Power load prediction method and device of power system and electronic equipment

Abstract

The invention discloses a power load prediction method and device of a power system and electronic equipment. The method comprises the steps that historical power load data of a power system is acquired, the historical power load data comprises a plurality of power load time sequences, and different power load time sequences are used for representing time sequences corresponding to different variables influencing power loads; inputting the plurality of power load time sequences into a power load prediction model, and predicting the plurality of power load time sequences of the power system in a future time period by using the power load prediction model to obtain predicted power load data, the predicted power load data is used for representing the predicted power load data of the power system in the future time period. According to the invention, the technical problem of low power load prediction precision in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and in particular to a method, a device and an electronic device for predicting power load in a power system. Background Art

[0002] As an important basic work in the power sector, power load forecasting plays a key role in the smooth and efficient operation of the power system. With the access of large-scale distributed power sources and new loads such as electric vehicle charging piles, the volatility, time-varying and randomness of power loads have increased significantly, increasing the difficulty of power load forecasting in the power system. Traditional forecasting methods, due to their linear assumptions, are difficult to capture the complex nonlinear characteristics of power load data, so the forecasting accuracy is low.

[0003] In recent years, deep learning models have been used in related technologies to predict power load. Due to the powerful nonlinear relationship learning ability of deep learning models, they show significant advantages in processing long sequence data and multi-feature situations. However, existing models still have limitations in processing the uncertainty, randomness and complex time correlation of power load data, resulting in insufficient mining of power load time series characteristics, which in turn leads to low power load prediction accuracy in related technologies.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method, device and electronic equipment for predicting power load in a power system, so as to at least solve the technical problem of low power load prediction accuracy in the related art.

[0006] According to one aspect of an embodiment of the present invention, a method for predicting electric power load of an electric power system is provided, comprising: obtaining historical electric power load data of the electric power system, the historical electric power load data comprising a plurality of electric power load time series, wherein different electric power load time series are used to represent time series corresponding to different variables affecting the electric power load; inputting the plurality of electric power load time series into an electric power load prediction model, and using the electric power load prediction model to predict the plurality of electric load time series of the electric power system in a future time period to obtain predicted electric power load data, wherein the predicted electric power load data are used to represent predicted electric power load data of the electric power system in a future time period.

[0007] Furthermore, the power load forecasting model includes: a time series decomposition module, a feature extraction module, and a feature enhancement module. Multiple power load time series are input into the power load forecasting model, and the power load forecasting model is used to predict multiple power load time series of the power system in the future time period to obtain target predicted power load data, including: using the time series decomposition module to decompose multiple power load time series to obtain trend time series components and seasonal time series components, wherein the seasonal time series component is a component that shows periodic fluctuations in multiple power load time series, and the trend time series component is a component that shows a long-term stable change trend in multiple power load time series; using the feature extraction module to extract features from the seasonal time series components to obtain comprehensive time window load features; using the feature enhancement module to enhance features of the comprehensive time window load features and the trend time series components to obtain predicted power load data.

[0008] Furthermore, the feature extraction module includes: a time window embedding layer, a horizontal time window analyzer, and a vertical time window analyzer. The feature extraction module is used to extract features of seasonal time series components to obtain comprehensive time window load features, including: using the time window embedding layer to establish a time window load feature matrix for the first load point and the second load point in the seasonal time series component, wherein the second load point is used to represent the load point in the time window adjacent to the first load point, and the time window load feature matrix is ​​used to represent the time window load features at multiple moments; using the horizontal time window analyzer to extract the time window load features at multiple moments to obtain the horizontal time window load features at multiple moments; using the vertical time window analyzer to extract the horizontal time window load features at multiple moments to obtain the comprehensive time window load features at multiple moments.

[0009] Furthermore, the feature enhancement module is used to perform feature enhancement on the comprehensive time window load characteristics and the trend time series component to obtain predicted power load data, including: using the feature enhancement module to perform feature enhancement on the comprehensive time window load characteristics to obtain the seasonal time series component output result; using the feature enhancement module to perform feature enhancement on the trend time series component to obtain the trend time series component output result; based on the trend time series component output result and the seasonal time series component output result, the predicted power load data is determined.

