A power operation data processing method and system based on heterogeneous data fusion

Abstract

The application discloses a power operation data processing method and system based on heterogeneous data fusion, wherein the method comprises the following steps: acquiring charging behavior and power grid power supply historical data of a charging station; sequentially cleaning and filling the two kinds of data; extracting features of the charging behavior data, and analyzing the features according to a correlation analysis algorithm to obtain a target feature set; screening the charging behavior data according to the target feature set to obtain target charging behavior data; correcting the power grid power supply data according to a power grid terminal execution standard to obtain target power grid power supply data; fusing the two kinds of data based on a decision tree method; training a charging load prediction model by using the fused data; and predicting based on the model to obtain a charging load prediction result. The method provided by the application provides reliable technical support for power grid dispatching and charging station operation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric vehicle charging, in particular to a power operation data processing method and system based on heterogeneous data fusion. BACKGROUND

[0002] With the development of electric vehicle charging technology, multi-source data problems are particularly prominent. The charging process involves vehicle battery status, charging station facility information, power grid power supply status, and user charging behavior, etc. Various data types, these data sources are extensive and have various formats, making it difficult to obtain accurate charging load, and thus bringing challenges to the formulation of charging strategy, the improvement of charging efficiency and the optimization of charging facilities.

[0003] In the prior art, a simple data aggregation method is used for multi-source heterogeneous data to realize the prediction of charging load. This method ignores the correlation between data, and the data fusion effect is poor, and the data accuracy is low. The precision of the predicted charging load is not high.

[0004] Therefore, how to improve the prediction accuracy of the charging load, and then formulate a reasonable charging strategy, has become a technical problem to be solved by the technical personnel in the field. SUMMARY

[0005] The present application provides a power operation data processing method and system based on heterogeneous data fusion to improve the prediction accuracy of the charging load, and to realize the effect of accurately predicting the charging load and formulating a reasonable charging strategy.

[0006] In order to solve the above technical problems, the embodiment of the present application provides a power operation data processing method based on heterogeneous data fusion, comprising:

[0007] The charging behavior history data and the power grid power supply behavior history data of the target charging station obtained are sequentially subjected to data cleaning and data filling, respectively, to obtain first charging behavior data and first power grid power supply data;

[0008] According to the correlation analysis algorithm, the user features, time series features and charging mode features extracted from the first charging behavior data are analyzed to obtain a target feature set matched with the first charging behavior data;

[0009] The first charging behavior data is filtered based on the target feature set to obtain target charging behavior data; and the first power grid power supply data is corrected according to a pre-determined power grid terminal execution standard to obtain target power grid power supply data;

[0010] The target charging behavior data and the target power grid power supply data are fused based on a decision tree method to obtain a target sample set;

[0011] training the constructed charging load prediction model based on the target sample set to obtain a target charging load prediction model;

[0012] In the actual charging load prediction process, the target charging load prediction model is used for prediction to obtain a charging load prediction result.

[0013] As one of the preferred solutions, the user features, time series features and charging mode features are analyzed in sequence according to the correlation analysis algorithm to obtain a target feature set matched with the first charging behavior data, which includes:

[0014] The correlation coefficients between the user features, time series features and charging mode features and the first charging behavior data are calculated based on the Pearson correlation analysis algorithm;

[0015] Each correlation coefficient is filtered based on a pre-set correlation threshold, and the target feature set is constructed based on the filtering result.

[0016] As one of the preferred solutions, the grid terminal execution standard includes a power factor standard;

[0017] The first grid power supply data is corrected according to the pre-determined grid terminal execution standard to obtain target grid power supply data, which includes:

[0018] The active power output and the reactive power output in the extracted grid power supply behavior history data are compared and analyzed, and a target power factor is determined according to the analysis result;

[0019] The power factor standard is constructed based on the target power factor;

[0020] The current power factor in the first grid power supply data is extracted, and the current power factor is corrected based on the power factor standard to obtain the target grid power supply data after correction.

[0021] As one of the preferred solutions, the target charging load prediction model is obtained by training the constructed charging load prediction model based on the target sample set, which includes:

[0022] An initial charging load prediction model based on a long short-term memory network is constructed;

[0023] The target sample set is expanded according to a resampling method to obtain an expanded data set;

[0024] The initial charging load prediction model is trained based on the expanded data set and a Bayesian optimization algorithm to obtain a target charging load prediction model.

