A method and system for identifying key drivers of power imbalance funds
By integrating XGBoost and SHAP algorithms, a key driver identification method for power imbalance funds was constructed, which solved the problem of identifying and attributing imbalance funds in the power market, and improved prediction accuracy and risk management capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID IOT E-COMMERCE CO LTD
- Filing Date
- 2026-03-26
- Publication Date
- 2026-07-14
AI Technical Summary
In the electricity spot market, the analysis of the formation mechanism of imbalance funds and the quantitative attribution of responsibility are not yet mature, and there is a lack of systematic data-driven methods, which leads to market behavior distortion and price signal failure.
By employing the XGBoost machine learning algorithm and SHAP interpretability analysis, a structured dataset of multi-source power data and a key driver identification system are constructed. Through preprocessing, model training, and interpreting the model to generate identification results, accurate early warning and proactive risk management of unbalanced funds are achieved.
It improves the accuracy of predicting the scale of unbalanced funds, enables attribution analysis of various influencing factors, and provides data-driven decision support for the stable and healthy development of the power market.
Smart Images
Figure CN122390776A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of power market technology, and in particular to a method and system for identifying key driving factors of power imbalance funding. Background Technology
[0002] In the context of the electricity spot market, the parallel existence of planned and market-based systems has led to discrepancies in power generation and consumption, price differences, and market volatility, highlighting the problem of imbalanced funds and making it a key factor affecting market stability and healthy development. Simultaneously, with the accelerated entry of new energy sources into the market and the full implementation of mechanism-based electricity pricing policies, the electricity market structure and pricing mechanism are becoming increasingly complex, introducing new uncertainties into the formation of imbalanced funds. If imbalanced funds lack systematic analysis and tracing, it will be difficult to detect and address abnormal deviations in a timely manner, easily leading to distortions in market behavior and weakening the effectiveness of market price signals.
[0003] Currently, research on imbalanced funding mainly focuses on the analysis of its formation mechanism, policy and institutional design, and optimization of deviation assessment rules, lacking in-depth exploration of systematic attribution of complex, multi-source data. Furthermore, the identification of the causes of imbalanced funding in the electricity market and the quantification of responsibility are still in their initial stages, and a mature technical approach has not yet been developed. Summary of the Invention
[0004] The purpose of this application is to provide a method and system for identifying key driving factors of power imbalance funds in order to solve the above-mentioned technical problems, and to provide data-driven decision support for accurate early warning and proactive risk management of power imbalance funds.
[0005] In some embodiments of this application, by integrating the XGBoost machine learning algorithm with SHAP interpretability analysis, a structured dataset covering multi-source power data and a key driver identification system are constructed, which effectively improves the prediction accuracy of the scale of power imbalance funds, realizes the attribution analysis of various influencing factors in the formation of imbalance funds, and provides data-driven decision support for accurate early warning and proactive risk management of power imbalance funds.
[0006] In some embodiments of this application, a method for identifying key driving factors of electricity imbalance funding is provided, including: Acquire multidimensional historical data and generate a structured dataset based on the preprocessing results of the multidimensional historical data; Based on the preset basic prediction model and structured dataset, a model training strategy is set, and a power imbalance funding prediction model is established based on the model training strategy. The power imbalance funding prediction model outputs a prediction dataset based on real-time collected data, and generates recognition results based on a preset interpretation model and the prediction dataset.
[0007] In some embodiments of this application, a preset basic prediction model is included, including: Let the time series dataset for the first T time steps be (y t-1 y t-2 , ..., y t-N The current data is y. t The recognition result is ; Building a basic prediction model: Where n is time, K is the number of tree models to obtain the best fit target, F is the set of regression trees of the model, f(y) and w(y) are the leaf weights of each tree, q is the number of independent numbers in the model, and Q is the number of nodes in the leaves.
[0008] In some embodiments of this application, the establishment of the power imbalance funding prediction model includes: Training and test data packages are defined based on structured data; The basic prediction model is trained using training data packages; Output the model to be verified based on the training results, and generate the verification parameter set of the model to be verified based on the test data package; The verification parameter set includes: Mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE); Generate model accuracy evaluation values for the model to be validated based on the validation parameter set; Preset precision threshold; If the model accuracy evaluation value is greater than the preset model accuracy evaluation value threshold, the model to be verified is set as the power imbalance funding prediction model. If the model accuracy evaluation value is less than the preset model accuracy evaluation value threshold, an iterative training instruction is generated.