[0010] Furthermore, the feature enhancement module includes: a first convolution layer and a second convolution layer, wherein the feature enhancement module is used to enhance the comprehensive time window load characteristics to obtain a seasonal time series component output result, and the feature enhancement module is used to enhance the trend time series component to obtain a trend time series component output result, including: using the first convolution layer to extract the characteristics of the trend time series component and the comprehensive time window load characteristics respectively to obtain a first trend characteristic and a first seasonal characteristic; using the second convolution layer to extract the characteristics of the trend time series component and the comprehensive time window load characteristics respectively to obtain a second trend characteristic and a second seasonal characteristic, wherein the convolution kernel size of the second convolution layer is larger than the convolution kernel size of the first convolution layer; determining the first trend convolution parameters based on the first activation function, the first trend characteristics and the second trend characteristics, and determining the first seasonal convolution parameters based on the first activation function, the first seasonal characteristics and the second seasonal characteristics; determining the second trend convolution parameters based on the second activation function, the second trend characteristics and the first trend characteristics, and determining the second seasonal convolution parameters based on the second activation function, the second seasonal characteristics and the first seasonal characteristics; obtaining the trend time series component output result based on the first trend convolution parameters and the second trend convolution parameters; obtaining the seasonal time series component output result based on the first seasonal convolution parameters and the second seasonal convolution parameters.

[0011] Furthermore, a time window load characteristic matrix is ​​established for the first load point and the second load point in the seasonal time series component using a time window embedding layer, including: constructing preset dimensional spatial characteristics of the seasonal time series component based on the first load point, the second load point and the preset time window parameters, wherein the preset dimensional spatial characteristics are used to represent the spatial characteristics in the preset dimension; performing a linear transformation on the preset dimensional spatial characteristics to obtain a time window load characteristic matrix.

[0012] Furthermore, the horizontal time window analyzer includes: a local multilayer perceptron, a first normalization layer, and the horizontal time window analyzer is used to analyze the local correlation of the time window load characteristics at multiple moments to obtain the horizontal time window load characteristics at multiple moments, including: using the local multilayer perceptron to extract features of the time window load characteristics at multiple moments to obtain the first time window load characteristics at multiple moments; using the first normalization layer to normalize the first time window load characteristics at multiple moments to obtain the second time window load characteristics at multiple moments; transposing the second time window load characteristics at multiple moments to obtain the horizontal time window load characteristics at multiple moments.

[0013] Furthermore, the longitudinal time window analyzer includes: a global multilayer perceptron, a second normalization layer, and the longitudinal time window analyzer is used to extract the transverse time window load characteristics of multiple moments to obtain the comprehensive time window load characteristics of multiple moments, including: using the global multilayer perceptron to extract the transverse time window load characteristics of multiple moments to obtain the third time window load characteristics; using the second normalization layer to normalize the third time window load characteristics of multiple moments to obtain the fourth time window load characteristics of multiple moments; transposing the fourth time window load characteristics of multiple moments to obtain the comprehensive time window load characteristics of multiple moments.

[0014] According to another aspect of an embodiment of the present invention, there is also provided an electric power load prediction device for an electric power system, comprising: an acquisition module, used to acquire historical electric power load data of the electric power system, the historical electric power load data comprising a plurality of electric power load time series, wherein different electric power load time series are used to represent time series corresponding to different variables affecting the electric power load; a prediction module, used to input the plurality of electric power load time series into an electric power load prediction model, and use the electric load prediction model to predict the plurality of electric load time series of the electric power system in a future time period to obtain predicted electric power load data, wherein the predicted electric power load data are used to represent predicted electric power load data of the electric power system in a future time period.