[0025] As one of the preferred solutions, after the target charging load prediction model is obtained, the method further comprises:

[0026] The target charging load prediction model is verified based on a pre-selected verification data set, and the target charging load prediction model is optimized based on the verification result.

[0027] Another embodiment of the present application provides a power operation data processing system based on heterogeneous data fusion, characterized in that it comprises:

[0028] The acquisition module is configured to sequentially perform data cleaning and data filling on the acquired charging behavior history data and power grid power supply behavior history data of the target charging station, respectively, to obtain first charging behavior data and first power grid power supply data.

[0029] The analysis module is configured to analyze user features, time series features and charging mode features extracted from the first charging behavior data according to a correlation analysis algorithm, to obtain a target feature set matched with the first charging behavior data.

[0030] The extraction module is configured to filter the first charging behavior data based on the target feature set, to obtain target charging behavior data; and correct the first power grid power supply data according to a pre-determined power grid terminal execution standard, to obtain target power grid power supply data.

[0031] The fusion module is configured to fuse the target charging behavior data and the target power grid power supply data based on a decision tree method, to obtain a target sample set.

[0032] The training module is configured to train a constructed charging load prediction model based on the target sample set, to obtain a target charging load prediction model.

[0033] The prediction module is configured to perform prediction based on the target charging load prediction model in an actual charging load prediction process, to obtain a charging load prediction result.

[0034] As one of the preferred solutions, the analysis module is specifically configured to:

[0035] Calculate correlation coefficients between the user features, the time series features and the charging mode features and the first charging behavior data based on a Pearson correlation analysis algorithm.

[0036] Filter each of the correlation coefficients based on a pre-set correlation threshold, and construct the target feature set based on the filtering result.

[0037] As one of the preferred solutions, the power grid terminal execution standard comprises a power factor standard.

[0038] The extraction module is specifically configured to:

[0039] The active power output and the reactive power output in the extracted power grid power supply behavior history data are compared and analyzed, and a target power factor is determined according to an analysis result;

[0040] The target power factor is used to construct the power factor standard;

[0041] The current power factor in the first power grid power supply data is extracted, and the current power factor is corrected based on the power factor standard to obtain the target power grid power supply data after correction.

[0042] As one of the preferred solutions, the prediction module is specifically configured to:

[0043] An initial charging load prediction model based on a long short-term memory network is constructed;

[0044] The target sample set is expanded according to a resampling method to obtain an expanded data set;

[0045] The initial charging load prediction model is trained based on the expanded data set and a Bayesian optimization algorithm to obtain a target charging load prediction model.

[0046] As one of the preferred solutions, the prediction module is further configured to:

[0047] The target charging load prediction model is verified based on a preselected verification data set, and the target charging load prediction model is optimized based on a verification result.

[0048] Compared with the prior art, the beneficial effects of the embodiments of the application are at least one of the following:

[0049] (1) The application can filter out key features that have an important influence on charging load prediction through correlation analysis, thereby improving the prediction accuracy and efficiency of the model.

[0050] (2) The application can fully utilize the complementarity between two kinds of heterogeneous data through the fusion strategy of the decision tree method, improve the overall quality and prediction value of the data, and at the same time, the intuitiveness and interpretability of the decision tree method also make the fusion process clearer and easier to understand.

[0051] (3) The model obtained by training can accurately reflect the change rule of the charging load, providing reliable prediction support for the operation and dispatching of the power system, and at the same time, the prediction result of the model can also provide an important reference for the operation optimization of the charging station and the improvement of user experience.

[0052] (4) The charging load prediction method based on heterogeneous data fusion proposed in the application realizes accurate prediction and efficient management of the charging load. This method not only improves the accuracy and stability of the prediction, but also provides strong support for the optimized operation of the power system and the operation optimization of the charging station. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is a flowchart of the power operation data processing method based on heterogeneous data fusion in an embodiment of the application;

[0054] Figure 2 is a module schematic diagram of the power operation data processing system based on heterogeneous data fusion in an embodiment of the application. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. The purpose of providing these embodiments is to make the disclosure of the application more thorough and comprehensive. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the application.

