Electrical load regulation potential evaluation method and device based on multi-model fusion, medium and equipment

Abstract

The invention discloses an electrical load regulation potential evaluation method and device based on multi-model fusion, a medium and equipment, and the method comprises the steps: obtaining the to-be-predicted electrical load data of an owner for each owner in a plurality of owners, and constructing the to-be-predicted electrical load characteristics based on the to-be-predicted electrical load data, the to-be-predicted electrical load characteristics comprise electrical load characteristics and time period characteristics; inputting the to-be-predicted electrical load characteristics into a multi-model fusion model, performing electrical load prediction on the to-be-predicted electrical load through each electrical load prediction model in the multi-model fusion model, and fusing results predicted by the electrical load prediction models to obtain an electrical load prediction fusion value; and determining the maximum interruptible power, the transferable electric quantity and the power adjustment range of the owner based on the electric load prediction fusion value, and determining the comprehensive electric load adjustment potential of the owner based on the maximum interruptible power, the transferable electric quantity and the power adjustment range.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method, apparatus, storage medium, and computer equipment for assessing the electrical load regulation potential based on multi-model fusion. Background Technology

[0002] In my country's electricity sector, load management among large users has long been a crucial development direction supported and promoted. Load management plays an indispensable and critical role in balancing electricity supply and demand, improving grid operating efficiency, and ensuring a safe and stable power supply. By rationally regulating the electricity load of large users, power shortages during peak hours can be effectively avoided, the occurrence of power outages and rationing can be reduced, and the utilization efficiency of electricity resources can be improved, while reducing energy waste.

[0003] In the operation and management of power systems, accurate assessment of load regulation potential is crucial, as it relates to the stable and efficient operation of the power system and also affects the rational allocation and effective utilization of energy. With the continuous development of the electricity market and the widespread application of emerging technologies such as distributed energy resources and smart grids, the complexity and uncertainty of electrical loads have increased significantly, bringing many challenges to the assessment of load regulation potential.

[0004] How to meet the requirements of modern power systems for assessing load regulation potential is a key issue in this field. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, apparatus, storage medium, and computer device for evaluating electrical load regulation potential based on multi-model fusion.

[0006] According to one aspect of this application, a method for assessing electrical load regulation potential based on multi-model fusion is provided, the method comprising: For each of the multiple property owners, obtain the property owner's electricity load data to be predicted, and construct the electricity load characteristics to be predicted based on the electricity load data, wherein the electricity load characteristics to be predicted include electricity load characteristics and time period characteristics; The electrical load characteristics to be predicted are input into a multi-model fusion model. Each electrical load prediction model in the multi-model fusion model performs electrical load prediction on the electrical load characteristics to be predicted, and the results predicted by each electrical load prediction model are fused to obtain a fused electrical load prediction value. The electrical load prediction model includes XGBoost model, random forest model, support vector regression model and generalized additive model. Based on the combined electrical load forecast values, the owner's maximum interruptible power, transferable power, and power adjustment range are determined respectively, and the owner's comprehensive electrical load adjustment potential is determined based on the maximum interruptible power, transferable power, and power adjustment range.

[0007] Optionally, before inputting the electrical load characteristics to be predicted into the multi-model fusion model, the method further includes: Obtain electrical load training samples, and use the electrical load training samples to train each electrical load prediction model respectively; The fusion weights of each load prediction model are optimized using the Bayesian optimization method to obtain the fusion weights of each load prediction model. Accordingly, the results predicted by each electrical load forecasting model are fused to obtain a fused electrical load forecast value, including: The results predicted by each load prediction model are fused based on the fusion weights of each load prediction model to obtain the fused load prediction value.

[0008] Optionally, the step of optimizing the fusion weights corresponding to each electrical load prediction model using a Bayesian optimization method to obtain the fusion weights of each electrical load prediction model includes: With the goal of minimizing the mean square error of the multi-model fusion model, the fusion weights of each electric load prediction model are optimized using Bayesian optimization methods based on the electric load validation samples to obtain the fusion weights of each electric load prediction model.

[0009] Optionally, after obtaining the fused electrical load forecast values, the method further includes: The electrical load prediction fusion values ​​are grouped based on a preset threshold, and the electrical load prediction fusion values ​​are compressed based on the mean of the electrical load prediction fusion values ​​within the group, so as to update the electrical load prediction fusion values ​​based on the compression results.

[0010] Optionally, based on the combined electrical load forecast value, the owner's maximum interruptible power, transferable power, and power regulation range are determined respectively, and the owner's comprehensive electrical load regulation potential is determined based on the maximum interruptible power, transferable power, and power regulation range, including: The maximum interruptible power of the owner is determined by the product of the preset interruption ratio coefficient and the user's basic load power corresponding to the power load prediction fusion value; the transferable power of the owner is determined by the product of the preset power transfer ratio coefficient and the user's daily power consumption corresponding to the power load prediction fusion value; and the power adjustment range of the owner is determined by the user's maximum power consumption function and the user's minimum power consumption power in the power load prediction fusion value. For each property owner, an interruptibility potential index is determined based on the owner's maximum interruptible power, the preset optimal value of the maximum interruptible power, and the preset worst value of the maximum interruptible power; a transferability potential index is determined based on the owner's transferable power, the preset optimal value of the transferable power, and the preset worst value of the transferable power; an adjustment potential index is determined based on the owner's power adjustment range, the preset optimal value of the power adjustment range, and the preset worst value of the power adjustment range; and the owner's interruptibility potential index, transferability potential index, and adjustment potential index are combined to determine the owner's comprehensive power load adjustment potential.

