Explainable daily load curve forecasting method based on adaptive reference curve optimization

Abstract

The application discloses an interpretable daily load curve prediction method based on adaptive reference curve optimization, which comprises the following steps: selecting load characteristic indexes with physical significance; for each load characteristic index, input feature selection is performed by using a random forest algorithm, and the load characteristic index is predicted by combining a gated recurrent unit network to obtain a load characteristic index of a day to be predicted; based on the load characteristic index of the day to be predicted, an adaptive cosine distance nearest neighbor algorithm is used to select a historical daily load curve similar to a daily load curve shape of the day to be predicted from a historical daily load curve library, and an adaptive reference daily load curve is constructed; and based on the adaptive reference daily load curve, the load characteristic index of the day to be predicted is introduced as a constraint condition, and an optimization algorithm is used to predict a daily load curve, so that a daily load curve prediction result of the day to be predicted is finally obtained. The application can guarantee prediction accuracy and enhance the interpretability of the prediction process.

Description

Technical Field

[0001] This invention belongs to the field of power systems, specifically relating to an interpretable daily load curve prediction method based on adaptive baseline curve optimization. Background Technology

[0002] With the rapid development of power systems, load forecasting plays a crucial role in power system dispatching, planning, and safe operation. Traditional load forecasting methods mainly rely on statistical models. While these methods can provide a certain level of forecast accuracy, they are unable to effectively cope with complex load changes and high volatility issues as the electricity market evolves. In recent years, artificial intelligence-based forecasting models, especially those based on deep learning and recurrent neural networks, have improved forecast accuracy but face the "black box" problem, lacking interpretability.

[0003] Currently, interpretability of complex models, especially in load curve forecasting, remains a significant challenge. To improve forecasting accuracy while maintaining model interpretability, increasing research is focusing on interpretable artificial intelligence methods. Based on this, this invention proposes an interpretable daily load curve forecasting method based on adaptive baseline curve optimization, aiming to address the balance between accuracy and interpretability in existing load forecasting models. Summary of the Invention

[0004] The main objective of this invention is to provide an interpretable daily load curve prediction method based on adaptive baseline curve optimization, aiming to improve the accuracy and interpretability of load forecasting. This method introduces a feature-based adaptive baseline curve optimization mechanism, combined with artificial intelligence techniques such as random forests and gated recursive units, to effectively predict daily load curves and exhibit strong interpretability.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides an interpretable daily load curve prediction method based on adaptive baseline curve optimization, the method comprising: Step 1: Select load characteristic indicators with physical meaning. For each load characteristic indicator, use the random forest algorithm to select input features and combine it with a gated recurrent unit network to predict the load characteristic indicator to obtain the load characteristic indicator for the day to be predicted. Step 2: Based on the load characteristic indicators of the day to be predicted, the adaptive cosine distance nearest neighbor algorithm is used to select historical day load curves with similar shapes to the day to be predicted from the historical day load curve library to construct an adaptive base day load curve; Step 3: Based on the adaptive reference day load curve, the load characteristic index of the day to be predicted is introduced as a constraint condition. The daily load curve is predicted by an optimization algorithm, and the prediction result of the daily load curve of the day to be predicted is finally obtained.

[0006] Preferably, the load characteristic indicators include the daily maximum load factor, the daily load factor, and the daily minimum load factor, and the calculation formula is as follows: ; ; ; In the formula: For date; For a specific moment; for Daily maximum load factor; for The actual load at that time; for Daily load factor; for Daily minimum load factor; and These represent the maximum and minimum values, respectively.

[0007] Preferably, for each load characteristic index, a random forest algorithm is used to select input features, and a gated recurrent unit network is used to predict the load characteristic index, specifically including: Input features include historical load data, meteorological data, and date type information; For each load feature index, the random forest model uses multiple decision trees to statistically analyze the importance of features. Based on the contribution of features in the node splitting process, the input features are ranked, and the features with the highest importance for each load feature index are selected as the input to the subsequent gated recurrent unit network. The gated cyclic unit network uses update gate and reset gate structures to predict each load characteristic index based on the input characteristics corresponding to each load characteristic index.