[0056] In the description of the application, the terms "first", "second", "third", etc. are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second", "third", etc. can explicitly or implicitly include one or more of the features. In the description of the application, unless otherwise specified, the meaning of "multiple" is two or more.

[0057] In the description of the application, it should be noted that, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the communication between two elements. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used in this paper are only for the purpose of description, and cannot be understood as indicating or implying that the devices or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the application. The term "and / or" used in this paper includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meaning of the above terms in the application can be understood according to the specific circumstances.

[0058] In the description of the present application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present application have the same meanings as commonly understood by one of ordinary skill in the art to which this technology belongs. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The above terms can be understood in the specific context by one of ordinary skill in the art.

[0059] The embodiment of the present application provides a power operation data processing method based on heterogeneous data fusion, and specifically, please refer to Figure 1 , Figure 1 The flowchart of the power operation data processing method based on heterogeneous data fusion in one embodiment of the present application is shown, which includes steps S1-S6:

[0060] S1: The obtained charging behavior history data and power grid power supply behavior history data of the target charging station are sequentially subjected to data cleaning and data filling, respectively, to obtain first charging behavior data and first power grid power supply data;

[0061] In one embodiment of the present application, the first charging behavior data mainly reflects the charging activities of electric vehicle users at the charging station, specifically including:

[0062] User information: such as user ID, user type (such as private user, operation user, etc.), user vehicle information (such as vehicle type, battery capacity, etc.).

[0063] Charging time information: including start charging time, end charging time, charging duration, etc. These data are helpful to understand the charging habits, charging period distribution, etc. of users.

[0064] Charging amount information: such as the amount of electricity for each charging, the remaining amount of electricity before charging, the amount of electricity after charging, etc. These data reflect the charging demand and charging efficiency of users.

[0065] Charging station information: such as charging station ID, charging station location, charging pile type (such as alternating current charging pile, direct current charging pile), etc.

[0066] Charging cost information: such as the cost of each charging, charging unit price, etc.

[0067] The first power grid power supply data mainly reflects the power supply situation and stability of the power grid, specifically including:

[0068] Power grid load information: such as real-time load of power grid, load peak, load fluctuation, etc. These data are helpful to understand the power supply capacity and stability of the power grid.

[0069] Supply Quality Information: such as voltage stability, frequency stability, power factor, etc. These data reflect the power supply quality of the power grid.

[0070] Grid Failure Information: such as grid failure type, failure time, failure impact range, etc. These data are of great significance for evaluating the reliability and safety of the power grid.

[0071] Electricity Price Information: such as electricity price in different time periods, peak-valley electricity price difference, etc. These data reflect the price mechanism of the electricity market and the electricity cost of users.

[0072] Among them, the data cleaning is preferably abnormal value processing, duplicate data deletion, data consistency processing and data normalization, etc., and the data filling is preferably mean filling, median filling, interpolation method, multiple imputation, K-nearest neighbor filling and regression prediction filling, etc., which are not limited in the embodiments of the present application.

[0073] S2: analyzing the user features, time series features and charging mode features extracted from the first charging behavior data according to a correlation analysis algorithm to obtain a target feature set matched with the first charging behavior data;

[0074] Preferably, in an embodiment of the present application, the user features, time series features and charging mode features are analyzed in sequence according to the correlation analysis algorithm to obtain the target feature set matched with the first charging behavior data, including:

[0075] calculating the correlation coefficients between the user features, time series features and charging mode features and the first charging behavior data based on the Pearson correlation analysis algorithm;

[0076] screening each correlation coefficient based on a pre-set correlation threshold to construct the target feature set with the screening result.

[0077] Among them, the Pearson correlation analysis algorithm is a statistical method for measuring the degree of linear correlation between two continuous variables. In an embodiment of the present application, calculating the correlation coefficient based on the Pearson correlation analysis algorithm includes:

[0078] extracting the user feature data set, such as user type, vehicle type, charging frequency, etc. For each user feature, the Pearson correlation analysis algorithm is used to calculate the correlation coefficient between it and the first charging behavior data (such as charging amount, charging times, etc.).

[0079] extracting the time series feature data set, such as daily charging amount trend, weekly or monthly charging amount fluctuation, etc. For each time series feature, the Pearson correlation analysis algorithm is also used to calculate the correlation coefficient between it and the first charging behavior data.