[0011] Optionally, the method further includes: Obtain the owner scope corresponding to the virtual power plant; Based on the sum of the interruptibility potential index, transferability potential index, and adjustability potential index of each owner within the owner scope, the owner's adjustability potential is determined, and the total adjustability potential corresponding to the virtual power plant is determined based on the sum of the owner's adjustability potential of all owners within the owner scope.

[0012] Optionally, the method further includes: Based on the combined electricity load forecast values ​​of each property owner, typical electricity consumption patterns of each property owner are clustered. For each typical electricity consumption pattern, the interruptibility potential index, transferability potential index, and adjustability potential index of each owner in the typical electricity consumption pattern are clustered to obtain multiple interruptibility potential index classes, multiple transferability potential index classes, and multiple adjustability potential index classes. Based on each cluster, each owner in the typical electricity consumption pattern is divided into different levels of interruptibility potential owners, different levels of transferability potential owners, and different levels of adjustability potential owners. Based on the real-time load characteristics corresponding to the virtual power plant and the interruptibility potential level, transferability potential level, and adjustability potential level of each owner within the owner's scope, power dispatch is carried out for each owner.

[0013] According to another aspect of this application, a device for assessing electrical load regulation potential based on multi-model fusion is provided, the device comprising: The feature construction module is used to obtain the predicted electricity load data of each owner among multiple owners, and construct the predicted electricity load features based on the predicted electricity load data, wherein the predicted electricity load features include electricity load features and time period features. The load forecasting module is used to input the electrical load characteristics to be predicted into a multi-model fusion model, and to perform electrical load forecasting on the electrical load characteristics to be predicted by each electrical load forecasting model in the multi-model fusion model, and to fuse the results predicted by each electrical load forecasting model to obtain a fused electrical load forecast value; wherein, the electrical load forecasting model includes XGBoost model, random forest model, support vector regression model and generalized additive model; The potential assessment module is used to determine the owner's maximum interruptible power, transferable power, and power adjustment range based on the power load prediction fusion value, and to determine the owner's comprehensive power load adjustment potential based on the maximum interruptible power, transferable power, and power adjustment range.

[0014] Optionally, the apparatus further includes: a model training module, used for: Obtain electrical load training samples, and use the electrical load training samples to train each electrical load prediction model respectively; The fusion weights of each load prediction model are optimized using the Bayesian optimization method to obtain the fusion weights of each load prediction model. Accordingly, the load forecasting module is also used for: The results predicted by each load prediction model are fused based on the fusion weights of each load prediction model to obtain the fused load prediction value.

[0015] Optionally, the model training module is further configured to: With the goal of minimizing the mean square error of the multi-model fusion model, the fusion weights of each electric load prediction model are optimized using Bayesian optimization methods based on the electric load validation samples to obtain the fusion weights of each electric load prediction model.

[0016] Optionally, the load forecasting module is further configured to: The electrical load prediction fusion values ​​are grouped based on a preset threshold, and the electrical load prediction fusion values ​​are compressed based on the mean of the electrical load prediction fusion values ​​within the group, so as to update the electrical load prediction fusion values ​​based on the compression results.

[0017] Optionally, the potential assessment module is further configured to: The maximum interruptible power of the owner is determined by the product of the preset interruption ratio coefficient and the user's basic load power corresponding to the power load prediction fusion value; the transferable power of the owner is determined by the product of the preset power transfer ratio coefficient and the user's daily power consumption corresponding to the power load prediction fusion value; and the power adjustment range of the owner is determined by the user's maximum power consumption function and the user's minimum power consumption power in the power load prediction fusion value. For each property owner, an interruptibility potential index is determined based on the owner's maximum interruptible power, the preset optimal value of the maximum interruptible power, and the preset worst value of the maximum interruptible power; a transferability potential index is determined based on the owner's transferable power, the preset optimal value of the transferable power, and the preset worst value of the transferable power; an adjustment potential index is determined based on the owner's power adjustment range, the preset optimal value of the power adjustment range, and the preset worst value of the power adjustment range; and the owner's interruptibility potential index, transferability potential index, and adjustment potential index are combined to determine the owner's comprehensive power load adjustment potential.

[0018] Optionally, the potential assessment module is further configured to: Obtain the owner scope corresponding to the virtual power plant; Based on the sum of the interruptibility potential index, transferability potential index, and adjustability potential index of each owner within the owner scope, the owner's adjustability potential is determined, and the total adjustability potential corresponding to the virtual power plant is determined based on the sum of the owner's adjustability potential of all owners within the owner scope.

