Intelligent load forecasting method, system, electronic equipment, medium and chip for substation area

Abstract

The present disclosure relates to the technical field of load prediction, and specifically to a method, system, electronic device, medium and chip for intelligent load prediction in a substation. The method for intelligent load prediction in a substation includes: obtaining power load data in the substation; extracting training features from the power load data in the substation; wherein the training features include historical power load data and characteristic parameters that affect the power load; the characteristic parameters are configured with weight factors; using the historical power load data, characteristic parameters and weight factors as training input data of a recurrent neural network, training the recurrent neural network to obtain a substation load prediction model; and using the substation load prediction model to perform power load prediction. In the above technical solution, power load prediction is performed by training the substation load prediction model, thereby improving the accuracy of load prediction in complex environments.

Description

Technical Field

[0001] The present disclosure relates to the technical field of load forecasting, and in particular to a method, system, electronic device, medium and chip for intelligent load forecasting in a substation. Background Art

[0002] The rapid development of the power market has made the factors affecting power load changes more complex. These factors mainly include the adjustment of economic structure, changes in power consumption structure, changes in electricity prices, changes in weather, etc., which have a profound impact on the effectiveness and scope of application of traditional power load forecasting methods. Traditional forecasting methods are mainly implemented through mathematical and statistical models. The forecasting method is simple and fast. When the accuracy requirement is not high, its forecast results have a certain reference value. With the diversification of the power market, power load data is becoming increasingly complex. Traditional forecasting methods are difficult to adapt to the nonlinearity, randomness and chaos of power load data. Summary of the invention

[0003] In order to solve the problems in the related art, the embodiments of the present disclosure provide a method, system, electronic device, medium and chip for intelligent load prediction in a substation.

[0004] In a first aspect, an embodiment of the present disclosure provides a method for intelligent load forecasting in a substation area, comprising:

[0005] Obtain area power load data;

[0006] Extracting training features from the power load data of the substation area; wherein the training features include historical power load data and characteristic parameters affecting the power load; the characteristic parameters are configured with weight factors;

[0007] The power load historical data, characteristic parameters and weight factors are used as training input data of a recurrent neural network, and the recurrent neural network is trained to obtain a load prediction model for the substation;

[0008] The power load forecasting model is used to forecast the power load.

[0009] In one implementation of the present disclosure, the intelligent load forecasting method for the substation area further includes:

[0010] The step of configuring weight factors for the feature parameters.

[0011] In one implementation of the present disclosure, the step of configuring a weight factor for the characteristic parameter includes:

[0012] The characteristic parameters are taken as a characteristic set, the power load history data is taken as a category set, and the weight factors of the characteristic parameters are determined based on the attribute-weighted naive Bayes algorithm according to the characteristic set and the category set.

[0013] In an implementation of the present disclosure, the characteristic parameter is one or more of a meteorological parameter, an area parameter, and a date type.

[0014] In one implementation of the present disclosure, the recurrent neural network is an LSTM network; the LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

[0015] In one implementation of the present disclosure, the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and at the same time introduces the unit state at time t-1 into the input signal of the forgetting gate and the output gate at time t.

[0016] In one implementation of the present disclosure, extracting training features from the power load data of the substation includes:

[0017] Performing data preprocessing on the power load data of the substation area to obtain a sample set;

[0018] Training features are extracted from the sample set.

[0019] In an implementation of the present disclosure, the data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

[0020] In a second aspect, the present disclosure provides an intelligent load forecasting system for a substation, including:

[0021] An acquisition module is configured to acquire power load data of a substation area;

[0022] An extraction module is configured to extract training features from the power load data of the substation area; wherein the training features include power load historical data and characteristic parameters affecting the power load; and the characteristic parameters are configured with weight factors;

[0023] A training module is configured to use the power load historical data, characteristic parameters and weight factors as training input data of a recurrent neural network, train the recurrent neural network, and obtain a load prediction model for the substation;

[0024] The prediction module is configured to use the substation load prediction model to perform power load prediction.

[0025] In one implementation of the present disclosure, the intelligent load forecasting system for the substation area further includes:

[0026] The configuration module is configured to configure a weight factor for the feature parameter.

