Power system load prediction method and device and nonvolatile storage medium

Abstract

The invention discloses a power system load prediction method and device and a nonvolatile storage medium. The method comprises the following steps: acquiring historical power load data of a target area, the historical power load data comprising load information, weather data and a preset type of economic index in each historical time period; determining an external feature associated with the power system load of the target area; and iteratively executing a preset number of times to input the historical power load data and the real-time updated external features into the extended long-short-term memory network model to obtain a load prediction value output by the extended long-short-term memory network model, the extended long-short-term memory network comprising a convolutional layer for extracting local features and an attention mechanism module. The technical problem that the load of the power system cannot be predicted due to the fact that complex, nonlinear and uncertain power load data cannot be processed in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of electrical engineering, and in particular, to a method, device and non-volatile storage medium for predicting load in a power system. Background Art

[0002] In the related art, when predicting the load of a power system, the existing load prediction model is unable to accurately predict the load of the power system due to the high complexity, nonlinearity and uncertainty of the power load data.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a method, device and non-volatile storage medium for predicting the load of an electric power system, so as to at least solve the technical problem that the load of an electric power system cannot be predicted due to the inability of related technologies to handle complex, nonlinear and uncertain electric load data.

[0005] According to one aspect of an embodiment of the present application, a method for predicting a load on a power system is provided, comprising: obtaining historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in various historical time periods; determining external features associated with the power system load in the target area; iteratively executing a preset number of times the step of inputting the historical power load data and the real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the load prediction value outputted in each iteration corresponds to a different time point.

[0006] Optionally, obtaining historical power load data of the target area includes: obtaining initial historical load data, wherein the initial historical load data includes initial load information, initial weather data and initial economic indicators of preset types within each historical time period; cleaning the initial historical load data to replace missing values ​​and abnormal values ​​in the initial historical load data; and normalizing the cleaned initial historical load data to obtain historical power load data.

[0007] Optionally, determining the external characteristics associated with the power system load of the target area includes: determining multiple alternative external characteristics associated with the power system load of the target area, wherein the feature type of the alternative external characteristics includes at least one of the following: average daily temperature, average daily humidity, wind speed, season, and whether it is a holiday; using a feature selection algorithm to perform importance evaluation on each of the multiple alternative external characteristics to obtain an importance evaluation index for each alternative external feature, wherein the importance evaluation index is used to reflect the correlation between the alternative external characteristics and the power system load, and the correlation and the importance evaluation index are positively correlated; selecting an external feature from the multiple alternative external characteristics according to the size of the importance evaluation index.

[0008] Optionally, the extended long short-term memory network model also includes multiple standard long short-term memory network units and multiple extended long short-term memory network units, wherein the standard long short-term memory network units are used to extract a first dependency relationship between data input into the extended long short-term memory network model, and the extended long short-term memory network units are used to extract a second dependency relationship between data input into the extended long short-term memory network model, and the time scale corresponding to the second dependency relationship is longer than the time scale corresponding to the first dependency relationship.

[0009] Optionally, the steps of iteratively executing a preset number of times to input historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model include: a first step of obtaining real-time updated external features; a second step of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model; a third step of adding the load prediction value to the historical power load data when the number of iterations is less than a preset number of iterations, and jumping to the first step.

[0010] Optionally, after iteratively executing a preset number of times of inputting historical power load data and real-time updated external features into the extended long short-term memory network model to obtain the load prediction value output by the extended long short-term memory network model, the power system load prediction method also includes: performing inverse normalization processing on the load prediction value to obtain an inverse normalization processing result; comparing the inverse normalization processing result with the actual load value, and determining the accuracy of the extended long short-term memory network model based on the comparison result.

[0011] Optionally, after iteratively executing a preset number of times of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, the power system load prediction method also includes: adjusting the power generation plan of the power system according to the load prediction value.

[0012] According to another aspect of an embodiment of the present application, there is also provided a power system load forecasting device, including: a first processing module, used to obtain historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in each historical time period; a second processing module, used to determine external features associated with the power system load of the target area; a third processing module, used to iteratively execute a preset number of times of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load forecast value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the load forecast value outputted in each iteration corresponds to a different time point.