[0019] Optionally, the potential assessment module is further configured to: Based on the combined electricity load forecast values ​​of each property owner, typical electricity consumption patterns of each property owner are clustered. For each typical electricity consumption pattern, the interruptibility potential index, transferability potential index, and adjustability potential index of each owner in the typical electricity consumption pattern are clustered to obtain multiple interruptibility potential index classes, multiple transferability potential index classes, and multiple adjustability potential index classes. Based on each cluster, each owner in the typical electricity consumption pattern is divided into different levels of interruptibility potential owners, different levels of transferability potential owners, and different levels of adjustability potential owners. Based on the real-time load characteristics corresponding to the virtual power plant and the interruptibility potential level, transferability potential level, and adjustability potential level of each owner within the owner's scope, power dispatch is carried out for each owner.

[0020] According to another aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for assessing electrical load regulation potential based on multi-model fusion.

[0021] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described method for assessing electrical load regulation potential based on multi-model fusion.

[0022] By employing the aforementioned technical solution, the load regulation potential is assessed through multi-model fusion, overcoming the limitations of single models in processing complex load data. Different models can capture the characteristics and patterns of load from their unique perspectives. Fusing their prediction results helps improve the accuracy of load forecasting, thereby making the maximum interruptible power, transferable power, and power regulation range determined based on the prediction results more accurate and reliable. The final comprehensive load regulation potential assessment result is also more scientific and comprehensive, helping the power system to more accurately grasp the load regulation capabilities of owners, achieve optimal allocation and efficient utilization of power resources, improve the stability and reliability of the power system, and better meet the actual needs of the power market in terms of demand response.

[0023] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a method for assessing electrical load regulation potential based on multi-model fusion, provided in an embodiment of this application, is shown. Figure 2 This paper illustrates a flowchart of another method for evaluating electrical load regulation potential based on multi-model fusion, provided in an embodiment of this application. Figure 3 The diagram shows a structural schematic of an electrical load regulation potential assessment device based on multi-model fusion provided in an embodiment of this application. Detailed Implementation

[0025] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0026] Example 1 This embodiment provides a method for evaluating the electrical load regulation potential based on multi-model fusion, such as... Figure 1 As shown, the method includes: Step 101: For each of the multiple owners, obtain the owner's predicted electricity load data, and construct the predicted electricity load features based on the predicted electricity load data, wherein the predicted electricity load features include electricity load features and time period features. Step 102: Input the electrical load characteristics to be predicted into the multi-model fusion model, and use each electrical load prediction model in the multi-model fusion model to predict the electrical load characteristics. Then, fuse the prediction results of each electrical load prediction model to obtain the electrical load prediction fusion value. The electrical load prediction model includes XGBoost model, random forest model, support vector regression model and generalized additive model. Step 103: Based on the predicted and integrated electrical load values, determine the owner's maximum interruptible power, transferable power, and power adjustment range, and determine the owner's comprehensive electrical load adjustment potential based on the maximum interruptible power, transferable power, and power adjustment range.

[0027] In this embodiment, firstly, for each of the multiple owners, their predicted electrical load data is obtained, and based on this data, predicted electrical load features are constructed. These features encompass electrical load characteristics and time period characteristics, aiming to comprehensively capture the characteristics of electrical load across different dimensions. In a specific application scenario, the predicted electrical load features include: electrical load data for the corresponding time points of the previous seven days; electrical load values ​​for the same time points of the past seven days; whether it is a weekend (binary feature); whether it is a holiday (binary feature); time of day (continuous variable); day of the week (categorical variable); and quarter (categorical variable).

[0028] Next, the constructed electrical load characteristics to be predicted are input into the multi-model fusion model, such as... Figure 2 As shown, this model integrates multiple electricity load prediction models, including XGBoost, Random Forest, Support Vector Regression, and Generalized Additive Model. Each model predicts the electricity load separately, and then the prediction results are fused to obtain a fused electricity load prediction value. In assessing the electricity load regulation potential, different electricity load prediction models have their own characteristics. For example, the XGBoost model excels at handling structured data and feature interactions, and can quickly fit data; the Random Forest model has strong anti-overfitting capabilities and can handle high-dimensional data; the Support Vector Regression model has advantages in small sample and high-dimensional data; and the Generalized Additive Model can flexibly handle nonlinear relationships. Multi-model fusion, by integrating the prediction results of multiple models, can effectively reduce the impact of single model errors on the overall prediction. Even if a model's prediction is biased, the prediction results of other models can correct it, making the final prediction result closer to the true value. This method can construct a collaborative prediction framework of heterogeneous hybrid models. By integrating models with different learning mechanisms, it generates differentiated preliminary electricity load prediction results from multiple perspectives, combining the advantages of different models to improve prediction accuracy.

[0029] Finally, based on the combined value of the electricity load forecast, the owner's maximum interruptible power, transferable power, and power regulation range are determined respectively. Based on these three key indicators, the owner's comprehensive electricity load regulation potential is determined, providing a scientific basis for the load management and resource allocation of the power system.