[0027] In an implementation of the present disclosure, the part of the configuration module that configures the weight factor for the characteristic parameter is configured as follows:

[0028] The characteristic parameters are taken as a characteristic set, the power load history data is taken as a category set, and the weight factors of the characteristic parameters are determined based on the attribute-weighted naive Bayes algorithm according to the characteristic set and the category set.

[0029] In an implementation of the present disclosure, the characteristic parameter is one or more of a meteorological parameter, an area parameter, and a date type.

[0030] In one implementation of the present disclosure, the recurrent neural network is an LSTM network; the LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

[0031] In one implementation of the present disclosure, the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and at the same time introduces the unit state at time t-1 into the input signal of the forgetting gate and the output gate at time t.

[0032] In one implementation of the present disclosure, the extraction module includes:

[0033] A preprocessing unit is configured to perform data preprocessing on the power load data of the substation area to obtain a sample set;

[0034] The extraction unit is configured to extract training features from the sample set.

[0035] In an implementation of the present disclosure, the data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

[0036] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer instructions, and wherein the one or more computer instructions are executed by the processor to implement a method as described in any one of the first aspects.

[0037] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the method described in any one of the first aspects is implemented.

[0038] In a fifth aspect, an embodiment of the present disclosure provides a chip, the chip including a processor, the processor being used to call a computer program in a memory to execute the method steps of the above-mentioned substation intelligent load forecasting method.

[0039] The technical effects provided by the embodiments of the present disclosure may include the following beneficial effects:

[0040] According to the technical solution provided by the embodiment of the present disclosure, the intelligent load prediction method for the substation includes: obtaining the power load data of the substation; extracting training features from the power load data of the substation; wherein the training features include power load history data and characteristic parameters affecting the power load; the characteristic parameters are configured with weight factors; the power load history data, characteristic parameters and weight factors are used as training input data of the recurrent neural network, and the recurrent neural network is trained to obtain a substation load prediction model; and the power load is predicted using the substation load prediction model. In the above technical solution, the power load is predicted by training the substation load prediction model, which improves the accuracy of load prediction in complex environments, and when training the substation load prediction model, weight factors are configured for the characteristic parameters affecting the power load, so that for special dates such as holidays and weekends, or for different seasons such as spring and winter, high weights can be given to the characteristic parameters of special dates or seasons, which further improves the accuracy of load prediction compared to all characteristic parameters not being configured with weight factors, that is, using the same weight and being unable to adjust.

[0041] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings.

[0043] Figure 1 A flow chart of a method for intelligent load forecasting in a substation according to an embodiment of the present disclosure is shown.

[0044] Figure 2 A schematic diagram of the structure of a neuron according to an embodiment of the present disclosure is shown.

[0045] Figure 3 A flowchart of the substation load prediction model training according to an embodiment of the present disclosure is shown.

[0046] Figure 4 A structural block diagram of a substation intelligent load forecasting system according to an embodiment of the present disclosure is shown.

[0047] Figure 5A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0048] Figure 6 A schematic diagram showing the structure of a computer system suitable for implementing the method according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0049] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0050] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of features, numbers, steps, behaviors, components, parts, or a combination thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, parts, or a combination thereof exist or are added.

[0051] It should also be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0052] As mentioned above, traditional forecasting methods are mainly implemented through mathematical and statistical models. The forecasting methods are simple and fast. When the accuracy requirements are not high, the forecasting results have certain reference value. With the diversification of the power market, power load data is becoming increasingly complex. Traditional forecasting methods are difficult to adapt to the nonlinearity, randomness and chaos of power load data.

[0053] Taking the above-mentioned defects into consideration, the intelligent load forecasting method for substations provided in the present invention performs power load forecasting by training substation load forecasting models, thereby improving the accuracy of load forecasting in complex environments, and configuring weight factors for characteristic parameters that affect power load when training substation load forecasting models. In this way, for special dates such as holidays and weekends, or for different seasons such as spring and winter, high weights can be given to characteristic parameters of special dates or seasons. Compared with the situation where all characteristic parameters are not configured with weight factors, that is, the same weights are used and cannot be adjusted, the accuracy of load forecasting is further improved.