[0015] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0016] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0017] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0018] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0019] In an embodiment of the present invention, first, the historical power load data of the power system is obtained, and then the historical power load data is input into the power load prediction model, and the power load prediction model is used to predict multiple power load time series of the power system in the future time period to obtain predicted power load data. It is easy to notice that the historical power load data contains multiple power load time series, each power load time series corresponds to a variable that affects the power load prediction. By considering multiple variables that affect the power load, the power load model is designed at the same time so that it can capture the global correlation and local correlation between the power load time series, so as to achieve the purpose of deep mining of power load characteristics, thereby achieving the technical effect of improving the prediction accuracy, and thus solving the technical problem of low power load prediction accuracy in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0021] Figure 1 is a flow chart of a method for predicting power load in a power system according to an embodiment of the present invention;

[0022] Figure 2 is a flow chart of an optional method for predicting power load in a power system according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of a power load prediction device for a power system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] According to an embodiment of the present invention, an embodiment of a method for predicting power load in an electric power system 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.

[0027] Figure 1 is a flow chart of a method for predicting power load in a power system according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0028] Step S102, obtaining historical power load data of the power system, the historical power load data including a plurality of power load time series, wherein different power load time series are used to represent time series corresponding to different variables affecting the power load.

[0029] The historical power load data in the above steps is used to describe the power consumption record of the power system in a certain period of time in the past. The historical load data is presented in the form of a time series. The load data at each time point reflects the demand of the power system at that moment. The historical power load data includes multiple power load time series. The role of historical power load data can include but is not limited to power load forecasting, analyzing power demand patterns, and formulating electricity price strategies.

[0030] The power load time series in the above steps is used to describe a series of load values ​​arranged in chronological order in the historical power load data. Different power load time series represent the changes in power load corresponding to different variables over time, where different variables are used to describe multiple variables affecting the power load. The multiple variables may include but are not limited to temperature, humidity, power consumption, commercial electricity, and electricity prices, etc. The power load time series may include but are not limited to temperature-related load series, humidity-related load series, power consumption-related time series, etc. The specific power load time series needs to be determined based on the specific influencing variables, which is not limited here.

[0031] The acquisition methods in the above steps may include but are not limited to the following methods:

[0032] The first method is to monitor and record the operation status of the power network in real time through the Supervisory Control and Data Acquisition (SCADA) system to obtain historical power load data.

[0033] The second method is to obtain historical power load data by obtaining user-side power consumption information through smart meters and obtaining historical electricity price information through public data of power companies.

[0034] The above method for obtaining historical power load data is only an example. The specific method for obtaining it needs to be determined according to the actual situation of the power system and is not limited here.

[0035] In an optional embodiment, the historical power load data of the power system is obtained through multiple channels such as smart meters and public electricity prices of power companies. The historical power load data contains multiple power load time series, and each series represents the changes over time of different variables affecting the power load. For example, the temperature-related load series reflects the impact of ambient temperature fluctuations on power demand, and the social activity-related load series records the change pattern of power consumption caused by special events or holidays. Through the training and analysis of multiple power load time series, the periodicity, seasonality and event-driven characteristics of historical power load data can be understood, so as to facilitate the subsequent power load data prediction based on historical power load data.

[0036] In an optional embodiment, the historical power load data will be passed through a reversible instance normalization module, which is used to remove non-stationary information in the data and solve the problem of data distribution drift. Reversible instance normalization includes normalization and denormalization. First, the initial input data is normalized. For each input power load time series x (i) ∈R 1×M (i=1,2,…,N), uses its own mean and variance to perform instance normalization at time t, and the calculation formula is as follows:

[0037]

[0038] The input power load time series x is analyzed by using the mean and variance (i) Standardization is performed and the calculation formula is as follows:

[0039]

[0040] In the formula, represents the representation of the input power load time series at time t, M represents x (i) The length of the power load time series, μ (i) and σ (i) Respectively represent x (i) Calculate the mean and variance, α and β are the learnable parameters of the model, ε is the random error, For the time series x (i) Standardized results. After standardization, the power load time series is expressed as The corresponding output is

[0041] Accordingly, the output result of the power load forecasting model, i.e. the predicted power load data, is inversely standardized, and the calculation process is as follows:

[0042]

[0043] In the formula, This is the output result of the model before inverse normalization. For input variables The corresponding output result.

[0044] Step S104, input multiple power load time series into the power load prediction model, use the power load prediction model to predict multiple power load time series of the power system in the future time period, and obtain predicted power load data, wherein the predicted power load data is used to represent the predicted power load data of the power system in the future time period.