[0013] According to another aspect of an embodiment of the present application, a non-volatile storage medium is further provided, in which a program is stored, wherein when the program is running, a device where the non-volatile storage medium is located is controlled to execute a power system load prediction method.

[0014] According to another aspect of an embodiment of the present application, there is also provided an electronic device, including: a memory and a processor, the processor being configured to run a program stored in the memory, wherein the power system load prediction method is executed when the program is run.

[0015] According to another aspect of an embodiment of the present application, a computer program product is also provided, including a computer program, and the computer program implements a power system load prediction method when executed by a processor.

[0016] In an embodiment of the present application, the historical power load data of the target area is obtained, wherein the historical power load data includes load information, weather data and economic indicators of preset types in each historical time period; external features associated with the power system load of the target area are determined; and the steps of iteratively executing a preset number of times to input the historical power load data and the real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model are performed, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the load prediction value outputted at each iteration corresponds to a different time point. By processing the historical power load data with the extended long short-term memory network model, the purpose of obtaining high-precision load prediction results at different time points is achieved, thereby achieving the technical effect of accurately predicting the load of the power system, and further solving the technical problem that the load of the power system cannot be predicted due to the inability to process complex, nonlinear and uncertain power load data in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 It is a structural schematic diagram of a computer terminal (mobile terminal) provided according to an embodiment of the present application;

[0019] Figure 2 It is a flow chart of a method for predicting load of a power system provided according to an embodiment of the present application;

[0020] Figure 3 It is a schematic diagram of an architecture of an extended long short-term memory network provided according to an embodiment of the present application;

[0021] Figure 4 It is a flowchart of a power system load prediction process provided according to an embodiment of the present application;

[0022] Figure 5 It is a schematic diagram of prediction results of a long short-term memory network provided according to an embodiment of the present application;

[0023] Figure 6 It is a schematic diagram of prediction results of an extended long short-term memory network provided according to an embodiment of the present application;

[0024] Figure 7 It is a structural schematic diagram of a power system load prediction device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] In modern power systems, power load forecasting is crucial to optimize the allocation of power resources and improve the reliability and economy of the power grid. Accurate load forecasting helps power companies to effectively dispatch power generation under different load conditions, thereby reducing operating costs and reducing environmental impact. Power load forecasting is usually divided into short-term, medium-term and long-term forecasts. Short-term forecasts involve load changes from a few hours to a few days; medium-term forecasts focus on trends from weeks to months; and long-term forecasts look at load changes over several years. Since load is affected by many factors such as weather changes, economic activities and social behavior, accurately capturing these changes is crucial to achieving efficient power management.

[0028] Current load forecasting methods mainly include:

[0029] Traditional time series models: such as autoregressive integrated moving average (ARIMA). Although they can capture the trend of load changes to a certain extent, they assume linear relationships and cannot handle complex nonlinear patterns. In addition, they are sensitive to outliers and missing data, which affects the stability of predictions.

[0030] Machine learning algorithms: such as support vector machines (SVM), random forests, and gradient boosting trees. These methods are helpful in improving prediction performance, but they usually rely on large amounts of historical data, are sensitive to noise and outliers, and have difficulty coping with complex power load changes.

[0031] Deep learning models: such as long short-term memory (LSTM) networks. Although LSTM can capture long-term dependencies, it still has limitations when dealing with complex nonlinear relationships.

[0032] Hybrid Model:

[0033] Combining traditional methods with machine learning or deep learning techniques can improve accuracy, but model design and training are relatively complex and are still limited by the shortcomings of each component.

[0034] In summary, the prediction methods in related technologies all have the following disadvantages:

[0035] Limited prediction accuracy: When faced with large data discreteness and grayscale, the prediction accuracy of many methods will drop significantly. This is mainly because these methods have limitations in dealing with complex, nonlinear, and highly uncertain load data.