[0054] Figure 1 A flow chart of a method for intelligent load forecasting in a substation according to an embodiment of the present disclosure is shown.

[0055] like Figure 1 As shown, the intelligent load prediction method for the substation area includes the following steps S110-S140:

[0056] In step S110, the power load data of the substation area is obtained;

[0057] In step S120, training features are extracted from the power load data of the substation area; wherein the training features include power load historical data and characteristic parameters affecting the power load; and the characteristic parameters are configured with weight factors;

[0058] In step S130, the power load history data, characteristic parameters and weight factors are used as training input data of a recurrent neural network, and the recurrent neural network is trained to obtain a load prediction model for the substation;

[0059] In step S140, the power load forecasting model is used to perform power load forecasting.

[0060] In one implementation of the present disclosure, the step S120 of extracting training features from the power load data of the substation area includes:

[0061] Performing data preprocessing on the power load data of the substation area to obtain a sample set;

[0062] Training features are extracted from the sample set.

[0063] The data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

[0064] Specifically, the power load data of the substation includes but is not limited to meteorological information, substation information and date type information. Meteorological information includes special weather, season, temperature, humidity changes, substation information includes substation type, scale classification, substation power load history curve, date type information includes weekdays, holidays and weekends, etc.

[0065] After obtaining the power load data of the substation, the load data needs to be preprocessed first, mainly including data quantization processing, outlier processing and data normalization processing, and the preprocessed data is used as a sample set, and then training features are extracted from the sample set, and then the extracted training features are used as training input data for the recurrent neural network.

[0066] The training features in the disclosed embodiment include power load history data and characteristic parameters that affect power load. The power load history data can be obtained from the power load history curve of the substation in the substation information, and the characteristic parameters can be obtained from the meteorological information, substation information and date type information, that is, the characteristic parameters can be one or more of the meteorological parameters, substation parameters and date types.

[0067] After extracting the historical data of power load and the characteristic parameters affecting the power load, weight factors can also be configured for the characteristic parameters based on the above training features, that is, in one implementation of the present disclosure, the intelligent load prediction method for the substation area further includes:

[0068] The step of configuring weight factors for the feature parameters.

[0069] Specifically, the step of configuring a weight factor for the characteristic parameter includes:

[0070] The characteristic parameters are taken as a characteristic set, the power load history data is taken as a category set, and the weight factors of the characteristic parameters are determined based on the attribute-weighted naive Bayes algorithm according to the characteristic set and the category set.

[0071] The naive Bayes algorithm assumes that the attributes are independent of each other under given classification feature conditions. Based on the premise of attribute condition independence, it also implies another assumption, that is, the importance of each condition attribute to the decision classification is the same, and this assumption is often not true in practical applications. Specific to the present disclosure, the characteristic parameters that affect the power load, such as the power load on weekdays and holidays in the date type, are usually different, while the power load of the substations of the same scale and classification is usually not much different, so the contribution of each characteristic parameter to the power load is different. The attribute-weighted naive Bayes algorithm assigns a suitable weight to the characteristic parameter according to the contribution of the feature set to the category set. The larger the weight of the characteristic parameter, the greater its impact on the power load prediction. In the present disclosure, the attribute-weighted naive Bayes algorithm used can refer to the technical details of the prior art, which will not be repeated here. As another implementation method, it is also possible to configure weight factors for feature parameters in the form of expert experience, or to configure weight factors in other ways, and the present disclosure does not limit this.

[0072] In the disclosed method, at least two of the characteristic parameters are configured with different weight factors. After the weight factors are configured for the characteristic parameters, the weight factors are added to the sample set as training input data for the recurrent neural network, further improving the accuracy of load prediction.

[0073] In one implementation of the present disclosure, the recurrent neural network is an LSTM network; the LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

[0074] In the disclosed method, considering that the number of hidden layers will greatly affect the performance of the recurrent neural network, by verifying different numbers of hidden layers and numbers of neurons, and considering the comprehensive prediction accuracy and online calculation time, when the number of hidden layers is three and the number of neurons in the above layers is used, the model accuracy is the highest and the calculation time is short.