[0080] Extract the charging mode feature dataset, such as the time of charging start and end, charging speed, charging power, etc. For each charging mode feature, use the Pearson correlation analysis algorithm to calculate the correlation coefficient between it and the first charging behavior data.

[0081] Further, according to the business needs and data characteristics, a reasonable correlation threshold is set. Generally, this threshold can be determined according to experience or business needs, for example, 0.3, 0.5 or 0.7, etc. are selected as the threshold. According to the set correlation threshold, the user features, time series features and charging mode features with correlation coefficients greater than or equal to the threshold are screened out from the first charging behavior data. These screened features constitute the target feature set.

[0082] S3: screening the first charging behavior data based on the target feature set to obtain the target charging behavior data; correcting the first power grid supply data according to the pre-determined power grid terminal execution standard to obtain the target power grid supply data;

[0083] The target feature set obtained from step S2 accurately reflects the key elements of charging behavior, such as user type, charging period, charging amount, etc. In an embodiment of the present application, the target feature set is used to screen the first charging behavior data. Specifically, records matching the target feature set are selected from the original data set. For example, if the target feature set emphasizes the charging behavior of a specific user type (such as private users) in a specific period (such as evening), the charging records that meet these conditions are screened out. The screened data is checked to ensure its accuracy, completeness and consistency. This includes checking whether the data format, data type, data range, etc. meet the expectations and correcting any possible errors, which are not specifically limited in the embodiment of the present application.

[0084] Preferably, in an embodiment of the present application, the power factor standard includes a power factor standard;

[0085] According to the pre-determined power grid terminal execution standard, the first power grid supply data is corrected to obtain the target power grid supply data, including:

[0086] Comparative analysis of the active power output and the reactive power output in the extracted power grid supply behavior history data, and determining the target power factor according to the analysis result;

[0087] Constructing a power factor standard based on the target power factor;

[0088] Extracting the current power factor in the first power grid supply data, and correcting the current power factor based on the power factor standard to obtain the corrected target power grid supply data.

[0089] It should be noted that in the power system, the grid terminal performs the standard for ensuring the stable operation and efficient power supply of the power grid. Among them, the power factor standard is a key implementation standard, which reflects the proportional relationship between active power and apparent power in the power grid, and is an important indicator to measure the degree of effective use of electricity.

[0090] In this step, the active power and reactive power are compared and analyzed to understand the power factor changes of the power grid under different time periods and different load conditions. According to the comparison and analysis results, combined with the actual situation of power grid operation and business demand, a reasonable target power factor is determined. Generally, the target power factor should be higher than or equal to the value specified in the national standard to ensure the efficient operation of the power grid. For general industrial and commercial electricity, the power factor of the power grid with rated voltage below 1000V should be above 0.9 and not less than 0.8; for power grids with rated voltage exceeding 1000V, the power factor should be above 0.95 and not less than 0.9.

[0091] The current power factor data is extracted from the first power grid power supply data. This can be obtained in real time through the power grid monitoring system or data acquisition system. The extracted current power factor is compared with the constructed power factor standard. If the current power factor is lower than the specified value in the standard, it needs to be corrected.

[0092] The correction method may include adjusting the load distribution of the power grid, increasing reactive power compensation devices (such as capacitor banks, static reactive power compensators, etc.) to improve the power factor, which is not limited herein. During the correction process, the change of the power factor should be monitored in real time to ensure that the corrected power factor meets the standard requirements.

[0093] S4: Based on the decision tree method, the target charging behavior data and the target power grid power supply data are fused to obtain a target sample set;

[0094] Among them, the decision tree algorithm is a commonly used classification and regression method, which mainly classifies or predicts data through a series of questions (i.e. selection and division of features). Its model structure is similar to the shape of a tree, where each node of the tree represents a test on a feature, each branch represents a test result, and the leaf nodes of the tree represent a class or output value.

[0095] It can be understood that first, the most important feature is selected as the root node, which should be able to best divide the data set. For example, the application needs to obtain the influence of charging behavior on the power grid at different time periods, so time can be used as a root node feature.