[0036] Insufficient long-term forecasting capabilities: For long-term forecasting of power systems, many methods fail to fully consider changes in long-term influencing factors such as politics, economy, and climate, resulting in inaccurate forecast results.

[0037] Poor adaptability: Some methods have poor adaptability when facing emergencies or changes in conditions and are unable to adjust the prediction model in a timely manner, thus affecting the accuracy of the prediction results.

[0038] The main reasons for the above shortcomings are as follows:

[0039] Data complexity: Power load data is highly complex, nonlinear, and uncertain, which makes it difficult for many traditional forecasting methods to process this data. Uncertainty of long-term influencing factors: The long-term load forecasting of the power system is subject to a variety of long-term influencing factors, such as policy adjustments, economic development, population growth, etc. The changes in these factors are highly uncertain and difficult to accurately predict. Limitations of the method itself: Each forecasting method has its specific scope of application and conditions. When these scopes or conditions are exceeded, the forecasting effect of the method will be affected. For example, although the neural network method has strong adaptive and nonlinear processing capabilities, it cannot give full play to its advantages when processing certain specific types of load data. Although LSTM has potential in theory, it still faces challenges in practical applications, such as high computing resource requirements, difficulty in feature selection, and sensitivity to loss functions.

[0040] Existing methods generally face the problem of insufficient capture of nonlinear relationships and complex patterns, especially when dealing with seasonal and sudden load changes. At the same time, many methods have high requirements on data quality, are easily affected by noise and outliers, and have deficiencies in computational efficiency and model interpretability.

[0041] In order to solve the above problems, the embodiments of the present application provide relevant solutions, which improve the traditional LSTM model, introduce convolutional layers and attention mechanisms, make full use of the nonlinear characteristics of power load data, improve prediction accuracy, and achieve accurate prediction of multiple future time steps. The following is a detailed description.

[0042] According to an embodiment of the present application, a method embodiment of a power system load forecasting method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0043] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing a power system load prediction method is shown. Figure 1 As shown, the computer terminal 10 (or mobile device 10) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0044] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0045] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the power system load prediction method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the above-mentioned power system load prediction method is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0046] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0047] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0048] Under the above operating environment, the embodiment of the present application provides a method for predicting load of a power system, such as Figure 2 As shown, the method comprises the following steps:

[0049] Step S202, obtaining historical power load data of the target area, wherein the historical power load data includes load information, weather data and preset economic indicators in each historical time period;

[0050] In the technical solution provided in step S202, the step of obtaining the historical power load data of the target area includes: obtaining initial historical load data, wherein the initial historical load data includes initial load information, initial weather data and initial economic indicators of preset types in each historical time period; cleaning the initial historical load data to replace missing values ​​and abnormal values ​​in the initial historical load data; normalizing the cleaned initial historical load data to obtain historical power load data.

[0051] As an optional implementation, power load data in the target area over the past few years can be collected. The data includes hourly load information, weather data (such as temperature, humidity, wind speed) and related economic indicators. The collected data is then cleaned, missing values ​​and outliers are processed to ensure the integrity and accuracy of the data. Common processing methods include interpolation, mean substitution, etc. In addition, in order to eliminate the impact of different magnitudes, the data can be normalized to ensure that all input features are in the same range. Min-Max normalization or Z-score normalization can be used for normalization.

[0052] Step S204, determining external features associated with the power system load in the target area;

[0053] In the technical solution provided in step S204, the step of determining the external features associated with the power system load of the target area includes: determining multiple alternative external features associated with the power system load of the target area, wherein the feature type of the alternative external features includes at least one of the following: average daily temperature, average daily humidity, wind speed, season, and whether it is a holiday; using a feature selection algorithm to perform importance evaluation on each of the multiple alternative external features to obtain an importance evaluation index for each alternative external feature, wherein the importance evaluation index is used to reflect the correlation between the alternative external feature and the power system load, and the correlation and the importance evaluation index are positively correlated; selecting an external feature from the multiple alternative external features based on the size of the importance evaluation index.