[0009] In some embodiments of this application, generating the verification parameter set of the model to be verified includes: Obtain the predicted values of each sample in the test dataset generated by the model to be validated; Generate the mean absolute error (MAE): Generate mean squared error (MSE): Generate the root mean square error (RMSE): Where yi is the true value of the i-th sample, f(xi) is the predicted value of the model, and m is the total number of samples.
[0010] In some embodiments of this application, the preset interpretation model includes: The SHAP value algorithm is used to parse samples in a structured dataset; The shapely values corresponding to each feature in each sample are obtained through parsing.
[0011] Where: n is the number of features, and k represents the k-th sample in the structured dataset. The i-th feature in sample k; It is the model's prediction baseline value for the sample, representing the model's expectation of the recognition result for any sample, as well as the label value of all samples; S represents the SHAP value of the model features, used to quantify the impact of feature x on the model output; S represents the value excluding features. Feature subset, This represents the expected output of the model when only a subset of features is considered. This represents the marginal gain resulting from adding new features to a subset; An explanatory model is constructed based on all shapely values.
[0012] In some embodiments of this application, the step of constructing an interpretation model based on all shapely values includes: in: The i-th feature represents the global importance of the i-th feature; K represents the total number of samples in the structured dataset. .
[0013] In some embodiments of this application, a key driver identification system for electricity imbalance funding is provided, comprising: The data processing unit is used to acquire multidimensional historical data and generate a structured dataset based on the preprocessing results of the multidimensional historical data. The training unit is used to set the model training strategy based on the preset basic prediction model and structured dataset, and to establish a power imbalance funding prediction model based on the model training strategy. The power imbalance funding prediction model outputs a prediction dataset based on real-time collected data. The analysis unit is used to generate recognition results based on a preset interpretation model and initial prediction parameters.
[0014] In some embodiments of this application, the training unit includes: The first training module is used to set the time series dataset for the first T time steps as (y t-1 y t-2 , ..., y t-N The current data is y. t The prediction result is ; Building a basic prediction model: Where n is time, K is the number of tree models to obtain the best fit target, F is the set of regression trees of the model, f(y) and w(y) are the leaf weights of each tree, q is the number of independent numbers in the model, and Q is the number of nodes in the leaves.
[0015] In some embodiments of this application, the training unit further includes: The second training module is used to set up training and test data packages based on structured data; The basic prediction model is trained using training data packages; Output the model to be verified based on the training results, and generate the verification parameter set of the model to be verified based on the test data package; The verification parameter set includes: Mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE); Generate model accuracy evaluation values for the model to be validated based on the validation parameter set; Preset precision threshold; If the model accuracy evaluation value is greater than the preset model accuracy evaluation value threshold, the model to be verified is set as the power imbalance funding prediction model. If the model accuracy evaluation value is less than the preset model accuracy evaluation value threshold, an iterative training instruction is generated. The set of validation parameters for generating the model to be validated includes: Obtain the predicted values of each sample in the test dataset generated by the model to be validated; Generate the mean absolute error (MAE): Generate mean squared error (MSE): Generate the root mean square error (RMSE): Where yi is the true value of the i-th sample, f(xi) is the predicted value of the model, and m is the total number of samples.
[0016] In some embodiments of this application, the analysis unit is further configured to: The SHAP value algorithm is used to parse samples in a structured dataset; The shapely values corresponding to each feature in each sample are obtained through parsing.
[0017] Where: n is the number of features, and k represents the k-th sample in the structured dataset. The i-th feature in sample k; It is the model's prediction baseline value for the sample, representing the model's expected prediction result for any sample, as well as the label value of all samples; S represents the SHAP value of the model features, used to quantify the impact of feature x on the model output; S represents the value excluding features. Feature subset, This represents the expected output of the model when only a subset of features is considered. This represents the marginal gain resulting from adding new features to a subset; An explanatory model is constructed based on all shapely values.
