (5) Evaluate the building demand flexibility regulation potential based on the high-precision surrogate model, and calculate the relative change rate under different climate conditions to characterize the adaptability and stability of the demand response strategy under future climate conditions; (6) Establish a global sensitivity analysis method for the influence of building design and operation variables on demand response potential, and select the demand response strategy that still maintains high demand response potential stability under future climate scenarios as the building's early demand response design and later operation optimization scheme.
2. The method for evaluating and optimizing building demand response strategies considering future climate scenarios according to claim 1, characterized in that, The implementation process of step (1) is as follows: Based on the typical meteorological year EPW data file of the assessment area, the CCWorldWeatherGen tool or climate data downscaling tool is used to import future climate prediction data, dynamically generate future meteorological parameters representing different life cycle stages of buildings, and form a future EPW file library corresponding to different climate scenarios.
3. The method for evaluating and optimizing building demand response strategies considering future climate scenarios according to claim 1, characterized in that, The implementation process of step (2) is as follows: Select typical building prototypes including small pure office buildings, large pure office buildings, small commercial office buildings and large commercial office buildings, and use the building energy consumption simulation software DesignBuilder to model and set their geometry, floor plan and enclosure structure type, and output the IDF building performance calculation model for subsequent parameter setting and batch simulation.
4. The method for evaluating and optimizing building demand response strategies considering future climate scenarios according to claim 1, characterized in that, The implementation process of step (3) is as follows: Construct a multi-dimensional demand response scenario parameter space, including meteorological variables, design variables, and operational variables; the meteorological variables include: meteorological period variables and demand response meteorological day variables; the meteorological period variables include the current climate scenario, the climate scenario for the next 30 years, and the climate scenario for the next 50 years; the design variables include: wall heat transfer coefficient, roof heat transfer coefficient, window heat transfer coefficient, and window solar heat gain coefficient; the operational variables include: adjustment temperature, adjustment start time, and adjustment duration; use the Latin hypercube sampling method to uniformly sample the above demand response scenarios to obtain demand response schemes; based on the obtained demand response schemes, combine the third-party Python library Eppy to batch adjust the IDF model files output by DesignBuilder, generate demand response scenario IDF files, and perform parallel computation simulation output to obtain load reduction values under various demand response scenarios, thereby forming a "variable-performance" dataset.
5. The method for evaluating and optimizing building demand response strategies considering future climate scenarios according to claim 1, characterized in that, The implementation process of step (4) is as follows: data preprocessing, dividing the dataset into training set and test set, and using root mean square error coefficient of variation RMSE, standardized mean deviation error NMBE and coefficient of determination R² as the core evaluation indicators of model accuracy. ； ； ;in, These are simulated values. These are the predicted values from the XGBoost model. The average value of the simulated values is used; a basic XGBoost model is built using the Python third-party library Scikit-learn, and initial training is performed using a pre-divided dataset; reasonable ranges are set for each hyperparameter to be optimized, and the hyperparameters of the XGBoost model are iteratively updated using the NSGA-II algorithm to continuously narrow the gap between the model's predicted data and the simulated data in terms of cv(RMSE), NMBE, and R. 2 Error; when the error meets the preset requirements, determine the final model parameters, complete the establishment and verification of the XGBoost model, and ensure that the model can accurately reflect the thermal characteristics of the building.
6. The method for evaluating and optimizing building demand response strategies considering future climate scenarios according to claim 1, characterized in that, The process of step (5) is as follows: using unit area load reduction Characterizing the adjustable load per unit area of a building: ;in, Total building load under the reference temperature setting condition. The total building load after implementing the response strategy, For the building's air-conditioned area; define the relative rate of change of the building's adjustable load capacity from the near term to the long term. index: ;like A value >0 indicates that the building or the regulation strategy has enhanced flexibility and superior regulation capabilities under future climate conditions; if =0 indicates that the greater the potential sustainability of the building or the regulation strategy, the more likely it is to maintain its existing regulation capacity under future climate conditions; if <0 indicates that the potential of the building or the control strategy is diminishing, and The smaller the value, the greater the degree of attenuation, and the worse the adaptability to future climate conditions.
7. The method for evaluating and optimizing building demand response strategies considering future climate scenarios according to claim 1, characterized in that, The implementation process of step (6) is as follows: Latin hypercube sampling is performed again, with a sample size of 5000. After obtaining the samples, the adjustable capacity of the building unit area load is calculated using the XGBoost high-precision model. Based on a third-party Python library, a global sensitivity analysis method for DMIM was established using a resampled dataset to conduct sensitivity analysis on meteorological variables, design variables, and operational variables. The sensitivity analysis results distinguish the first-order sensitivity index, which reflects the direct contribution of input variables to output variables, and ranks the meteorological, design, and operational variables that affect the building demand response index. The sensitivity analysis results also distinguish the total-order sensitivity index, which reflects the importance of the interaction between variables to the output variable under multiple input variable changes. The variables are ranked using this index, thus obtaining the importance ranking of the variables, which provides a reference for subsequent building design and maintenance.
8. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by at least one processor, implements the steps of the building demand response strategy assessment and optimization method considering future climate scenarios as described in any one of claims 1 to 7.
9. An electronic device, characterized in that, It includes a memory and a processor, wherein: the memory is used to store a computer program that can run on the processor; the processor is used to execute, when running the computer program, the steps of the building demand response strategy assessment and optimization method considering future climate scenarios as described in any one of claims 1 to 7.