[0030] By applying the technical solution of this embodiment, the load regulation potential is evaluated through multi-model fusion, overcoming the limitations of a single model in processing complex load data. Different models can capture the characteristics and patterns of load from their own unique perspectives. After fusing their prediction results, the accuracy of load prediction is improved, making the owner's maximum interruptible power, transferable power, and power regulation range determined based on the prediction results more accurate and reliable. The final comprehensive load regulation potential assessment result is also more scientific and comprehensive, helping the power system to more accurately grasp the owner's load regulation capabilities, achieve optimal allocation and efficient utilization of power resources, improve the stability and reliability of the power system, and better meet the actual needs of the power market in terms of demand response.

[0031] Example 2 In another embodiment of this application, optionally, before inputting the electrical load features to be predicted into the multi-model fusion model, the method further includes: obtaining electrical load training samples, training each electrical load prediction model using the electrical load training samples respectively; and optimizing the fusion weights corresponding to each electrical load prediction model using a Bayesian optimization method to obtain the fusion weights of each electrical load prediction model. Accordingly, the results predicted by each load forecasting model are fused to obtain a fused load forecast value, including: fusing the results predicted by each load forecasting model based on the fusion weight of each load forecasting model to obtain a fused load forecast value.

[0032] In this embodiment, firstly, training samples of electrical load are obtained. These samples serve as the foundational data for training the electrical load prediction model, and their data structure is identical to the characteristics of the electrical load to be predicted. Next, each electrical load prediction model is trained using these training samples. These models include XGBoost, Random Forest, Support Vector Regression, and Generalized Additive Models. Training allows each model to learn the characteristics and patterns of the electrical load data. Then, Bayesian optimization is used to optimize the fusion weights corresponding to each electrical load prediction model. Bayesian optimization is a highly efficient optimization algorithm that, within a given parameter space, finds the optimal parameter values ​​for model performance through continuous iteration and evaluation; in this case, it involves finding the optimal fusion weights for each electrical load prediction model. After obtaining the fusion weights for each electrical load prediction model, when the characteristics of the electrical load to be predicted are input into the multi-model fusion model, and each electrical load prediction model makes its prediction and outputs results, the results predicted by each electrical load prediction model are fused based on the previously optimized fusion weights, ultimately yielding the fused electrical load prediction value. This application embodiment trains each load prediction model separately by acquiring load training samples, allowing each model to fully learn the characteristics of the load data and improve its own load prediction ability, laying the foundation for accurate predictions. Furthermore, using Bayesian optimization to optimize the fusion weights of each load prediction model, compared to traditional weight setting methods, can more scientifically and accurately determine the contribution of each model in the fusion process, avoiding the subjectivity and arbitrariness of manually setting weights. Fusing the prediction results of each model based on the optimized fusion weights allows for a more comprehensive utilization of the advantages of different models. By integrating the prediction information from each model, the errors that may be caused by a single model prediction can be reduced, thereby improving the accuracy and reliability of the fused load prediction value. This provides a strong guarantee for accurately determining the owner's comprehensive load regulation potential and helps the power system to more accurately manage load and allocate resources.

[0033] Example 3 In another embodiment of this application, optionally, the step of optimizing the fusion weights corresponding to each load prediction model using a Bayesian optimization method to obtain the fusion weights of each load prediction model includes: using load validation samples to optimize the fusion weights corresponding to each load prediction model using a Bayesian optimization method with the goal of minimizing the prediction mean square error of the multi-model fusion model, thereby obtaining the fusion weights of each load prediction model.

[0034] In this embodiment, a Bayesian optimization method is used to minimize the mean squared error of the multi-model fusion model. The fusion weights of each electricity load prediction model are optimized using electricity load validation samples to obtain the optimal fusion weights. Dynamic optimization of the fusion weights allows for a more scientific quantification of the contribution of each individual model (such as XGBoost, Random Forest, etc.) in the fusion process, avoiding the subjectivity of manually setting weights. Simultaneously, minimizing the mean squared error of the prediction is directly linked to the prediction accuracy index, ensuring that the deviation between the fused electricity load prediction value and the actual value is minimized. Specifically, the fused electricity load prediction value is represented as... , where w m Let m be the weights of model m, and M represents the number of electrical load prediction models. Bayesian optimization uses the mean square error of the validation set (MSE) as the objective function: The optimal weight w is searched iteratively using a Gaussian process surrogate model. m N represents the total number of electrical load verification samples, y i This represents the actual electrical load value of the i-th sample. This represents the predicted electrical load value for the i-th sample.

[0035] In another embodiment of this application, optionally, after obtaining the fused electrical load prediction value, the method further includes: grouping the fused electrical load prediction value based on a preset threshold, compressing the fused electrical load prediction value based on the mean of the fused electrical load prediction values ​​within the group, and updating the fused electrical load prediction value based on the compression result.