[0018] Compared with existing technologies, the key driving factors identification method and system for power imbalance funding proposed in this application have the following advantages: By integrating the XGBoost machine learning algorithm with SHAP interpretability analysis, a structured dataset covering multi-source power data and a key driver identification system were constructed. This effectively improved the accuracy of predicting the scale of power imbalance funds and enabled attribution analysis of various influencing factors in the formation of imbalance funds. This provides data-driven decision support for accurate early warning and proactive risk management of power imbalance funds. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating a method for identifying key driving factors of power imbalance funding in a preferred embodiment of this application. Detailed Implementation
[0020] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.
[0021] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.
[0022] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0023] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0024] like Figure 1 As shown in the preferred embodiment of this application, a method for identifying key driving factors of power imbalance funding includes: Acquire multidimensional historical data and generate a structured dataset based on the preprocessing results of the multidimensional historical data; Based on the preset basic prediction model and structured dataset, a model training strategy is set, and a power imbalance funding prediction model is established based on the model training strategy. The power imbalance funding prediction model outputs a prediction dataset based on real-time collected data, and generates recognition results based on a preset interpretation model and the prediction dataset.
[0025] Specifically, the multidimensional data includes five influencing factors of power imbalance funding: load dimension, planning dimension, price dimension, operation dimension, and assessment dimension. These five influencing factors are used as independent variables to predict power imbalance funding.
[0026] Specifically, the multidimensional historical data includes: for the load dimension, the data corresponds to the system load forecast time series data; for the planning dimension, the data corresponds to the time series data of priority power generation from new energy sources and priority power purchase by users; for the price dimension, the data corresponds to the time series data of spot electricity prices, medium- and long-term contract prices, and price difference; for the operation dimension, the data corresponds to the time series data of predicted output from new energy sources, the time series data of variable costs of traditional energy units, and the time series data of adjustable output of traditional energy units; and for the assessment dimension, the data corresponds to the time series data of deviation assessment costs and ancillary service and allocated costs.
[0027] Specifically, the preprocessing of the collected multidimensional historical data includes the following steps: Data cleaning: using 3 The criteria involve cleaning key time-series data and calculating the mean of the data series. and standard deviation Eliminate those that meet the requirements Anomalies; Missing value imputation: Missing data is imputed using Lagrange interpolation to ensure temporal continuity; Time granularity alignment and standardization: All dimensional data are unified to the hour for time granularity alignment, and the Min-Max standardization method is used to eliminate the influence of units. Feature selection: Correlation and collinearity tests are performed on the preprocessed data variables to select variables with a correlation greater than 0.8 and a variance inflation factor greater than 10. These variables are identified as factors strongly associated with imbalance funding prediction. The selected variables are used to form a feature vector, and the imbalance funding within the target region during the corresponding time period is used as the label to construct a structured dataset.
[0028] In a preferred embodiment of this application, a preset basic prediction model is included, comprising: Let the time series dataset for the first T time steps be (y t-1 y t-2 , ..., y t-N The current data is y. t The recognition result is ; Building a basic prediction model: Where n is time, K is the number of tree models to obtain the best fit target, F is the set of regression trees of the model, f(y) and w(y) are the leaf weights of each tree, q is the number of independent numbers in the model, and Q is the number of nodes in the leaves.
[0029] Specifically, the XGBoost algorithm is used for time series forecasting to build the corresponding basic forecasting model.
[0030] Specifically, the XGBoost model can model nonlinear relationships by combining multiple decision trees. Therefore, the electricity imbalance funding prediction method based on the XGBoost model is more flexible than the commonly used linear regression model. Furthermore, the XGBoost model has good robustness, can handle various types of data, and has good fault tolerance for outliers and noise, thus achieving higher prediction accuracy. The XGBoost model also has efficient parallel computing capabilities, can handle large-scale datasets and high-dimensional features, and has fast training and prediction speeds. In addition, the XGBoost model has an adaptive learning strategy that can automatically adjust the model parameters according to the data distribution, making the model more adaptable to the data characteristics.
[0031] Specifically, establishing a funding forecasting model for power imbalance includes: Training and test data packages are defined based on structured data; The basic prediction model is trained using training data packages; Output the model to be verified based on the training results, and generate the verification parameter set of the model to be verified based on the test data package; The verification parameter set includes: Mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE); Generate model accuracy evaluation values for the model to be validated based on the validation parameter set; Preset precision threshold; If the model accuracy evaluation value is greater than the preset model accuracy evaluation value threshold, the model to be verified is set as the power imbalance funding prediction model. If the model accuracy evaluation value is less than the preset model accuracy evaluation value threshold, an iterative training instruction is generated.