[0008] Preferably, step 2 includes: The historical daily load curves are normalized, and a historical daily load curve library is constructed. The load characteristic index is concatenated with the input features selected by the random forest algorithm to form a feature vector for measuring similarity; Obtain the target feature vector for the day to be predicted, calculate the cosine distance between the target feature vector and the feature vectors of each historical day load curve in the historical day load curve library, and select several historical day load curves with the smallest distance. The weight of each selected historical daily load curve is calculated using a soft maximum function, and the adaptive baseline daily load curve for the day to be predicted is obtained by weighted averaging.

[0009] Preferably, the weight of each selected historical daily load curve is calculated using a soft maximum function, and the adaptive baseline load curve for the day to be predicted is obtained by weighted averaging, including: Obtain the cosine distance between the eigenvectors of several selected historical daily load curves and the target eigenvector of the day to be predicted, and calculate the weight of each selected historical daily load curve using a soft maximum function. ; In the formula: The number of historical daily load curves selected. For the first The weights of historical daily load curves; The target date for the predicted date. For the first The cosine distance between the eigenvectors of historical daily load curves and the target eigenvector of the day to be predicted; Based on the weights of each selected historical daily load curve, an adaptive baseline load curve for the day to be predicted is obtained through weighted averaging: ; In the formula: To adapt the baseline day load curve Normalized value of hourly load; For the first Historical daily load curves Normalized value of hourly load.

[0010] Preferably, step 2 also includes: The number of selected historical daily load curves is adaptively determined based on the goodness-of-fit index, as follows: Calculate the goodness-of-fit index under different quantity conditions: ; In the formula: The goodness-of-fit index; To adapt the baseline day load curve Normalized value of hourly load; After normalization The actual load value at that time; Daily load factor; Based on the goodness-of-fit index under different quantity conditions, the optimal number of historical daily load curves is determined.

[0011] Preferably, step 3 includes: The objective function is to minimize the modification amount of the adaptive base date load curve, specifically satisfying the following: ; In the formula: For date, The objective function is... To be predicted Normalized value of hourly load; To adapt the baseline day load curve Normalized value of hourly load; The load characteristic indicators of the day to be predicted are introduced as constraints, including: The predicted maximum value of the daily load curve satisfies the predicted daily maximum load factor; The average value of the predicted daily load curve satisfies the predicted daily load factor; The minimum value of the predicted daily load curve satisfies the predicted daily minimum load factor; The prediction problem is transformed into an optimization problem, and the daily load curve for the day to be predicted is obtained by solving the optimization algorithm.

[0012] Preferably, the optimization algorithm includes the Gurobi commercial solver.

[0013] In a second aspect, the present invention provides a computer device / apparatus / system, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0014] Thirdly, the present invention provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0015] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: This invention decomposes the daily load curve prediction process into three stages: prediction of load characteristic indicators with clear physical meaning, construction of an adaptive reference daily load curve, and constraint optimization solution. This effectively avoids the error accumulation problem in traditional hourly prediction methods and significantly enhances the interpretability of the prediction process while ensuring prediction accuracy. Furthermore, the adaptive reference curve method based on cosine distance can dynamically characterize the changes in the shape of the daily load curve, overcoming the shortcomings of traditional clustering methods in adapting to the shape of the load curve. In addition, by introducing physical constraints such as the daily maximum load factor, daily load factor, and daily minimum load factor during the load curve optimization process, the prediction results conform to the operating laws of the power system and have clear engineering significance, thus achieving an effective unification of prediction accuracy, model interpretability, and engineering feasibility. Attached Figure Description

[0016] Figure 1 This is a diagram illustrating an interpretable daily load curve prediction framework based on adaptive baseline curve optimization, according to an embodiment of the present invention. Figure 2 This is a graph showing the daily maximum load input filtering results according to an embodiment of the present invention; Figure 3 This is a graph showing the daily load rate input filtering results according to an embodiment of the present invention; Figure 4 This is a graph showing the daily minimum load rate input filtering results according to an embodiment of the present invention; Figure 5 This is a graph showing the daily maximum load prediction results according to an embodiment of the present invention; Figure 6 This is a graph showing the daily load factor prediction results according to an embodiment of the present invention; Figure 7 This is a graph showing the predicted daily minimum load rate according to an embodiment of the present invention; Figure 8 This is a graph showing the predicted daily load curve according to an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments provided by this invention without inventive effort are within the scope of protection of this invention.