[0054] As an optional implementation, external factors related to load changes can be selected as model input features according to the characteristics of the power load. Common features include average daily temperature, humidity, wind speed, season, holidays, etc.

[0055] In order to accurately screen out suitable external features, feature selection algorithms (such as random forest, LASSO regression, principal component analysis) can be used to evaluate the importance of features to determine the features that have the greatest impact on load forecasting, and finally select a preset number (such as 10) of the most important features for modeling.

[0056] Step S206, iteratively executes a preset number of times to input historical power load data and real-time updated external features into the extended long short-term memory network model to obtain the load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the time point corresponding to the output load prediction value in each iteration is different.

[0057] In the technical solution provided in step S206, if Figure 3As shown, the extended long short-term memory network model also includes multiple standard long short-term memory network units and multiple extended long short-term memory network units, wherein the standard long short-term memory network units are used to extract a first dependency relationship between data input into the extended long short-term memory network model, and the extended long short-term memory network units are used to extract a second dependency relationship between data input into the extended long short-term memory network model, and the time scale corresponding to the second dependency relationship is longer than the time scale corresponding to the first dependency relationship.

[0058] Optionally, by introducing a convolutional layer before the SLSTM layer and the XLSTM layer, local features of the input data can be extracted to improve the model's sensitivity to short-term fluctuations. In addition, the self-attention mechanism is integrated into the model, which enables the model to pay more attention to key features when making multi-step predictions, helping to improve the overall performance of the model.

[0059] It should also be noted that the extended long short-term memory network unit in the embodiment of the present application includes a scalar long short-term memory network unit and a matrix long short-term memory network unit, wherein: the scalar long short-term memory network unit retains the characteristics of scalar storage, and introduces a multi-layer hybrid mechanism so that the memory of different layers interacts with each other, which is used to achieve the flexibility of storing information flow. The matrix long short-term memory network unit uses a memory unit in the form of a matrix to improve the storage capacity of the model and store sparse information. The matrix memory unit is updated by the covariance update rule to process complex information for the model. In addition, an exponential gating mechanism is introduced in the extended long short-term memory network model to provide information filtering capabilities and improve memory and forgetting processes. And a residual stacking architecture similar to Transformer is introduced to integrate the scalar long short-term memory network unit and the matrix long short-term memory network unit into the residual block to form a deep network in a residual stacking manner.

[0060] In some embodiments of the present application, when training the extended long short-term memory network model, the data set can be divided into a training set, a validation set, and a test set for model training and performance evaluation. The training set accounts for 80%, the validation set accounts for 10%, and the test set accounts for 10%. During the training process, the training set can be used to train the XLSTM model, the cross-validation method can be used to select the best hyperparameters, and the validation set can be used to evaluate the model performance.

[0061] In addition, you can also use adaptive learning rate optimization algorithms (such as Adam, RMSprop) to improve the training efficiency of the model and reduce fluctuations during training.

[0062] As an optional implementation, the step of iteratively executing a preset number of times to input historical power load data and real-time updated external features into the extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model includes:

[0063] The first step is to obtain external features that are updated in real time;

[0064] The second step is to input the historical power load data and the real-time updated external features into the extended long short-term memory network model to obtain the load forecast value output by the extended long short-term memory network model;

[0065] In the third step, when the number of iterations is less than the preset number of iterations, the load prediction value is added to the historical power load data and the process jumps to the first step.

[0066] In this way, the trained XLSTM model can be used to predict the power load for multiple time steps in the future. The model will perform comprehensive calculations based on historical data and external features and output future load forecast values. Specifically, when making predictions, the already obtained prediction results can be used as input data for iterative predictions. For example, the power data of the previous 23 hours can be used as the input of the XLSTM network to predict the data of the 24th hour.

[0067] In some embodiments of the present application, after iteratively executing a preset number of times of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, the power system load prediction method also includes: performing inverse normalization processing on the load prediction value to obtain an inverse normalization processing result; comparing the inverse normalization processing result with the actual load value, and determining the accuracy of the extended long short-term memory network model based on the comparison result.