[0032] Specifically, the obtained structured dataset is divided into training and testing data sets in an 8:2 ratio. The training data set is used to train the model, and the testing data set is used to optimize the model's hyperparameters. The basic prediction model is trained and tested using the structured dataset, and the model to be validated and the validation parameter set are output after each iteration.
[0033] Specifically, the accuracy thresholds for mean absolute error, mean square error, and root mean square error are set according to actual needs. When the mean absolute error, mean square error, and root mean square error are all less than their respective set accuracy thresholds, the model accuracy evaluation value is set to 1; otherwise, the model accuracy evaluation value is set to 0.
[0034] Specifically, the threshold value for model accuracy evaluation is in the range of (0,1), and is preferably 0.5 in this application.
[0035] Specifically, when the mean absolute error, mean square error, and root mean square error are all less than their respective set accuracy thresholds, the iteration stops, and the model to be verified is the optimal power imbalance funding prediction model, which is then output as the power imbalance funding prediction model. Otherwise, the model to be verified continues to be optimized and iterated according to the training instructions.
[0036] Specifically, after training the power imbalance funding prediction model, real-time data is collected and converted into feature variables, which are then input into the power imbalance funding prediction model to output the predicted power imbalance funding value. A prediction dataset is then constructed based on all output parameters.
[0037] Specifically, the validation parameter set for the model to be validated is generated, including: Obtain the predicted values of each sample in the test dataset generated by the model to be validated; Generate the mean absolute error (MAE): Generate mean squared error (MSE): Generate the root mean square error (RMSE): Where yi is the true value of the i-th sample, f(xi) is the predicted value of the model, and m is the total number of samples.
[0038] Specifically, the mean absolute error is calculated by summing the absolute differences between the predicted and actual values for each sample and then taking the average; the mean square error is calculated by summing the squares of the differences between the predicted and actual values for each sample and then taking the average; and the root mean square error is calculated by taking the square root of the mean square error.
[0039] In a preferred embodiment of this application, the preset interpretation model includes: The SHAP value algorithm is used to parse samples in a structured dataset; The shapely values corresponding to each feature in each sample are obtained through parsing.
[0040] Where: n is the number of features, and k represents the k-th sample in the structured dataset. The i-th feature in sample k; It is the model's prediction baseline value for the sample, representing the model's expectation of the recognition result for any sample, as well as the label value of all samples; S represents the SHAP value of the model features, used to quantify the impact of feature x on the model output; S represents the value excluding features. Feature subset, This represents the expected output of the model when only a subset of features is considered. This represents the marginal gain resulting from adding new features to a subset; An explanatory model is constructed based on all shapely values.
[0041] Specifically, an explanatory model is constructed based on all shapely values, including: in: The i-th feature represents the global importance of the i-th feature; K represents the total number of samples in the structured dataset. .
[0042] Specifically, for By sorting the values from largest to smallest, we can identify the most important influencing factors globally. Specifically, since each feature in each sample corresponds to a shapely value, the local contribution of each feature to the predicted value of each sample can be explained by interpreting the model and the shapely value corresponding to each feature in each sample.
[0043] It is understood that, in the above embodiments, by integrating the XGBoost machine learning algorithm and SHAP interpretability analysis, a structured dataset covering multi-source power data and a key driving factor identification system are constructed, which effectively improves the prediction accuracy of the scale of power imbalance funds, realizes the attribution analysis of various influencing factors in the formation of imbalance funds, and provides data-driven decision support for accurate early warning and proactive risk management of power imbalance funds.
[0044] In another preferred embodiment of the method for identifying key driving factors of power imbalance funding based on any of the above preferred embodiments, this preferred embodiment provides a system for identifying key driving factors of power imbalance funding, comprising: The data processing unit is used to acquire multidimensional historical data and generate a structured dataset based on the preprocessing results of the multidimensional historical data. The training unit is used to set the model training strategy based on the preset basic prediction model and structured dataset, and to establish a power imbalance funding prediction model based on the model training strategy. The power imbalance funding prediction model outputs a prediction dataset based on real-time collected data. The analysis unit is used to generate recognition results based on a preset interpretation model and initial prediction parameters.