[0075] In one implementation of the present disclosure, the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and at the same time introduces the unit state at time t-1 into the input signal of the forgetting gate and the output gate at time t.

[0076] Figure 2 A schematic diagram of the structure of a neuron according to an embodiment of the present disclosure is shown.

[0077] The mathematical model of the improved LSTM neural network module can be expressed as:

[0078]

[0079] in, for The forgetfulness factor of the moment, for The output signal of the output gate at all times, for The candidate cell states at time t, for The cell state at a moment in time, for The hidden layer state at time t, is the weight matrix of the forget gate, is the weight matrix of the output gate, For input The weight matrix of is the bias vector of the forget gate, is the bias vector of the output gate, for The bias vector of is the sigmoid function, is the tanh function, represents matrix element multiplication, The input of time, Represents the historical data of power load and the characteristic parameters that affect power load, Represents the weight factor.

[0080] The above-mentioned improved LSTM neural network module cancels the input gate and combines the input gate with the forget gate. On the one hand, it increases the current unit's information utilization intensity of the historical unit state, and on the other hand, it reduces the impact of abnormal cell states on the results during network training, making the convergence of the deep model training process more stable and the output results better.

[0081] In one implementation of the present disclosure, when the power load is predicted using the substation load prediction model in step S140, the power load data of the substation to be predicted is obtained, and then the historical data before time t-1 in the power load data of the substation to be predicted is input into the substation load prediction model to predict the power load at time t.

[0082] Figure 3 A flowchart of the substation load prediction model training according to an embodiment of the present disclosure is shown.

[0083] like Figure 3 The training process of the substation load forecasting model includes data preprocessing, Bayesian optimization, and model training and model optimization processes.

[0084] Specifically, in the data preprocessing part, the first step is to collect power load data and related information, mainly including the historical load data of the study area and the main factors affecting the power load, and then quantify the influencing factors, process the abnormal data, and normalize the input and output data, and divide the preprocessed historical data into training samples and test samples;

[0085] In the Bayesian optimization part, the naive Bayesian algorithm based on attribute weighting is used to extract features from the historical data of power load in the substation, calculate the weight factors of the characteristic parameters affecting the power load, and add their data to the training samples and test samples.

[0086] In the model training part, the LSTM model is trained using the training samples. In the model optimization part, after the training is completed, the test samples are used for testing. If the test error is greater than the first preset threshold, the training samples are reused for training until the test error meets the requirements. Furthermore, the trained LSTM model is tested using the real-time data of the power load in the substation. If the error is large, the training samples are reused for training until the difference between the comparison result of the prediction result and the real-time load data is less than the second preset threshold, indicating that the comparison is correct, and the training is terminated to obtain a trained substation load prediction model. Among them, the second preset threshold and the first preset threshold are set according to experience, and the two can be the same or different, and the present disclosure does not limit this.

[0087] Effect embodiment:

[0088] The data used in this example are the historical load data of a low-voltage substation in Chengdu, Sichuan Province, and the characteristic parameters that affect the power load, including six characteristics: special weather, season, temperature, humidity, substation size, and date type. The sampling period is from March 6, 2022 to March 6, 2023, with a sampling interval of 4 hours, and a total of 17,650 data. Before the experiment, the data were normalized. In this experiment, a total of three indicators were used to evaluate the model prediction effect, namely Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Accuracy (Accuracy).

[0089] The disclosed load forecasting model (denoted as NB-LSTM) is compared with the support vector machine (SVM), BP neural network, LSTM, CNN-LSTM, and GRU models, and the prediction evaluation index values ​​of each model are shown in the following table:

[0090]

[0091] As can be seen from the table above, the NB-LSTM model has the smallest prediction error compared to the other five models. Although it has the shortest training time, it has higher accuracy in prediction effect. Specifically, compared with LSTM, GRU and NB-LSTM models, the training time is shorter, indicating that removing the input gate can further accelerate the convergence speed, proving the effectiveness of the improvement of LSTM. At the same time, NB-LSTM retains the unit state link of LSTM very well, making its own use of historical information better than GRU, while CNN-LSTM further extracts deep nonlinear features, which can enhance the learning of more detailed features of time and influencing factors by the later prediction layer, thereby improving the prediction accuracy of the model in scenarios with large load fluctuations. Compared with CNN-LSTM, the NB-LSTM model combines the LSTM input gate and the forget gate into one, neutralizes the mutant unit state, and makes full use of historical information to make it the model with the best prediction effect. In summary, the NB-LSTM model is the best in terms of prediction error, effectively shortens the training time and has high prediction accuracy.