[0036] In this embodiment, after obtaining the fused load forecast value through multi-model fusion and Bayesian optimization to determine the fusion weights, the fused value is not directly used for subsequent operations. Instead, further data processing is performed. Specifically, the fused load forecast value is first grouped according to a preset threshold. A load level discretization method oriented towards scheduling decisions is used to standardize the high-precision load sequence obtained from the fused forecast. This is converted into a series of standardized schedulable command points, greatly simplifying the decision space of the scheduling model while preserving key trends. Data compression is then performed on the predicted load. Based on a preset threshold, continuous loads with differences within the preset threshold are processed into the same load value, reducing computational load. The combined predicted load value is compressed using a threshold. Group the load values ​​and use the group mean to represent the compression value. Formula: , n represents the number of data points for the load forecast fusion value in the group, L j L represents the j-th predicted combined value of electrical load. compThis represents the compression result of the grouping. Specifically, after grouping, the mean of all electrical load forecasts within each group is calculated. This mean represents the central tendency of the data in that group and reflects the overall level of the data within the group. Compression can be achieved by replacing each data point in the group with the group's mean, or by converging towards the mean according to certain rules. Finally, the original electrical load forecasts are updated based on the compression results, resulting in updated electrical load forecasts. Subsequent operations, such as determining the owner's maximum interruptible power and transferable power, will be based on this updated value.

[0037] Example 4 In another embodiment of this application, optionally, based on the combined electrical load forecast value, the owner's maximum interruptible power, transferable power, and power regulation range are determined respectively, and the owner's comprehensive electrical load regulation potential is determined based on the maximum interruptible power, transferable power, and power regulation range, including: The maximum interruptible power of the owner is determined by the product of the preset interruption ratio coefficient and the user's basic load power corresponding to the power load prediction fusion value; the transferable power of the owner is determined by the product of the preset power transfer ratio coefficient and the user's daily power consumption corresponding to the power load prediction fusion value; and the power adjustment range of the owner is determined by the user's maximum power consumption function and the user's minimum power consumption power in the power load prediction fusion value. For each property owner, an interruptibility potential index is determined based on the owner's maximum interruptible power, the preset optimal value of the maximum interruptible power, and the preset worst value of the maximum interruptible power; a transferability potential index is determined based on the owner's transferable power, the preset optimal value of the transferable power, and the preset worst value of the transferable power; an adjustment potential index is determined based on the owner's power adjustment range, the preset optimal value of the power adjustment range, and the preset worst value of the power adjustment range; and the owner's interruptibility potential index, transferability potential index, and adjustment potential index are combined to determine the owner's comprehensive power load adjustment potential.

[0038] In this embodiment, the maximum interruptible power is determined by multiplying a preset interruption ratio coefficient by the user's basic load power, which considers both the user's actual electricity consumption characteristics and ensures the safety of interruption operations. The transferable electricity is calculated by multiplying the electricity transfer ratio coefficient by the daily electricity consumption, effectively quantifying the load's spatiotemporal transfer capability. The power adjustment range is determined using the difference between the user's maximum and minimum electricity consumption, intuitively reflecting the load fluctuation elasticity. Furthermore, a preset optimal / worst value is introduced as a benchmark, transforming each dimension's indicators into a standardized potential index, avoiding evaluation bias caused by differences in dimensions. Finally, through the fusion calculation of the potential index, complementary information from the three dimensions of interruptibility, transferability, and adjustment range is integrated, overcoming the one-sidedness of single-indicator evaluation and improving the robustness of the evaluation results through multi-source data fusion. Specifically, the comprehensive load regulation potential is a weighted sum of the interruptibility potential index, the transferability potential index, and the adjustability potential index; the interruptibility potential index = (maximum interruptible power - preset optimal maximum interruptible power) / (preset worst maximum interruptible power - preset optimal maximum interruptible power); the transferability potential index = (transferable quantity - preset optimal transferable quantity) / (preset worst transferable quantity - preset optimal transferable quantity); the adjustability potential index = (power regulation range - preset optimal power regulation range) / (preset worst power regulation range - preset optimal power regulation range). Each power regulation range is mapped to a specific value before calculation. This evaluation method not only provides a quantitative basis for the load regulation capacity of the power system owner, supporting the formulation of demand response strategies and the optimized scheduling of virtual power plants, but also enables refined management and efficient utilization of regulation resources through dynamic tracking of the potential indices, ultimately improving the flexibility and economy of the power system.

[0039] Example 5 In another embodiment of this application, the method may optionally further include: obtaining the owner scope corresponding to the virtual power plant; determining the owner adjustable potential of each owner based on the sum of the interruptibility potential index, transferability potential index and adjustable potential index of each owner within the owner scope; and determining the total adjustable potential corresponding to the virtual power plant based on the sum of the owner adjustable potentials of all owners within the owner scope.