[0018] Obviously, the accompanying drawings described below are merely some examples or embodiments of the present invention. Those skilled in the art can apply the present invention to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this invention, modifications to design, manufacturing, or production based on the technical content disclosed in this invention are merely conventional technical means and should not be construed as insufficient disclosure of the present invention.

[0019] In this invention, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this invention may be combined with other embodiments without conflict.

[0020] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "a," "an," "an," "the," and similar words used in this invention do not indicate quantity limitation and may indicate singular or plural. The terms "comprising," "including," "having," and any variations thereof used in this invention are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms "connected," "linked," "coupled," and similar words used in this invention are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "A plurality" used in this invention refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships may exist; for example, "A and / or B" can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects have an "or" relationship. The terms "first," "second," and "third" used in this invention are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0021] This invention provides an interpretable daily load curve prediction method based on adaptive baseline curve optimization. The method first selects load characteristic indicators with clear physical meaning, such as the daily maximum load factor, daily load factor, and daily minimum load factor, and uses a random forest algorithm to filter input features, combined with a gated recursive unit network to predict these load characteristic indicators. Subsequently, based on the predicted load characteristic indicators, an adaptive baseline daily load curve is constructed from a historical daily load curve database using an adaptive nearest neighbor method based on cosine distance. Finally, based on the baseline curve, load characteristic indicator constraints are introduced, and the final daily load curve prediction result is obtained through an optimization algorithm. This invention enhances the interpretability of the prediction process while ensuring prediction accuracy, effectively characterizing the overall shape changes of the daily load curve, and is suitable for short-term load forecasting and operational decision support in power systems.

[0022] like Figure 1 As shown, the interpretable daily load curve prediction method based on adaptive baseline curve optimization of the present invention specifically includes the following steps: Step 1: Select three physically meaningful load characteristic indicators, including the daily maximum load factor, daily load factor, and daily minimum load factor, to reflect the characteristics of the daily load curve. For each load characteristic indicator, use the random forest algorithm for feature selection, and combine it with a gated recurrent unit network to predict these three load characteristic indicators.

[0023] Step 2: Combining the load characteristic index prediction results obtained in Step 1, the adaptive cosine distance nearest neighbor algorithm is used to select the most suitable benchmark curve from the historical daily load curve library. The adaptive cosine distance nearest neighbor algorithm calculates the similarity between the historical daily load curve and the daily load curve of the day to be predicted based on the cosine distance, and calculates the weight of each historical curve through the soft maximum method to obtain the benchmark curve that is most similar to the daily load curve of the day to be predicted.

[0024] Step 3: Based on the predicted load characteristic indicators obtained in Step 1 and the adaptive baseline curve obtained in Step 2, the final prediction of the daily load curve is performed using an optimization algorithm. The optimization objective is to minimize the mean square error of the objective function, which considers constraints such as the daily maximum load factor, daily load factor, and daily minimum load factor.

[0025] As a preferred embodiment, step 1 is described as follows: Step 11: Select load characteristic indicators, as follows: To mitigate the error accumulation problem introduced by directly predicting daily load curves hourly, the daily load curve is first characterized by selecting load characteristic indicators with clear physical meaning for prediction. These load characteristic indicators include at least the daily peak load (i.e., daily maximum load factor), daily load factor (i.e., daily load factor), and daily minimum load factor (i.e., daily minimum load factor), calculated using the following formulas: ; ; ; In the formula: For date; For a specific moment; for The daily maximum load; for The actual load at that time; for Daily load factor; for Daily minimum load factor; and These represent the maximum and minimum values, respectively.