[0045] In a preferred embodiment of this application, the training unit includes: The first training module is used to set the time series dataset for the first T time steps as (y t-1 y t-2 , ..., y t-N The current data is y. t The prediction result is ; Building a basic prediction model: Where n is time, K is the number of tree models to obtain the best fit target, F is the set of regression trees of the model, f(y) and w(y) are the leaf weights of each tree, q is the number of independent numbers in the model, and Q is the number of nodes in the leaves.
[0046] Specifically, the training unit also includes: The second training module is used to set up training and test data packages based on structured data; The basic prediction model is trained using training data packages; Output the model to be verified based on the training results, and generate the verification parameter set of the model to be verified based on the test data package; The verification parameter set includes: Mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE); Generate model accuracy evaluation values for the model to be validated based on the validation parameter set; Preset precision threshold; If the model accuracy evaluation value is greater than the preset model accuracy evaluation value threshold, the model to be verified is set as the power imbalance funding prediction model. If the model accuracy evaluation value is less than the preset model accuracy evaluation value threshold, an iterative training instruction is generated. The set of validation parameters for generating the model to be validated includes: Obtain the predicted values of each sample in the test dataset generated by the model to be validated; Generate the mean absolute error (MAE): Generate mean squared error (MSE): Generate the root mean square error (RMSE): Where yi is the true value of the i-th sample, f(xi) is the predicted value of the model, and m is the total number of samples.
[0047] In a preferred embodiment of this application, the analysis unit is further configured to: The SHAP value algorithm is used to parse samples in a structured dataset; The shapely values corresponding to each feature in each sample are obtained through parsing.
[0048] Where: n is the number of features, and k represents the k-th sample in the structured dataset. The i-th feature in sample k; It is the model's prediction baseline value for the sample, representing the model's expected prediction result for any sample, as well as the label value of all samples; S represents the SHAP value of the model features, used to quantify the impact of feature x on the model output; S represents the value excluding features. Feature subset, This represents the expected output of the model when only a subset of features is considered. This represents the marginal gain resulting from adding new features to a subset; An explanatory model is constructed based on all shapely values.
[0049] Based on the first concept of this application, by integrating the XGBoost machine learning algorithm and SHAP interpretability analysis, a structured dataset covering multi-source power data and a key driver identification system are constructed. This effectively improves the accuracy of predicting the scale of power imbalance funds, realizes the attribution analysis of various influencing factors in the formation of imbalance funds, and provides data-driven decision support for accurate early warning and proactive risk management of power imbalance funds.
[0050] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application.
Claims
1. A method for identifying key driving factors of power imbalance funding, characterized in that, include: Acquire multidimensional historical data and generate a structured dataset based on the preprocessing results of the multidimensional historical data; Based on the preset basic prediction model and structured dataset, a model training strategy is set, and a power imbalance funding prediction model is established based on the model training strategy. The power imbalance funding prediction model outputs a prediction dataset based on real-time collected data, and generates recognition results based on a preset interpretation model and the prediction dataset.
2. The method for identifying key driving factors of power imbalance funding as described in claim 1, characterized in that, The preset basic prediction model includes: Let the time series dataset for the first T time steps be (y t-1 y t-2 , ..., y t-N The current data is y. t The recognition result is ; Building a basic prediction model: Where n is time, K is the number of tree models to obtain the best fit target, F is the set of regression trees of the model, f(y) and w(y) are the leaf weights of each tree, q is the number of independent numbers in the model, and Q is the number of nodes in the leaves.
3. The method for identifying key driving factors of power imbalance funding as described in claim 2, characterized in that, The establishment of the power imbalance funding forecasting model includes: Training and test data packages are defined based on structured data; The basic prediction model is trained using training data packages; Output the model to be verified based on the training results, and generate the verification parameter set of the model to be verified based on the test data package; The verification parameter set includes: Mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE); Generate model accuracy evaluation values for the model to be validated based on the validation parameter set; Preset precision threshold; If the model accuracy evaluation value is greater than the preset model accuracy evaluation value threshold, the model to be verified is set as the power imbalance funding prediction model. If the model accuracy evaluation value is less than the preset model accuracy evaluation value threshold, an iterative training instruction is generated.