[0096] The dataset is divided into different subsets according to the feature values of the root nodes. For each subset, a new feature is recursively selected as the division criterion for the child nodes until the stopping conditions are met (such as the number of samples within a node being less than a certain threshold, or the purity improvement after division is no longer significant). The leaf nodes represent the final decision results or categories. In the context of fusion data, the leaf nodes can represent a specific combination of charging behavior and grid power supply state.

[0097] In this step, starting from the root node, the decision tree is traversed according to the feature values of the target charging behavior data and the target grid power supply data. A sample is generated every time a leaf node is reached. This sample contains the values of all features from the root node to the leaf node. All generated samples are combined to form the target sample set. This set contains the fused charging behavior and grid power supply data.

[0098] S5: Train the constructed charging load prediction model based on the target sample set to obtain a target charging load prediction model;

[0099] Preferably, in an embodiment of the present application, training the constructed charging load prediction model based on the target sample set to obtain a target charging load prediction model comprises:

[0100] Constructing an initial charging load prediction model based on a long short-term memory network;

[0101] According to the resampling method, the target sample set is expanded to obtain an expanded data set;

[0102] Training the initial charging load prediction model based on the expanded data set and the Bayesian optimization algorithm to obtain the target charging load prediction model.

[0103] In this step, an LSTM network is constructed using a deep learning framework (such as TensorFlow, PyTorch, etc.). The input layer, LSTM layer (which can contain multiple stacked LSTM units), fully connected layer, and output layer of the network are set. Appropriate activation functions (such as ReLU, Sigmoid, etc.) and loss functions (such as MSE, MAE, etc.) are selected.

[0104] Among them, according to the distribution and characteristics of the data, select appropriate resampling methods, such as random oversampling, random undersampling, SMOTE (synthetic minority over-sampling technique), etc., which are not limited here. In an embodiment of the present application, the resampled data set is evaluated by statistical indicators (such as class distribution, data volume, etc.) and visualization methods (such as scatter plots, histograms, etc.) to determine whether it meets the requirements.

[0105] Specifically, the initial LSTM charging load prediction model is trained using the augmented dataset. Among them, the Bayesian optimization algorithm is selected as the hyperparameter optimization method. Bayesian optimization approximates the target function (the loss function of the model in this embodiment) by constructing a probability model, and uses this model to select the next combination of hyperparameters to be evaluated. The optimal combination of hyperparameters is selected from the Bayesian optimization process, and the LSTM model is retrained using the combination.

[0106] Preferably, in one embodiment of the present application, the method further comprises:

[0107] Based on the pre-selected validation dataset, the target charging load prediction model is verified, and the target charging load prediction model is optimized based on the verification result.

[0108] Specifically, the trained model is evaluated, including performance on the test set, stability and interpretability of the model, etc. If the requirements are met, it will be deployed and used as the target charging load prediction model.

[0109] S6: In the actual charging load prediction process, the target charging load prediction model is used for prediction to obtain the charging load prediction result.

[0110] In this step, data related to charging load prediction is collected from relevant data sources (e.g. charging station monitoring system, power grid dispatching system, etc.) in real time or periodically. The prepared input data is input into the target charging load prediction model for prediction calculation. The model will output the corresponding charging load prediction result according to the characteristics of the input data.

[0111] The prediction results are obtained from the model. These results include the charging load prediction values in the future period, the confidence interval of the prediction values, etc. In one embodiment of the present application, the prediction results can be visualized, such as drawing a prediction curve graph, so as to more intuitively understand the prediction results.

[0112] Specifically, the prediction results can be applied to the operation and management of charging stations, such as formulating charging plans, optimizing charging station resource allocation, etc.; or providing the prediction results to the power grid dispatching department to help them better understand the future charging load demand, so as to formulate more reasonable power grid dispatching plans. Or use the prediction results as the basis for policy making and planning, such as formulating electric vehicle charging infrastructure construction planning, optimizing charging policies, etc.