[0092] Figure 4 The structural block diagram of the intelligent load forecasting system for a substation according to an embodiment of the present disclosure is shown. The system can be implemented as part or all of an electronic device through software, hardware or a combination of both.

[0093] like Figure 4 As shown, the substation intelligent load prediction system 400 includes an acquisition module 410, an extraction module 420, a training module 430 and a prediction module 440.

[0094] The acquisition module 410 is configured to acquire the power load data of the substation area;

[0095] The extraction module 420 is configured to extract training features from the power load data of the substation area; wherein the training features include power load historical data and characteristic parameters affecting the power load; the characteristic parameters are configured with weight factors;

[0096] The training module 430 is configured to use the power load history data, characteristic parameters and weight factors as training input data of the recurrent neural network, train the recurrent neural network, and obtain a load prediction model for the substation;

[0097] The prediction module 440 is configured to perform power load prediction using the substation load prediction model.

[0098] The intelligent load prediction system for substations provided by the present invention predicts power load by training substation load prediction models, thereby improving the accuracy of load prediction in complex environments, and configuring weight factors for characteristic parameters that affect power load when training substation load prediction models. In this way, for special dates such as holidays and weekends, or for different seasons such as spring and winter, high weights can be given to characteristic parameters of special dates or seasons. Compared with not configuring weight factors for all characteristic parameters, that is, using the same weights and not being able to adjust, the accuracy of load prediction is further improved.

[0099] In one implementation of the present disclosure, the extraction module 420 includes:

[0100] A preprocessing unit is configured to perform data preprocessing on the power load data of the substation area to obtain a sample set;

[0101] The extraction unit is configured to extract training features from the sample set.

[0102] The data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

[0103] Specifically, the power load data of the substation includes but is not limited to meteorological information, substation information and date type information. Meteorological information includes special weather, season, temperature, humidity changes, substation information includes substation type, scale classification, substation power load history curve, date type information includes weekdays, holidays and weekends, etc.

[0104] After obtaining the power load data of the substation, the load data needs to be preprocessed first, mainly including data quantization processing, outlier processing and data normalization processing, and the preprocessed data is used as a sample set, and then training features are extracted from the sample set, and then the extracted training features are used as training input data for the recurrent neural network.

[0105] The training features in the disclosed embodiment include power load history data and characteristic parameters that affect power load. The power load history data can be obtained from the power load history curve of the substation in the substation information, and the characteristic parameters can be obtained from the meteorological information, substation information and date type information, that is, the characteristic parameters can be one or more of the meteorological parameters, substation parameters and date types.

[0106] After extracting the historical data of power load and the characteristic parameters affecting the power load, weight factors can also be configured for the characteristic parameters based on the above training features, that is, in one implementation of the present disclosure, the intelligent load prediction system for the substation area further includes:

[0107] The configuration module is configured to configure a weight factor for the feature parameter.

[0108] Specifically, the part of the configuration module that configures the weight factor for the characteristic parameter is configured as follows:

[0109] The characteristic parameters are taken as a characteristic set, the power load history data is taken as a category set, and the weight factors of the characteristic parameters are determined based on the attribute-weighted naive Bayes algorithm according to the characteristic set and the category set.

[0110] The naive Bayes algorithm assumes that the attributes are independent of each other under given classification feature conditions. Based on the premise of attribute condition independence, it also implies another assumption, that is, the importance of each condition attribute to the decision classification is the same, and this assumption is often not true in practical applications. Specific to the present disclosure, the characteristic parameters that affect the power load, such as the power load on weekdays and holidays in the date type, are usually different, while the power load of the substations of the same scale and classification is usually not much different, so the contribution of each characteristic parameter to the power load is different. The attribute-weighted naive Bayes algorithm assigns a suitable weight to the characteristic parameter according to the contribution of the feature set to the category set. The larger the weight of the characteristic parameter, the greater its impact on the power load prediction. In the present disclosure, the attribute-weighted naive Bayes algorithm used can refer to the technical details of the prior art, which will not be repeated here. As another implementation method, it is also possible to configure weight factors for feature parameters in the form of expert experience, or to configure weight factors in other ways, and the present disclosure does not limit this.