[0068] As an optional implementation, after iteratively executing a preset number of times of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, the power system load prediction method also includes: adjusting the power generation plan of the power system based on the load prediction value.

[0069] In summary, the process of implementing load forecasting by extending the long short-term memory network model is as follows: Figure 4 As shown, the following steps are included:

[0070] Step S402, obtaining a load data set and preprocessing the load data set;

[0071] Step S404, training the extended long short-term memory network model based on the preprocessed load data set;

[0072] Step S406, performing evaluation test on the trained extended long short-term memory network model.

[0073] The method adopts the steps of obtaining historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in various historical time periods; determining external features associated with the power system load of the target area; iteratively executing a preset number of times of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the load prediction value outputted at each iteration corresponds to a different time point. By processing the historical power load data with the extended long short-term memory network model, the purpose of obtaining high-precision load prediction results at different time points is achieved, thereby realizing the technical effect of accurately predicting the load of the power system, and further solving the technical problem that the load of the power system cannot be predicted due to the inability to process complex, nonlinear and uncertain power load data in related technologies.

[0074] In addition, when the extended long short-term memory network model provided in the embodiment of the present application is used for load forecasting, the accuracy is higher than that of the traditional long short-term memory network model used in the related art. Figure 5 and Figure 6 As shown. Figure 5 To use the prediction results of the traditional long short-term memory network model, Figure 6 This is the prediction result of the extended long short-term memory network model. It can be seen that the waveform of the prediction result of the extended long short-term memory network model is closer to the actual detection result, that is, the prediction accuracy of the extended long short-term memory network model provided by the present application is significantly better than the prediction method in the related art.

[0075] Specifically, the extended long short-term memory network model provided in the embodiments of the present application has the following advantages:

[0076] Improve prediction accuracy: By introducing convolutional layers, the extended long short-term memory network model can more effectively extract local features in power load data. At the same time, combined with the attention mechanism, the model's ability to capture key information can be enhanced, thereby improving prediction accuracy.

[0077] Support multi-step prediction: The extended LSTM network model can handle multi-step prediction tasks and accurately predict the power load in multiple time steps in the future, meeting the actual needs of power companies in scheduling, management and decision-making.

[0078] Enhance model robustness: In response to outliers and noise in power load data, the extended long short-term memory network model improves anti-interference ability by optimizing algorithms and structures, ensuring the stability of prediction results and reducing the risks caused by inaccurate predictions.

[0079] Realize real-time prediction: The extended long short-term memory network model has been optimized in design and algorithm, which improves the training and prediction speed, meets the power system's requirements for real-time data processing, and realizes efficient application in real-time systems.

[0080] By extending the long short-term memory network model, the embodiment of the present application effectively solves the main problems faced in the field of power load forecasting, such as insufficient capture of nonlinear relationships, lack of multi-step forecasting capabilities, sensitivity to outliers and noise, high consumption of computing resources, and lack of dynamic adaptability. This method provides a more advanced and reliable solution for power load forecasting, which helps to promote the further development of smart grids and energy management. It provides strong support for the intelligence and efficiency of power systems.

[0081] The present application provides a power system load prediction device, Figure 7 is a schematic diagram of the structure of the device. Figure 7 It can be seen that the device includes: a first processing module 70, which is used to obtain historical power load data of the target area, wherein the historical power load data includes load information, weather data and preset types of economic indicators in each historical time period; a second processing module 72, which is used to determine the external features associated with the power system load of the target area; a third processing module 74, which is used to iteratively execute a preset number of times to input the historical power load data and the real-time updated external features into the extended long short-term memory network model to obtain the load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolution layer and an attention mechanism module for extracting local features, and the time point corresponding to the load prediction value output in each iteration is different.

[0082] In some embodiments of the present application, the step of the first processing module 70 acquiring historical power load data of the target area includes: acquiring initial historical load data, wherein the initial historical load data includes initial load information, initial weather data, and initial economic indicators of preset types within each historical time period; cleaning the initial historical load data to replace missing values ​​and abnormal values ​​in the initial historical load data; and normalizing the cleaned initial historical load data to obtain historical power load data.