[0113] Another embodiment of the present application provides a heterogeneous data fusion power operation data processing system. Specifically, please refer to Figure 2 , Figure 2 The module schematic diagram of the heterogeneous data fusion power operation data processing system in one embodiment of the present application is shown, which comprises:

[0114] The acquisition module 11 is configured to sequentially perform data cleaning and data filling on the acquired charging behavior history data and power grid power supply behavior history data of the target charging station respectively, to obtain first charging behavior data and first power grid power supply data respectively;

[0115] The analysis module 12 is configured to analyze the user features, time sequence features and charging mode features extracted from the first charging behavior data according to a correlation analysis algorithm, to obtain a target feature set matched with the first charging behavior data;

[0116] The extraction module 13 is configured to filter the first charging behavior data based on the target feature set, to obtain target charging behavior data; and correct the first power grid power supply data according to a pre-determined power grid terminal execution standard, to obtain target power grid power supply data;

[0117] The fusion module 14 is configured to fuse the target charging behavior data and the target power grid power supply data based on a decision tree method, to obtain a target sample set;

[0118] The training module 15 is configured to train a constructed charging load prediction model based on the target sample set, to obtain a target charging load prediction model;

[0119] The prediction module 16 is configured to perform prediction based on the target charging load prediction model in an actual charging load prediction process, to obtain a charging load prediction result.

[0120] Preferably, in an embodiment of the present application, the analysis module is specifically configured to:

[0121] calculate correlation coefficients between the user features, the time sequence features and the charging mode features and the first charging behavior data based on a Pearson correlation analysis algorithm;

[0122] filter the correlation coefficients based on a pre-set correlation threshold, to construct the target feature set based on the filtering result.

[0123] Preferably, in an embodiment of the present application, the power grid terminal execution standard comprises a power factor standard;

[0124] The extraction module is specifically configured to:

[0125] perform comparison analysis on the active power output and the reactive power output in the extracted power grid power supply behavior history data, to determine a target power factor according to an analysis result;

[0126] construct the power factor standard based on the target power factor;

[0127] The current power factor in the first power grid power supply data is extracted, and the current power factor is corrected based on a power factor standard to obtain corrected target power grid power supply data.

[0128] Preferably, in an embodiment of the present application, the prediction module is specifically used for:

[0129] An initial charging load prediction model based on a long short-term memory network is constructed.

[0130] The target sample set is expanded according to a resampling method to obtain an expanded data set.

[0131] The initial charging load prediction model is trained based on the expanded data set and a Bayesian optimization algorithm to obtain a target charging load prediction model.

[0132] Preferably, in an embodiment of the present application, the prediction module is further used for:

[0133] The target charging load prediction model is verified based on a preselected verification data set, and the target charging load prediction model is optimized based on a verification result.

[0134] The heterogeneous data fusion power operation data processing method and system provided by the embodiments of the present application have at least one of the following beneficial effects:

[0135] (1) The present application can filter out key features that have an important influence on charging load prediction through correlation analysis, thereby improving the prediction accuracy and efficiency of the model.

[0136] (2) The present application can make full use of the complementarity between the two kinds of heterogeneous data through the fusion strategy of the decision tree method, improve the overall quality and prediction value of the data, and at the same time, the intuitiveness and interpretability of the decision tree method also make the fusion process clearer and easier to understand.

[0137] (3) The model obtained by training can accurately reflect the change rule of the charging load, providing reliable prediction support for the operation and dispatching of the power system, and at the same time, the prediction result of the model can also provide an important reference for the operation optimization of the charging site and the improvement of user experience.

[0138] (4) The charging load prediction method based on heterogeneous data fusion proposed in the present application realizes accurate prediction and efficient management of the charging load. This method not only improves the accuracy and stability of the prediction, but also provides strong support for the optimized operation of the power system and the operation optimization of the charging site.

[0139] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent of the present application. It should be noted that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A power operation data processing method based on heterogeneous data fusion, characterized in that, include: The historical charging behavior data and historical power supply behavior data of the target charging station are cleaned and populated in sequence to obtain the first charging behavior data and the first power supply data, respectively. The user features, time series features, and charging mode features extracted from the first charging behavior data are analyzed using a correlation analysis algorithm to obtain a target feature set that matches the first charging behavior data. Based on the target feature set, the first charging behavior data is filtered to obtain target charging behavior data; According to the predetermined power grid terminal execution standard, the first power grid power supply data is corrected to obtain the target power grid power supply data; The target charging behavior data and the target power grid supply data are fused using the decision tree method to obtain the target sample set; The charging load prediction model is trained based on the target sample set to obtain the target charging load prediction model; In the actual charging load prediction process, the target charging load prediction model is used to make predictions and obtain the charging load prediction results. The standards implemented by the power grid terminals include power factor standards; The step of correcting the first power grid power supply data according to a predetermined power grid terminal execution standard to obtain the target power grid power supply data includes: The active power output and reactive power output in the extracted historical data of power grid supply behavior are compared and analyzed, and the target power factor is determined based on the analysis results. The power factor standard is constructed based on the target power factor; Extract the current power factor from the first power grid supply data, and correct the current power factor based on the power factor standard to obtain the corrected target power grid supply data.