[0026] Step 12: For each load characteristic index, a random forest model is used to filter the input features. The input features include, but are not limited to, historical load data, meteorological data, and date type information. The random forest model uses multiple decision trees to statistically analyze the importance of features, ranks the input features based on their contribution to the node splitting process, and selects several highly important features as inputs for subsequent prediction models. Specifically: For each load characteristic index, the three most important inputs are selected as follows: , , To make predictions separately.

[0027] See Figure 2 , Figure 3 and Figure 4 The random forest feature selection results show that the key input features corresponding to different load characteristic indicators are significantly different. Random forest can effectively identify meteorological and temporal features that have a significant impact on the daily maximum load factor, daily load factor, and daily minimum load factor, avoiding the problem that traditional linear correlation analysis cannot characterize nonlinear relationships. This provides a more reasonable and physically meaningful set of input features for subsequent prediction models. Figure 2 It can be seen that the most important influencing factor for daily maximum load is average temperature, followed by average humidity and average air pressure. Therefore, the inputs for predicting daily maximum load are average temperature, average humidity, and average air pressure. Figure 3 It can be seen that the most important influencing factor on daily load factor is average temperature, followed by average humidity and average air pressure. Therefore, the inputs for daily load factor prediction are average temperature, average humidity, and average air pressure. Figure 4 It can be seen that the most important influencing factor on the daily minimum load factor is the average temperature, followed by the average air pressure and the maximum temperature. Therefore, the input for daily load factor prediction is the average temperature, average air pressure, and maximum temperature.

[0028] Step 13: A gated recursive unit network is used to predict various load characteristic indicators. The gated recursive unit network utilizes its update and reset gate structures to model long-term dependencies in the time series data, thereby improving the prediction accuracy of load characteristic indicators. Specifically: The three load characteristic indicators for prediction are as follows: , , .

[0029] The prediction results of the gated cyclic unit network for load characteristic indicators show that the predicted curves can follow the actual indicator trends well, maintaining high consistency even during periods of large load fluctuations. This indicates that the gated cyclic unit network based on the selected features can accurately characterize the temporal variation patterns of load characteristic indicators. Figure 5 , Figure 6 , Figure 7 It can be seen that the gated cyclic unit network achieved the best prediction results for the daily maximum load, daily load rate, and daily minimum load rate, and had the highest degree of curve fitting, indicating that the model has significant advantages in time series fitting.

[0030] As a preferred embodiment, step 2, based on the load characteristic indicators obtained in step 1, introduces an adaptive cosine distance nearest neighbor method to construct an adaptive base day load curve, as detailed below: Step 21: Construct a historical daily load curve library and normalize the historical daily load data to eliminate the impact of load scale differences on curve similarity judgment, ensuring that the load curves only reflect morphological characteristics. Details are as follows: ; In the formula: After normalization The load value at that time.

[0031] Step 22: Concatenate the load feature index with the input features filtered by random forest to form a feature vector for measuring similarity. For the date to be predicted, the predicted load characteristic indicators and their corresponding characteristic vectors are used together as the target vector. .

[0032] It should be noted that the first half of the feature vector is the input feature selected by the random forest. In this embodiment, the union of the input features of three load characteristic indicators is taken, namely, average temperature, average humidity, average air pressure, and maximum temperature. The second half of the feature vector is the load characteristic indicator. Since the daily maximum load is normalized to 1, the feature vector does not include the daily maximum load.

[0033] Step 23: By calculating the cosine distance between the target feature vector and each historical vector in the historical load curve library, the similarity of different load curves in shape and feature space is measured. The cosine distance effectively characterizes the changing trend of the load curve without being affected by amplitude shifts. Specifically: ; In the formula: It is a vector with vector The cosine distance between them; It is the number of elements in the vector; It is a vector The Each element.

[0034] Step 24: Select several historical load curves with the smallest distance, and calculate the weight of each historical curve using a soft maximum function, so that curves with higher similarity receive greater weight. Specifically: ; In the formula: For the first The weights of the load curves; For the target date; This represents the number of load curves to be selected.