[0040] In this embodiment, a complete evaluation system from micro to macro is constructed by quantifying the individual adjustment potential of power plant owners and aggregating it into the overall potential of the virtual power plant. Specifically, firstly, the owner's adjustable potential is obtained by integrating the sum of the potential indices of three dimensions: interruptibility, transferability, and adjustability. This weighted fusion method avoids the limitations of a single indicator and can comprehensively reflect the owner's load adjustment capability. Furthermore, the adjustable potential of all owners within the owner's scope is accumulated to form the total adjustable potential of the virtual power plant, realizing a quantitative mapping from individual to overall. The beneficial effects of this evaluation method are: on the one hand, it provides virtual power plant operators with an intuitive assessment of the total amount of adjustment resources, facilitating the formulation of global demand response strategies and scheduling plans; on the other hand, through the linkage analysis of individual potential and overall potential, key adjustment resource nodes can be identified, optimizing resource allocation efficiency. In addition, dynamically tracking the changes in the adjustable potential of each owner and the virtual power plant as a whole can also support the real-time monitoring of the power system's load adjustment capability, improve the grid's adaptability to renewable energy fluctuations, and ultimately promote the safe and stable operation and economically efficient dispatch of the power system.

[0041] Example 6 Optionally, in another embodiment of this application, the method further includes: Based on the combined electricity load forecast values ​​of each property owner, typical electricity consumption patterns of each property owner are clustered. For each typical electricity consumption pattern, the interruptibility potential index, transferability potential index, and adjustability potential index of each owner in the typical electricity consumption pattern are clustered to obtain multiple interruptibility potential index classes, multiple transferability potential index classes, and multiple adjustability potential index classes. Based on each cluster, each owner in the typical electricity consumption pattern is divided into different levels of interruptibility potential owners, different levels of transferability potential owners, and different levels of adjustability potential owners. Based on the real-time load characteristics corresponding to the virtual power plant and the interruptibility potential level, transferability potential level, and adjustability potential level of each owner within the owner's scope, power dispatch is carried out for each owner.

[0042] In this embodiment, a multi-level clustering and dynamic scheduling strategy optimizes the entire process from electricity consumption behavior pattern identification to potential level classification and precise scheduling. Specifically, firstly, based on the fusion value of electricity load forecasts, typical electricity consumption patterns of property owners are clustered, revealing the electricity consumption behavior characteristics of different owners (such as commercial bi-peak patterns, industrial stable patterns, etc.), providing a behavioral pattern basis for subsequent potential assessment. Then, under the same electricity consumption patterns, a second clustering is performed on the interruptibility, transferability, and adjustability potential indices, further subdividing potential owners into different levels (such as high / medium / low interruptibility potential levels). This "pattern-level" dual clustering preserves the common characteristics of electricity consumption behavior while highlighting differences in adjustment capabilities. Finally, scheduling is performed by combining the real-time load characteristics of the virtual power plant with the owner's potential level, enabling "on-demand dispatch and precise matching." For example, during peak grid periods, priority is given to dispatching owners with high interruptibility potential, while during periods of renewable energy fluctuations, owners with high transferability potential are dispatched for load spatiotemporal balancing. The beneficial effects of this strategy are as follows: it improves the scenario adaptability of potential assessment by clustering electricity consumption patterns, avoiding a "one-size-fits-all" assessment; it enables refined hierarchical management of regulatory resources through potential level classification, improving the targeting of dispatch; and it enhances the system's response capability to emergencies by combining dynamic dispatch with real-time load characteristics, ultimately improving the operating efficiency and economy of virtual power plants and supporting the power system to achieve more flexible and reliable load management and resource optimization.

[0043] Furthermore, the virtual power plant can automatically track the target load curve through an event-driven stimulus-feedback mechanism. Its core is to construct a deep reinforcement learning (DRL) model, using the load tracking error e(t) = L_target(t) - L_actual(t) and the incentive cost C(t) as the reward function. R(t) = -[λ1 × e(t)² + λ2 × C(t)], where λ1 and λ2 are weighting coefficients, L_target(t) and L_actual(t) represent the target load value and actual load value at time t, respectively. Perfect tracking is indicated when e(t) = 0; e(t) > 0 indicates that the actual load is lower than the target, requiring increased output; e(t) < 0 indicates that the actual load is higher than the target, requiring reduced output. The incentive cost C(t) includes: interruptible load compensation cost, transferable load incentive cost, adjustable load compensation cost, and dispatch operation cost. Through DRL training, the virtual power plant can learn to maximize tracking accuracy while minimizing cost. To achieve accurate and economical tracking of the target load curve, this invention designs a precise dispatch strategy based on user potential grouping. The dispatch engine prioritizes the use of adjustable resources from the user group whose deviation characteristics best match the target load deviation e(t) at different time periods. For example, when short-term peak loads occur, the system will prioritize triggering the interruptible loads of the "commercial bi-peak type - high interruption potential" user group; when long-term load shifting is required, the system will primarily schedule the energy storage equipment of the "industrial stable type - high transfer potential" user group. Based on this, an adaptive learning mechanism can be introduced to dynamically optimize the calling strategy and incentive signals for each user group by analyzing historical scheduling instructions and tracking results, thereby achieving a comprehensive optimization of tracking accuracy and incentive costs.