2. The power operation data processing method based on heterogeneous data fusion as described in claim 1, characterized in that, The user characteristics, time series characteristics, and charging mode characteristics are analyzed sequentially using a correlation analysis algorithm to obtain a target feature set that matches the first charging behavior data, including: Based on the Pearson correlation analysis algorithm, the correlation coefficients between the user characteristics, time series characteristics, and charging mode characteristics and the first charging behavior data were calculated respectively. The correlation coefficients are filtered based on a pre-set correlation threshold, and the target feature set is constructed from the filtering results.

3. The power operation data processing method based on heterogeneous data fusion as described in claim 1, characterized in that, The process of training the constructed charging load prediction model based on the target sample set to obtain the target charging load prediction model includes: Construct an initial charging load prediction model based on a long short-term memory network; The target sample set is expanded using a resampling method to obtain an expanded dataset; The initial charging load prediction model is trained based on the expanded dataset and the Bayesian optimization algorithm to obtain the target charging load prediction model.

4. The power operation data processing method based on heterogeneous data fusion as described in claim 3, characterized in that, After obtaining the target charging load prediction model, the method further includes: The target charging load prediction model is validated based on a pre-selected validation dataset, and then optimized based on the validation results.

5. A power operation data processing system based on heterogeneous data fusion, characterized in that, include: The acquisition module is used to perform data cleaning and data filling on the acquired historical data of charging behavior and historical data of power grid supply behavior of the target charging station in sequence, so as to obtain the first charging behavior data and the first power grid supply data respectively. The analysis module is used to analyze the user features, time series features and charging mode features extracted from the first charging behavior data according to the correlation analysis algorithm, and obtain a target feature set that matches the first charging behavior data. The extraction module is used to filter the first charging behavior data based on the target feature set to obtain target charging behavior data; According to the predetermined power grid terminal execution standard, the first power grid power supply data is corrected to obtain the target power grid power supply data; The fusion module is used to fuse the target charging behavior data and the target power grid power supply data based on the decision tree method to obtain a target sample set; The training module is used to train the constructed charging load prediction model based on the target sample set to obtain the target charging load prediction model. The prediction module is used to make predictions based on the target charging load prediction model during the actual charging load prediction process, and obtain the charging load prediction results. The standards implemented by the power grid terminals include power factor standards; The extraction module is specifically used for: The active power output and reactive power output in the extracted historical data of power grid supply behavior are compared and analyzed, and the target power factor is determined based on the analysis results. The power factor standard is constructed based on the target power factor; Extract the current power factor from the first power grid supply data, and correct the current power factor based on the power factor standard to obtain the corrected target power grid supply data.

6. The power operation data processing system based on heterogeneous data fusion as described in claim 5, characterized in that, The analysis module is specifically used for: Based on the Pearson correlation analysis algorithm, the correlation coefficients between the user characteristics, time series characteristics, and charging mode characteristics and the first charging behavior data were calculated respectively. The correlation coefficients are filtered based on a pre-set correlation threshold, and the target feature set is constructed from the filtering results.

7. The power operation data processing system based on heterogeneous data fusion as described in claim 5, characterized in that, The prediction module is specifically used for: Construct an initial charging load prediction model based on a long short-term memory network; The target sample set is expanded using a resampling method to obtain an expanded dataset; The initial charging load prediction model is trained based on the expanded dataset and the Bayesian optimization algorithm to obtain the target charging load prediction model.

8. The power operation data processing system based on heterogeneous data fusion as described in claim 7, characterized in that, The prediction module is also used for: The target charging load prediction model is validated based on a pre-selected validation dataset, and then optimized based on the validation results.

Images