[0045] The power load forecasting model in the above steps is used to describe a deep learning model used to analyze historical power load data and predict future power demand. The types of power load forecasting models may include but are not limited to deep learning models including long short-term memory networks and convolutional neural networks. The specific power load forecasting model needs to be determined based on the nature of the power load data and actual demand, and is not limited here. The role of the power load forecasting model may include but is not limited to power load forecasting, demand response, and resource planning.

[0046] The future time period in the above steps is used to describe the time range in which the power load forecasting model will make predictions. The types of future time periods may include but are not limited to the next few hours, days, weeks or months. The specific future time period depends on the requirements of the forecasting task and the capabilities of the model, and is not limited here.

[0047] The predicted power load data in the above steps is used to describe the time series of future power system demand predicted by the power load forecasting model. The role of the predicted power load data can include but is not limited to operation optimization, economic decision-making and power system planning based on the predicted power load data.

[0048] In an optional embodiment, multiple power load time series are input into a power load forecasting model. The power load forecasting model can analyze and obtain the periodic characteristics of historical power load data by training multiple historical power load time series. Based on the periodic characteristics of the acquired historical power load data, the power load forecasting model can predict the power load time series for future time periods and obtain predicted power load data for future time periods. Furthermore, based on the predicted power load data, the power resource allocation and the economic dispatch strategy of the power market can be improved, thereby providing strong support for the efficient, stable and economical operation of the power system.

[0049] In an embodiment of the present invention, first, the historical power load data of the power system is obtained, and then the historical power load data is input into the power load prediction model, and the power load prediction model is used to predict multiple power load time series of the power system in the future time period to obtain predicted power load data. It is easy to notice that the historical power load data contains multiple power load time series, each power load time series corresponds to a variable that affects the power load prediction. By considering multiple variables that affect the power load, the power load model is designed at the same time so that it can capture the global correlation and local correlation between the power load time series, so as to achieve the purpose of deep mining of power load characteristics, thereby achieving the technical effect of improving the prediction accuracy, and thus solving the technical problem of low power load prediction accuracy in related technologies.

[0050] Optionally, the power load forecasting model includes: a time series decomposition module, a feature extraction module, and a feature enhancement module. Multiple power load time series are input into the power load forecasting model, and the power load forecasting model is used to predict multiple power load time series of the power system in the future time period to obtain target predicted power load data, including: using the time series decomposition module to decompose multiple power load time series to obtain trend time series components and seasonal time series components, wherein the seasonal time series component is a component that shows periodic fluctuations in multiple power load time series, and the trend time series component is a component that shows a long-term stable change trend in multiple power load time series; using the feature extraction module to extract features from the seasonal time series components to obtain comprehensive time window load features; using the feature enhancement module to enhance features of the comprehensive time window load features and the trend time series components to obtain predicted power load data.

[0051] The above-mentioned time series decomposition module is used to describe a technology that decomposes the original power load time series into trend time series components and seasonal time series components. The types of time series decomposition may include but are not limited to average pooling operations in deep learning models, wavelet transform decomposition in signal processing, and Fourier transform decomposition. The role of the time series decomposition module may include but is not limited to decomposition. The model can better understand and process different characteristics in the power load data, provide clear signal components for subsequent feature extraction and prediction, thereby improving the accuracy and stability of the prediction.

[0052] The above-mentioned feature extraction module is used to describe the technology of identifying and extracting key features from the decomposed seasonal time series components. The types of feature extraction modules may include but are not limited to convolutional neural networks, long short-term memory networks, etc. The specific feature extraction module needs to be determined according to the model structure design and is not limited here. The role of the feature extraction module may include but is not limited to extracting key information from the original data, providing more focused and meaningful input for the prediction model, and helping to enhance the model's expressiveness and prediction accuracy.

[0053] The above-mentioned feature enhancement module is used to describe the process of further strengthening feature representation based on feature extraction. The types of feature enhancement modules may include but are not limited to improved convolutional neural networks, residual connections, etc. The specific feature enhancement module needs to be determined according to the model design and feature type, and is not limited here. The role of the feature enhancement module may include but is not limited to helping the model to more comprehensively understand the inherent laws of power load data and improve prediction accuracy.