[0111] In the disclosed method, at least two of the characteristic parameters are configured with different weight factors. After the weight factors are configured for the characteristic parameters, the weight factors are added to the sample set as training input data for the recurrent neural network, further improving the accuracy of load prediction.

[0112] In one implementation of the present disclosure, the recurrent neural network is an LSTM network; the LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

[0113] In the disclosed method, considering that the number of hidden layers will greatly affect the performance of the recurrent neural network, by verifying different numbers of hidden layers and numbers of neurons, and considering the comprehensive prediction accuracy and online calculation time, when the number of hidden layers is three and the number of neurons in the above layers is used, the model accuracy is the highest and the calculation time is short.

[0114] In one implementation of the present disclosure, the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and introduces the cell state at time t-1 into the input signal of the forgetting gate and the output gate at time t. Figure 2 As shown, no further description is given here.

[0115] The above-mentioned improved LSTM neural network module cancels the input gate and combines the input gate with the forget gate. On the one hand, it increases the current unit's information utilization intensity of the historical unit state, and on the other hand, it reduces the impact of abnormal cell states on the results during network training, making the convergence of the deep model training process more stable and the output results better.

[0116] In one implementation of the present disclosure, when the prediction module 440 uses the substation load prediction model to perform power load prediction, it obtains the power load data of the substation to be predicted, and then inputs the historical data before time t-1 in the power load data of the substation to be predicted into the substation load prediction model to predict the power load at time t.

[0117] The present disclosure also discloses an electronic device, Figure 5 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0118] like Figure 5 As shown, the electronic device includes a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to an embodiment of the present disclosure.

[0119] Obtain area power load data;

[0120] Extracting training features from the power load data of the substation area; wherein the training features include historical power load data and characteristic parameters affecting the power load; the characteristic parameters are configured with weight factors;

[0121] The power load historical data, characteristic parameters and weight factors are used as training input data of a recurrent neural network, and the recurrent neural network is trained to obtain a load prediction model for the substation;

[0122] The power load forecasting model is used to forecast the power load.

[0123] In one implementation of the present disclosure, the intelligent load forecasting method for the substation area further includes:

[0124] The step of configuring weight factors for the feature parameters.

[0125] In one implementation of the present disclosure, the step of configuring a weight factor for the characteristic parameter includes:

[0126] The characteristic parameters are taken as a characteristic set, the power load history data is taken as a category set, and the weight factors of the characteristic parameters are determined based on the attribute-weighted naive Bayes algorithm according to the characteristic set and the category set.

[0127] In an implementation of the present disclosure, the characteristic parameter is one or more of a meteorological parameter, an area parameter, and a date type.

[0128] In one implementation of the present disclosure, the recurrent neural network is an LSTM network; the LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

[0129] In one implementation of the present disclosure, the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and at the same time introduces the unit state at time t-1 into the input signal of the forgetting gate and the output gate at time t.

[0130] In one implementation of the present disclosure, extracting training features from the power load data of the substation includes:

[0131] Performing data preprocessing on the power load data of the substation area to obtain a sample set;

[0132] Training features are extracted from the sample set.

[0133] In an implementation of the present disclosure, the data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

[0134] Figure 6 A schematic diagram showing the structure of a computer system suitable for implementing the method according to an embodiment of the present disclosure is shown.

[0135] like Figure 6 As shown, the computer system includes a processing unit, which can execute various methods in the above-mentioned embodiments according to a program stored in a read-only memory (ROM) or a program loaded from a storage part into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer system are also stored. The processing unit, ROM and RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0136] The following components are connected to the I / O interface: an input part including a keyboard, a mouse, etc.; an output part including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage part including a hard disk, etc.; and a communication part including a network interface card such as a LAN card, a modem, etc. The communication part performs a communication process via a network such as the Internet. The drive is also connected to the I / O interface as needed. Removable media, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive as needed so that the computer program read therefrom can be installed into the storage part as needed. Among them, the processing unit can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0137] In particular, according to an embodiment of the present disclosure, the method described above can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes a program code for executing the above method. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part, and / or installed from a removable medium.