[0044] Furthermore, as Figure 1 In terms of specific implementation, this application provides a device for assessing electrical load regulation potential based on multi-model fusion, such as... Figure 3 As shown, the device includes: The feature construction module is used to obtain the predicted electricity load data of each owner among multiple owners, and construct the predicted electricity load features based on the predicted electricity load data, wherein the predicted electricity load features include electricity load features and time period features. The load forecasting module is used to input the electrical load characteristics to be predicted into a multi-model fusion model, and to perform electrical load forecasting on the electrical load characteristics to be predicted by each electrical load forecasting model in the multi-model fusion model, and to fuse the results predicted by each electrical load forecasting model to obtain a fused electrical load forecast value; wherein, the electrical load forecasting model includes XGBoost model, random forest model, support vector regression model and generalized additive model; The potential assessment module is used to determine the owner's maximum interruptible power, transferable power, and power adjustment range based on the power load prediction fusion value, and to determine the owner's comprehensive power load adjustment potential based on the maximum interruptible power, transferable power, and power adjustment range.

[0045] Optionally, the apparatus further includes: a model training module, used for: Obtain electrical load training samples, and use the electrical load training samples to train each electrical load prediction model respectively; The fusion weights of each load prediction model are optimized using the Bayesian optimization method to obtain the fusion weights of each load prediction model. Accordingly, the load forecasting module is also used for: The results predicted by each load prediction model are fused based on the fusion weights of each load prediction model to obtain the fused load prediction value.

[0046] Optionally, the model training module is further configured to: With the goal of minimizing the mean square error of the multi-model fusion model, the fusion weights of each electric load prediction model are optimized using Bayesian optimization methods based on the electric load validation samples to obtain the fusion weights of each electric load prediction model.

[0047] Optionally, the load forecasting module is further configured to: The electrical load prediction fusion values ​​are grouped based on a preset threshold, and the electrical load prediction fusion values ​​are compressed based on the mean of the electrical load prediction fusion values ​​within the group, so as to update the electrical load prediction fusion values ​​based on the compression results.

[0048] Optionally, the potential assessment module is further configured to: The maximum interruptible power of the owner is determined by the product of the preset interruption ratio coefficient and the user's basic load power corresponding to the power load prediction fusion value; the transferable power of the owner is determined by the product of the preset power transfer ratio coefficient and the user's daily power consumption corresponding to the power load prediction fusion value; and the power adjustment range of the owner is determined by the user's maximum power consumption function and the user's minimum power consumption power in the power load prediction fusion value. For each property owner, an interruptibility potential index is determined based on the owner's maximum interruptible power, the preset optimal value of the maximum interruptible power, and the preset worst value of the maximum interruptible power; a transferability potential index is determined based on the owner's transferable power, the preset optimal value of the transferable power, and the preset worst value of the transferable power; an adjustment potential index is determined based on the owner's power adjustment range, the preset optimal value of the power adjustment range, and the preset worst value of the power adjustment range; and the owner's interruptibility potential index, transferability potential index, and adjustment potential index are combined to determine the owner's comprehensive power load adjustment potential.

[0049] Optionally, the potential assessment module is further configured to: Obtain the owner scope corresponding to the virtual power plant; Based on the sum of the interruptibility potential index, transferability potential index, and adjustability potential index of each owner within the owner scope, the owner's adjustability potential is determined, and the total adjustability potential corresponding to the virtual power plant is determined based on the sum of the owner's adjustability potential of all owners within the owner scope.

[0050] Optionally, the potential assessment module is further configured to: Based on the combined electricity load forecast values ​​of each property owner, typical electricity consumption patterns of each property owner are clustered. For each typical electricity consumption pattern, the interruptibility potential index, transferability potential index, and adjustability potential index of each owner in the typical electricity consumption pattern are clustered to obtain multiple interruptibility potential index classes, multiple transferability potential index classes, and multiple adjustability potential index classes. Based on each cluster, each owner in the typical electricity consumption pattern is divided into different levels of interruptibility potential owners, different levels of transferability potential owners, and different levels of adjustability potential owners. Based on the real-time load characteristics corresponding to the virtual power plant and the interruptibility potential level, transferability potential level, and adjustability potential level of each owner within the owner's scope, power dispatch is carried out for each owner.

[0051] It should be noted that other corresponding descriptions of the functional units involved in the multi-model fusion-based electrical load regulation potential assessment device provided in this application embodiment can be found in the following references. Figures 1 to 2 The corresponding descriptions in the method will not be repeated here.

[0052] This application also provides a computer device, specifically a personal computer, server, network device, etc. The computer device includes a bus, processor, memory, and communication interface, and may also include input / output interfaces and a display device. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores location information. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0053] Those skilled in the art will understand that the structure of the computer device described above is only a partial structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. A specific computer device may include more or fewer components, or combine certain components, or have different component arrangements.

[0054] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0055] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0056] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0057] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0058] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for assessing the electrical load regulation potential based on multi-model fusion, characterized in that, include: For each of the multiple property owners, the owner's predicted electricity load data is obtained. Based on the predicted electricity load data, a predicted electricity load feature is constructed, wherein the predicted electricity load feature includes electricity load characteristics and time period characteristics. The predicted electricity load feature is input into a multi-model fusion model. Each electricity load prediction model in the multi-model fusion model performs electricity load prediction on the predicted electricity load feature, and the prediction results of each electricity load prediction model are fused to obtain a fused electricity load prediction value. The electricity load prediction model includes XGBoost model, random forest model, support vector regression model, and generalized additive model. Based on the combined electrical load forecast values, the owner's maximum interruptible power, transferable power, and power adjustment range are determined, and the owner's comprehensive electrical load adjustment potential is also determined.