[0083] In some embodiments of the present application, the step of the second processing module 72 determining the external features associated with the power system load of the target area includes: determining multiple alternative external features associated with the power system load of the target area, wherein the feature type of the alternative external features includes at least one of the following: average daily temperature, average daily humidity, wind speed, season, and whether it is a holiday; using a feature selection algorithm to perform importance evaluation on each of the multiple alternative external features to obtain an importance evaluation index for each alternative external feature, wherein the importance evaluation index is used to reflect the correlation between the alternative external feature and the power system load, and the correlation and the importance evaluation index are positively correlated; selecting an external feature from the multiple alternative external features based on the size of the importance evaluation index.

[0084] In some embodiments of the present application, the extended long short-term memory network model also includes a plurality of standard long short-term memory network units and a plurality of extended long short-term memory network units, wherein the standard long short-term memory network units are used to extract a first dependency relationship between data input into the extended long short-term memory network model, and the extended long short-term memory network units are used to extract a second dependency relationship between data input into the extended long short-term memory network model, and the time scale corresponding to the second dependency relationship is longer than the time scale corresponding to the first dependency relationship.

[0085] In some embodiments of the present application, the third processing module 74 iteratively executes a preset number of times to input the historical power load data and the real-time updated external features into the extended long short-term memory network model, and the step of obtaining the load prediction value output by the extended long short-term memory network model includes: a first step, obtaining the real-time updated external features; a second step, inputting the historical power load data and the real-time updated external features into the extended long short-term memory network model, and obtaining the load prediction value output by the extended long short-term memory network model; a third step, when the number of iterations is less than the preset number of iterations, adding the load prediction value to the historical power load data and jumping to the first step.

[0086] In some embodiments of the present application, after iteratively executing a preset number of times of inputting historical power load data and real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, the power system load prediction device is also used to: perform inverse normalization processing on the load prediction value to obtain an inverse normalization processing result; compare the inverse normalization processing result with the actual load value, and determine the accuracy of the extended long short-term memory network model based on the comparison result.

[0087] In some embodiments of the present application, after iteratively executing the step of inputting historical power load data and real-time updated external features into an extended long short-term memory network model a preset number of times to obtain a load prediction value output by the extended long short-term memory network model, the power system load prediction device is also used to: adjust the power generation plan of the power system according to the load prediction value.

[0088] It should be noted that the various modules in the above-mentioned power system load forecasting device can be program modules (for example, a set of program instructions that implement a certain specific function) or hardware modules. For the latter, it can be expressed in the following forms, but is not limited to this: the expression form of each of the above-mentioned modules is a processor, or the functions of each of the above-mentioned modules are implemented by a processor.

[0089] According to an embodiment of the present application, a non-volatile storage medium is also provided, in which a program is stored, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the following power system load prediction method: obtaining historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in each historical time period; determining external features associated with the power system load of the target area; iteratively executing a preset number of times to input the historical power load data and the real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the load prediction value outputted in each iteration corresponds to a different time point.

[0090] According to an embodiment of the present application, an electronic device is also provided, including: a memory and a processor, the processor being used to run a program stored in the memory, wherein the following power system load forecasting method is executed when the program is running: obtaining historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in each historical time period; determining external features associated with the power system load of the target area; iteratively executing a preset number of times of inputting the historical power load data and the real-time updated external features into an extended long short-term memory network model to obtain a load forecast value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the load forecast value outputted in each iteration corresponds to a different time point.

[0091] According to an embodiment of the present application, a computer program product is also provided, including a computer program, which implements the following power system load forecasting method when executed by a processor: obtaining historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in each historical time period; determining external features associated with the power system load of the target area; iteratively executing a preset number of times to input the historical power load data and the real-time updated external features into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the time point corresponding to the load prediction value output in each iteration is different.