[0138] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0139] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or programmable hardware. The units or modules described may also be set in a processor, and the names of these units or modules do not constitute limitations on the units or modules themselves in some cases.

[0140] As another aspect, the present disclosure further provides a chip, which includes at least one processor and can be used to implement the method involved in the above system embodiment.

[0141] In one possible design, the chip also includes a memory, which is used to store program instructions and data, and the memory is located inside or outside the processor.

[0142] As another aspect, the present disclosure further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the electronic device or computer system in the above embodiment; or a computer-readable storage medium that exists independently and is not assembled into a device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the method described in the present disclosure.

[0143] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other.

Claims

1. A method for intelligent load forecasting in a substation, characterized in that: include: Obtain area power load data; Extracting training features from the power load data of the substation area; wherein the training features include power load historical data and characteristic parameters affecting power load; the characteristic parameters are configured with weight factors; the characteristic parameters are meteorological parameters, substation area parameters, and date types; Taking the characteristic parameters as a characteristic set and the power load history data as a category set, and determining the weight factors of the characteristic parameters based on the attribute-weighted naive Bayes algorithm according to the characteristic set and the category set; The power load historical data, characteristic parameters and weight factors are used as training input data of a recurrent neural network, and the recurrent neural network is trained to obtain a load prediction model for the substation; Using the load forecasting model to forecast power load; Among them, the recurrent neural network is an LSTM network; the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and at the same time introduces the unit state at time t-1 into the input signal of the forgetting gate and the output gate at time t.

2. The intelligent load forecasting method for a substation according to claim 1, characterized in that: The LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

3. The intelligent load forecasting method for a substation according to claim 1, characterized in that: The extracting of training features from the power load data of the substation area includes: Performing data preprocessing on the power load data of the substation area to obtain a sample set; Extract training features from the sample set.

4. The intelligent load forecasting method for a substation according to claim 3 is characterized in that: The data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

5. A smart load forecasting system for a substation, characterized in that: include: An acquisition module is configured to acquire power load data of a substation area; The extraction module is configured to extract training features from the power load data of the substation area; wherein the training features include power load historical data and characteristic parameters affecting the power load; the characteristic parameters are configured with weight factors; the characteristic parameters are meteorological parameters, substation area parameters, and date types; A configuration module is configured to use the characteristic parameters as a characteristic set and the power load history data as a category set, and determine a weight factor of the characteristic parameter based on a naive Bayes algorithm with attribute weighting according to the characteristic set and the category set; A training module is configured to use the power load historical data, characteristic parameters and weight factors as training input data of a recurrent neural network, train the recurrent neural network, and obtain a load prediction model for the substation; A prediction module, configured to perform power load prediction using the substation load prediction model; Among them, the recurrent neural network is an LSTM network; the structure of each neuron is an improved LSTM neural network module; the improved LSTM neural network module replaces the output signal of the input gate with the difference between 1 and the forgetting factor, and at the same time introduces the unit state at time t-1 into the input signal of the forgetting gate and the output gate at time t.

6. The intelligent load forecasting system for substations according to claim 5 is characterized in that: The LSTM network includes three hidden layers; wherein the number of neurons in the first hidden layer is 50, the number of neurons in the second hidden layer is 20, and the number of neurons in the third hidden layer is 10.

7. The intelligent load forecasting system for substations according to claim 5, characterized in that: The extraction module comprises: A preprocessing unit is configured to perform data preprocessing on the power load data of the substation area to obtain a sample set; The extraction unit is configured to extract training features from the sample set.

8. The intelligent load forecasting system for substations according to claim 7 is characterized in that: The data preprocessing includes one or more of data quantization processing, outlier processing, and data normalization processing.

9. An electronic device, characterized in that: The method comprises a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to any one of claims 1 to 4.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 4 is implemented.

11. A chip, characterized in that: include: At least one processor, configured to implement the method according to any one of claims 1 to 4.

Images