2. The method according to claim 1, characterized in that, Before inputting the electrical load characteristics to be predicted into the multi-model fusion model, the method further includes: Obtain electrical load training samples, and use the electrical load training samples to train each electrical load prediction model respectively; The fusion weights of each load prediction model are optimized using the Bayesian optimization method to obtain the fusion weights of each load prediction model.

3. The method according to claim 2, characterized in that, The results predicted by each electrical load forecasting model are fused to obtain a fused electrical load forecast value, including: The results predicted by each load prediction model are fused based on the fusion weights of each load prediction model to obtain the fused load prediction value.

4. The method according to claim 3, characterized in that, The process of optimizing the fusion weights of each electricity load prediction model using a Bayesian optimization method to obtain the fusion weights of each electricity load prediction model includes: With the goal of minimizing the mean square error of the multi-model fusion model, the fusion weights of each electric load prediction model are optimized using Bayesian optimization methods based on the electric load validation samples to obtain the fusion weights of each electric load prediction model.

5. The method according to claim 1, characterized in that, After obtaining the combined predicted electrical load value, the method further includes: The electrical load prediction fusion values ​​are grouped based on a preset threshold, and the electrical load prediction fusion values ​​are compressed based on the mean of the electrical load prediction fusion values ​​within the group, so as to update the electrical load prediction fusion values ​​based on the compression results.

6. The method according to claim 1, characterized in that, Based on the combined electrical load forecast values, the owner's maximum interruptible power, transferable power, and power regulation range are determined respectively. Furthermore, based on the maximum interruptible power, transferable power, and power regulation range, the owner's comprehensive electrical load regulation potential is determined, including: The maximum interruptible power of the owner is determined by the product of the preset interruption ratio coefficient and the user's basic load power corresponding to the power load prediction fusion value; the transferable power of the owner is determined by the product of the preset power transfer ratio coefficient and the user's daily power consumption corresponding to the power load prediction fusion value; and the power adjustment range of the owner is determined by the user's maximum power consumption function and the user's minimum power consumption power in the power load prediction fusion value. For each property owner, an interruptibility potential index is determined based on the owner's maximum interruptible power, the preset optimal value of the maximum interruptible power, and the preset worst value of the maximum interruptible power; a transferability potential index is determined based on the owner's transferable power, the preset optimal value of the transferable power, and the preset worst value of the transferable power; an adjustment potential index is determined based on the owner's power adjustment range, the preset optimal value of the power adjustment range, and the preset worst value of the power adjustment range; and the owner's interruptibility potential index, transferability potential index, and adjustment potential index are combined to determine the owner's comprehensive power load adjustment potential.

7. The method according to claim 6, characterized in that, The method further includes: Obtain the owner scope corresponding to the virtual power plant; Based on the sum of the interruptibility potential index, transferability potential index, and adjustability potential index of each owner within the owner scope, the owner's adjustability potential is determined, and the total adjustability potential corresponding to the virtual power plant is determined based on the sum of the owner's adjustability potential of all owners within the owner scope.

8. The method according to claim 7, characterized in that, The method further includes: Based on the combined electricity load forecast values ​​of each property owner, typical electricity consumption patterns of each property owner are clustered. For each typical electricity consumption pattern, the interruptibility potential index, transferability potential index, and adjustability potential index of each owner in the typical electricity consumption pattern are clustered to obtain multiple interruptibility potential index classes, multiple transferability potential index classes, and multiple adjustability potential index classes. Based on each cluster, each owner in the typical electricity consumption pattern is divided into different levels of interruptibility potential owners, different levels of transferability potential owners, and different levels of adjustability potential owners. Based on the real-time load characteristics corresponding to the virtual power plant and the interruptibility potential level, transferability potential level, and adjustability potential level of each owner within the owner's scope, power dispatch is carried out for each owner.

9. A device for assessing the electrical load regulation potential based on multi-model fusion, characterized in that, The device includes: The feature construction module is used to obtain the predicted electricity load data of each owner among multiple owners, and construct the predicted electricity load features based on the predicted electricity load data, wherein the predicted electricity load features include electricity load features and time period features. The load forecasting module is used to input the electrical load characteristics to be predicted into a multi-model fusion model, and to perform electrical load forecasting on the electrical load characteristics to be predicted by each electrical load forecasting model in the multi-model fusion model, and to fuse the results predicted by each electrical load forecasting model to obtain a fused electrical load forecast value; wherein, the electrical load forecasting model includes XGBoost model, random forest model, support vector regression model and generalized additive model; The potential assessment module is used to determine the owner's maximum interruptible power, transferable power, and power adjustment range based on the power load prediction fusion value, and to determine the owner's comprehensive power load adjustment potential based on the maximum interruptible power, transferable power, and power adjustment range.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 8.

11. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 8.

Images