[0092] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0094] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0095] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0096] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the relevant technology or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0097] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for predicting load in a power system, characterized in that: include: Acquire historical power load data of the target area, wherein the historical power load data includes load information, weather data, and economic indicators of preset types in each historical time period; determining external characteristics associated with power system loads in the target area; The step of iteratively executing a preset number of times to input the historical power load data and the external features updated in real time into an extended long short-term memory network model to obtain a load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the time point corresponding to the load prediction value output in each iteration is different.

2. The power system load prediction method according to claim 1, characterized in that: Obtaining historical power load data for the target area includes: Acquiring initial historical load data, wherein the initial historical load data includes initial load information, initial weather data, and initial economic indicators of a preset type in each of the historical time periods; Cleaning the initial historical load data to replace missing values ​​and abnormal values ​​in the initial historical load data; The initial historical load data after cleaning is normalized to obtain the historical power load data.

3. The power system load prediction method according to claim 1, characterized in that: Determining external characteristics associated with the power system load of the target area includes: Determine a plurality of candidate external features associated with the power system load of the target area, wherein the feature type of the candidate external features includes at least one of the following: average daily temperature, average daily humidity, wind speed, season, and whether it is a holiday; Using a feature selection algorithm to evaluate the importance of each candidate external feature among the multiple candidate external features, and obtain an importance evaluation index of each candidate external feature, wherein the importance evaluation index is used to reflect the correlation between the candidate external feature and the power system load, and the correlation is positively correlated with the importance evaluation index; The external feature is selected from a plurality of candidate external features according to the magnitude of the importance evaluation index.

4. The power system load prediction method according to claim 1, characterized in that: The extended long short-term memory network model also includes a stacked long short-term memory unit and an extended long short-term memory network unit, wherein the stacked long short-term memory unit is used to extract a first dependency relationship between data input into the extended long short-term memory network model, and the extended long short-term memory network unit is used to extract a second dependency relationship between data input into the extended long short-term memory network model, and the time scale corresponding to the second dependency relationship is longer than the time scale corresponding to the first dependency relationship.

5. The power system load prediction method according to claim 1, characterized in that: The step of iteratively executing a preset number of times to input the historical power load data and the external features updated in real time into the extended long short-term memory network model to obtain the load prediction value output by the extended long short-term memory network model comprises: The first step is to obtain the external features updated in real time; The second step is to input the historical power load data and the external features updated in real time into the extended long short-term memory network model to obtain the load prediction value output by the extended long short-term memory network model; The third step is to add the load prediction value to the historical power load data when the number of iterations is less than the preset number of iterations, and jump to the first step.

6. The power system load prediction method according to claim 1, characterized in that: After iteratively executing the step of inputting the historical power load data and the external features updated in real time into the extended long short-term memory network model for a preset number of times to obtain the load prediction value output by the extended long short-term memory network model, the power system load prediction method further includes: Performing inverse normalization processing on the load prediction value to obtain an inverse normalization processing result; The inverse normalization processing result and the actual load value are compared, and the accuracy of the extended long short-term memory network model is determined according to the comparison result.

7. The power system load prediction method according to claim 1, characterized in that: After iteratively executing the step of inputting the historical power load data and the external features updated in real time into the extended long short-term memory network model for a preset number of times to obtain the load prediction value output by the extended long short-term memory network model, the power system load prediction method further includes: The power generation plan of the power system is adjusted according to the load forecast value.

8. A power system load prediction device, characterized in that: include: A first processing module is used to obtain historical power load data of a target area, wherein the historical power load data includes load information, weather data and economic indicators of preset types in each historical time period; a second processing module for determining external characteristics associated with the power system load of the target area; The third processing module is used to iteratively execute the step of inputting the historical power load data and the external features updated in real time into the extended long short-term memory network model for a preset number of times to obtain the load prediction value output by the extended long short-term memory network model, wherein the extended long short-term memory network includes a convolutional layer and an attention mechanism module for extracting local features, and the time point corresponding to the load prediction value output in each iteration is different.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute the power system load prediction method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the processor is used to run a program stored in the memory, wherein the program, when running, executes the power system load forecasting method described in any one of claims 1 to 7.

11. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the power system load prediction method according to any one of claims 1 to 7.

Images