4. The method for identifying key driving factors of power imbalance funding as described in claim 3, characterized in that, The set of validation parameters for generating the model to be validated includes: Obtain the predicted values of each sample in the test dataset generated by the model to be validated; Generate the mean absolute error (MAE): Generate mean squared error (MSE): Root mean square error (RMSE) is generated: Where yi is the true value of the i-th sample, f(xi) is the predicted value of the model, and m is the total number of samples.
5. The method for identifying key driving factors of power imbalance funding as described in claim 3, characterized in that, The preset interpretation model includes: The SHAP value algorithm is used to parse samples in a structured dataset; The shapely values corresponding to each feature in each sample are obtained through parsing. Where: n is the number of features, and k represents the k-th sample in the structured dataset. The i-th feature in sample k; It is the model's prediction baseline value for the sample, representing the model's expectation of the recognition result for any sample, as well as the label value of all samples; S represents the SHAP value of the model features, used to quantify the impact of feature x on the model output; S represents the value excluding features. Feature subset, This represents the expected output of the model when only a subset of features is considered. This represents the marginal gain resulting from adding new features to a subset; An explanatory model is constructed based on all shapely values.
6. The method for identifying key driving factors of power imbalance funding as described in claim 5, characterized in that, The explanatory model is constructed based on all shapely values. include: in: The i-th feature represents the global importance of the i-th feature; K represents the total number of samples in the structured dataset. .
7. A system for identifying key driving factors of power imbalance funding, employing the method for identifying key driving factors of power imbalance funding as described in any one of claims 1-6, characterized in that, include: The data processing unit is used to acquire multidimensional historical data and generate a structured dataset based on the preprocessing results of the multidimensional historical data. The training unit is used to set the model training strategy based on the preset basic prediction model and structured dataset, and to establish a power imbalance funding prediction model based on the model training strategy. The power imbalance funding prediction model outputs a prediction dataset based on real-time collected data. The analysis unit is used to generate recognition results based on a preset interpretation model and initial prediction parameters.
8. The key driver factor identification system for power imbalance funding as described in claim 7, characterized in that, The training unit includes: The first training module is used to set the time series dataset for the first T time steps as (y t-1 y t-2 , ..., y t-N The current data is y. t The prediction result is ; Building a basic prediction model: Where n is time, K is the number of tree models to obtain the best fit target, F is the set of regression trees of the model, f(y) and w(y) are the leaf weights of each tree, q is the number of independent numbers in the model, and Q is the number of nodes in the leaves.
9. The key driver factor identification system for power imbalance funding as described in claim 8, characterized in that, The training unit also includes: The second training module is used to set up training and test data packages based on structured data; The basic prediction model is trained using training data packages; Output the model to be verified based on the training results, and generate the verification parameter set of the model to be verified based on the test data package; The verification parameter set includes: Mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE); Generate model accuracy evaluation values for the model to be validated based on the validation parameter set; Preset precision threshold; If the model accuracy evaluation value is greater than the preset model accuracy evaluation value threshold, the model to be verified is set as the power imbalance funding prediction model. If the model accuracy evaluation value is less than the preset model accuracy evaluation value threshold, an iterative training instruction is generated. The set of validation parameters for generating the model to be validated includes: Obtain the predicted values of each sample in the test dataset generated by the model to be validated; Generate the mean absolute error (MAE): Generate mean squared error (MSE): Root mean square error (RMSE) is generated: Where yi is the true value of the i-th sample, f(xi) is the predicted value of the model, and m is the total number of samples.
10. The key driver factor identification system for power imbalance funding as described in claim 9, characterized in that, The analysis unit is also used for: The SHAP value algorithm is used to parse samples in a structured dataset; The shapely values corresponding to each feature in each sample are obtained through parsing. Where: n is the number of features, and k represents the k-th sample in the structured dataset. The i-th feature in sample k; It is the model's prediction baseline value for the sample, representing the model's expected prediction result for any sample, as well as the label value of all samples; S represents the SHAP value of the model features, used to quantify the impact of feature x on the model output; S represents the value excluding features. Feature subset, This represents the expected output of the model when only a subset of features is considered. This represents the marginal gain resulting from adding new features to a subset; An explanatory model is constructed based on all shapely values.