[0035] Step 25: Calculate a weighted average of the selected historical load curves according to the weights to obtain the adaptive baseline load curve. By performing inverse normalization on the weighted result, the baseline curve for subsequent optimization can be obtained. ; In the formula: Reference load curve Normalized value of hourly load; For the first a curve Normalized value of hourly load.

[0036] Step 26: Calculate the goodness-of-fit index under different numbers of curves, and adaptively determine the number of historical curves selected to achieve a balance between computational efficiency and fitting accuracy. Specifically: ; In the formula: This is a goodness-of-fit index.

[0037] Therefore, based on the goodness-of-fit index under different quantity conditions, the optimal number of historical daily load curves can be determined.

[0038] As a preferred embodiment, the optimization process for the adaptive baseline load curve described in step 3 is as follows: Step 31: The objective function is to minimize the modification of the baseline load curve, specifically satisfying: ; In the formula: The objective function is... To be predicted Normalized value of hourly load.

[0039] Step 32: The predicted maximum value of the daily load curve satisfies the predicted daily maximum load factor. Since the optimization process is under normalization, the optimization constraint is that the maximum value is 1, specifically satisfying: ; Step 33: The average value of the predicted daily load curve satisfies the predicted daily load rate, specifically: ; Step 34: The minimum value of the predicted daily load curve satisfies the predicted daily minimum load rate, specifically: ; Step 35: Solve using the Gurobi commercial solver. After optimization, perform inverse normalization to convert the load forecast values ​​back to their original dimensions, specifically satisfying: ; In the formula: for Hourly load forecast.

[0040] This step, based on the predicted load characteristic indices obtained in step 1, corrects the adaptive baseline curve obtained in step 2 to obtain the final predicted value, transforming the prediction problem into an optimization problem. In this process, the objective function is to minimize the correction amount, and the predicted load characteristic indices are considered as constraints. The final solution is obtained using the Gurobi commercial solver.

[0041] Finally, taking Hubei Province's power and meteorological data as an example, we compare and study the daily load curve prediction results. Figure 8 As shown, the daily load curve fitting results obtained from the proposed optimized prediction method are highly consistent with the actual load curve in terms of overall shape, peak position and valley level. In particular, it has a good fitting effect in terms of peak-valley changes and intraday fluctuation characteristics. This indicates that by introducing an adaptive reference curve and combining it with load characteristic index constraints for optimization, the prediction accuracy of the daily load curve can be effectively improved while ensuring the rationality of the curve shape.

[0042] In summary, this invention decomposes the daily load curve prediction process into three stages: prediction of load characteristic indicators with clear physical meaning, construction of an adaptive reference daily load curve, and constraint optimization solution. This effectively avoids the error accumulation problem in traditional hourly prediction methods and significantly enhances the interpretability of the prediction process while ensuring prediction accuracy. Furthermore, the adaptive reference curve method based on cosine distance can dynamically characterize the changes in the shape of the daily load curve, overcoming the shortcomings of traditional clustering methods in adapting to the shape of the load curve. In addition, by introducing physical constraints such as the daily maximum load factor, daily load factor, and daily minimum load factor during the load curve optimization process, the prediction results conform to the operating laws of the power system and have clear engineering significance, thus achieving an effective unification of prediction accuracy, model interpretability, and engineering feasibility.

[0043] Furthermore, the present invention also provides a computer device / apparatus / system, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0044] The present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0045] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or block diagrams.

[0046] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more blocks in a block diagram.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, thereby providing steps for implementing the functions specified in one or more flowcharts and / or one or more blocks in a block diagram.

[0048] It should be noted that, depending on the implementation needs, the various steps / components described in this invention can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0049] Those skilled in the art will readily understand that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting interpretable daily load curves based on adaptive baseline curve optimization, characterized in that, The method includes: Step 1: Select load characteristic indicators with physical meaning. For each load characteristic indicator, use the random forest algorithm to select input features and combine it with a gated recurrent unit network to predict the load characteristic indicator to obtain the load characteristic indicator for the day to be predicted. Step 2: Based on the load characteristic indicators of the day to be predicted, the adaptive cosine distance nearest neighbor algorithm is used to select historical day load curves with similar shapes to the day to be predicted from the historical day load curve library to construct an adaptive base day load curve; Step 3: Based on the adaptive reference day load curve, the load characteristic index of the day to be predicted is introduced as a constraint condition. The daily load curve is predicted by an optimization algorithm, and the prediction result of the daily load curve of the day to be predicted is finally obtained.

2. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 1, characterized in that, The load characteristic indicators include the daily maximum load factor, the daily load factor, and the daily minimum load factor, and the calculation formula is as follows: ； ； ； In the formula: For date; For a specific moment; for Daily maximum load factor; for The actual load at that time; for Daily load factor; for Daily minimum load factor; and These represent the maximum and minimum values, respectively.

3. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 1, characterized in that, For each load characteristic index, a random forest algorithm is used to select input features, and a gated recurrent unit network is combined to predict the load characteristic index, specifically including: Input features include historical load data, meteorological data, and date type information; For each load feature index, the random forest model uses multiple decision trees to statistically analyze the importance of features. Based on the contribution of features in the node splitting process, the input features are ranked, and the features with the highest importance for each load feature index are selected as the input to the subsequent gated recurrent unit network. The gated cyclic unit network uses update gate and reset gate structures to predict each load characteristic index based on the input characteristics corresponding to each load characteristic index.

4. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 1, characterized in that, Step 2 includes: The historical daily load curves are normalized, and a historical daily load curve library is constructed. The load characteristic index is concatenated with the input features selected by the random forest algorithm to form a feature vector for measuring similarity; Obtain the target feature vector for the day to be predicted, calculate the cosine distance between the target feature vector and the feature vectors of each historical day load curve in the historical day load curve library, and select several historical day load curves with the smallest distance. The weight of each selected historical daily load curve is calculated using a soft maximum function, and the adaptive baseline daily load curve for the day to be predicted is obtained by weighted averaging.

5. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 4, characterized in that, The weight of each selected historical daily load curve is calculated using a soft maximum function, and an adaptive baseline load curve for the day to be predicted is obtained through weighted averaging, including: Obtain the cosine distance between the eigenvectors of several selected historical daily load curves and the target eigenvector of the day to be predicted, and calculate the weight of each selected historical daily load curve using a soft maximum function. ； In the formula: The number of historical daily load curves selected. For the first The weights of historical daily load curves; The target date for the predicted date. For the first The cosine distance between the eigenvectors of historical daily load curves and the target eigenvector of the day to be predicted; Based on the weights of each selected historical daily load curve, an adaptive baseline load curve for the day to be predicted is obtained through weighted averaging: ； In the formula: To adapt the baseline day load curve Normalized value of hourly load; For the first Historical daily load curves Normalized value of hourly load.

6. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 1, characterized in that, Step 2 also includes: The number of selected historical daily load curves is adaptively determined based on the goodness-of-fit index, as follows: Calculate the goodness-of-fit index under different quantity conditions: ； In the formula: The goodness-of-fit index; To adapt the baseline day load curve Normalized value of hourly load; After normalization The actual load value at that time; Daily load factor; Based on the goodness-of-fit index under different quantity conditions, the optimal number of historical daily load curves is determined.

7. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 2, characterized in that, Step 3 includes: The objective function is to minimize the modification amount of the adaptive base date load curve, specifically satisfying the following: ； In the formula: For date, The objective function is... To be predicted Normalized value of hourly load; To adapt the baseline day load curve Normalized value of hourly load; The load characteristic indicators of the day to be predicted are introduced as constraints, including: The predicted maximum value of the daily load curve satisfies the predicted daily maximum load factor; The average value of the predicted daily load curve satisfies the predicted daily load factor; The minimum value of the predicted daily load curve satisfies the predicted daily minimum load factor; The prediction problem is transformed into an optimization problem, and the daily load curve for the day to be predicted is obtained by solving the optimization algorithm.

8. The interpretable daily load curve prediction method based on adaptive baseline curve optimization according to claim 1, characterized in that, The optimization algorithm includes the Gurobi commercial solver.

9. A computer device / equipment / system, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 8.

Images