[0054] The above-mentioned comprehensive time window load characteristics are used to describe the characteristics extracted from the seasonal time series components and take into account the correlation between the load data within the time window.

[0055] In an optional embodiment, by inputting multiple power load time series into a power load forecasting model including a time series decomposition module, a feature extraction module and a feature enhancement module, the power load forecasting model first uses the time series decomposition module to decompose multiple power load time series into trend time series components and seasonal time series components, so that the model can effectively identify trends and seasonal changes in power load data, and then extract features for the characteristics of different time series components, and finally obtain highly accurate target forecast power load data. This time series processing method of the power load forecasting model can effectively improve the power load forecasting accuracy of the model, and has important practical application value for the optimized operation of the power system, the economic dispatch of the power market, and the intelligent power management of users.

[0056] In an optional embodiment, the time series decomposition module decomposes multiple power load time series into trend time series components and seasonal time series components, and then extracts features from the trend time series components and seasonal time series components respectively. First, the moving average method is used to calculate the average of the power load time series over a period of time in order to smooth the periodic terms and highlight the long-term trend of the power load time series. After obtaining the trend time series component of the power load time series, the seasonal time series component is obtained by subtracting the trend time series component from the input power load time series. The calculation formula for the moving average is as follows:

[0057]

[0058] In the formula, A t represents the sliding average at time point t, n LS is the sliding window size, B i represents the original power load time series at the i-th time step.

[0059] For a multivariate time series at time t The decomposition formula is as follows:

[0060]

[0061] X trend and X season They represent the trend time series component and seasonal time series component decomposed by the time series decomposition module. The trend time series component contains the long-term and rough information of the time series, while the seasonal time series component reflects the short-term and detailed information of the time series. After the time series is decomposed, it is more convenient for the model to process dynamically changing data.

[0062] Optionally, the feature extraction module includes: a time window embedding layer, a horizontal time window analyzer, and a vertical time window analyzer. The feature extraction module is used to extract features of seasonal time series components to obtain comprehensive time window load features, including: using the time window embedding layer to establish a time window load feature matrix for the first load point and the second load point in the seasonal time series component, wherein the second load point is used to represent the load point in the time window adjacent to the first load point, and the time window load feature matrix is ​​used to represent the time window load features at multiple moments; using the horizontal time window analyzer to extract the time window load features at multiple moments to obtain the horizontal time window load features at multiple moments; using the vertical time window analyzer to extract the horizontal time window load features at multiple moments to obtain the comprehensive time window load features at multiple moments.

[0063] The above-mentioned time window embedding layer is used to describe the network layer used to construct the time window load feature matrix in the feature extraction module, where the size of the time window may include but is not limited to 6, 12, 18, 24, etc. The specific size of the time window needs to be determined according to the sampling frequency of the data, which is not limited here. The function of the time window embedding layer may include but is not limited to associating the local information of the load data with the context, constructing the time window load feature matrix, constructing a high-dimensional feature space, etc.

[0064] The above-mentioned horizontal time window analyzer is used to describe the network layer used to extract the local correlation of the time window load feature matrix. The role of the horizontal time window analyzer may include but is not limited to helping the model to deeply understand the short-term fluctuations and local patterns of power load.

[0065] The above-mentioned longitudinal time window analyzer is used to describe the global correlation between the features of the horizontal time window for mining. The role of the longitudinal time window analyzer may include but is not limited to helping the model identify the long-term trend and long-distance dependence of the power load.

[0066] The first load point mentioned above is used to describe the load value at a specific time point in the time series.

[0067] The above-mentioned transverse time window load characteristics are used to describe the characteristics extracted from the time window load characteristic matrix by the transverse time window analyzer. The transverse time window load characteristics reflect the relationship between load points in the same time window.

[0068] In an optional embodiment, a feature extraction module for processing seasonal time series components in a power load forecasting model is constructed by a time window embedding layer, a horizontal time window analyzer, and a vertical time window analyzer. Then, the time window embedding layer is used to construct a time window load feature matrix, the horizontal time window analyzer is used to mine the local correlation of the time window load feature matrix to obtain the horizontal time window feature, and the vertical time window analyzer is used to mine the global correlation between the horizontal time window features to obtain the comprehensive time window feature. Through the above feature extraction structure, the model can capture the global correlation and local correlation of the power load data, thereby improving the accuracy of the power load forecast.

[0069] Optionally, a feature enhancement module is used to perform feature enhancement on the comprehensive time window load characteristics and trend time series components to obtain predicted power load data, including: using the feature enhancement module to perform feature enhancement on the comprehensive time window load characteristics to obtain a seasonal time series component output result; using the feature enhancement module to perform feature enhancement on the trend time series component to obtain a trend time series component output result; based on the trend time series component output result and the seasonal time series component output result, determining the predicted power load data.

[0070] The above-mentioned seasonal time series component output results are used to describe the feature results extracted and enhanced from the seasonal time series components by the feature enhancement module. The seasonal time series component output results reflect the periodicity and short-term fluctuation information in the load data, which can help the model identify and predict seasonal or periodic electricity demand.

[0071] The above trend time series component output results are used to describe the feature enhancement results obtained by the feature enhancement module when the model processes long-term trend information. The trend time series component output results reflect the overall change trend of electricity demand over time, which can help the model understand and predict long-term electricity load growth or decline.

[0072] In an optional embodiment, the power load forecasting model can capture the short-term fluctuations and long-term trends of power load from different angles by processing the seasonal time component and the trend time series component separately. In the forecasting stage of the model, the output results of the seasonal time component and the trend time series component are aggregated by summing, weighted averaging, etc. to generate the final forecasted power load data. This aggregation operation can integrate the model's understanding of short-term changes and long-term trends, provide a comprehensive forecast of power demand, help power companies to accurately plan and dispatch power resources, and also provide users with more sophisticated power consumption guidance, promoting the economic and efficient operation of the power system.

[0073] Optionally, the feature enhancement module includes: a first convolution layer and a second convolution layer, wherein the feature enhancement module is used to perform feature enhancement on the comprehensive time window load feature to obtain a seasonal time series component output result, and the feature enhancement module is used to perform feature enhancement on the trend time series component to obtain a trend time series component output result, including: using the first convolution layer to perform feature extraction on the trend time series component and the comprehensive time window load feature respectively to obtain a first trend feature and a first seasonal feature; using the second convolution layer to perform feature extraction on the trend time series component and the comprehensive time window load feature respectively to obtain a second trend feature and a second seasonal feature, wherein the convolution kernel size of the second convolution layer is The convolution kernel size is larger than that of the first convolution layer; the first trend convolution parameters are determined based on the first activation function, the first trend characteristics and the second trend characteristics, and the first seasonal convolution parameters are determined based on the first activation function, the first seasonal characteristics and the second seasonal characteristics; the second trend convolution parameters are determined based on the second activation function, the second trend characteristics and the first trend characteristics, and the second seasonal convolution parameters are determined based on the second activation function, the second seasonal characteristics and the first seasonal characteristics; based on the first trend convolution parameters and the second trend convolution parameters, the trend time series component output result is obtained; based on the first seasonal convolution parameters and the second seasonal convolution parameters, the seasonal time series component output result is obtained.

[0074] The above-mentioned first convolution layer and second convolution layer are convolution network layers used for feature enhancement in the feature enhancement module. The convolution kernel size of the second convolution layer is larger than the convolution kernel size of the first convolution layer. The size of the convolution kernel may include but is not limited to 1×1, 3×3, 5×5, etc. The specific convolution kernel size needs to be determined according to the network structure and feature structure, and is not limited here. In an optional case, the convolution kernel size of the first convolution layer can be set to 1×1, and the convolution kernel size of the second convolution kernel can be set to 3×3.

[0075] The above-mentioned first trend feature and second trend feature are used to describe that when the feature enhancement module processes the trend time series component, the first trend feature and the second trend feature respectively extracted by the first convolution layer and the second convolution layer are the results of the model learning trend information.

[0076] The above-mentioned first seasonal characteristics and second seasonal characteristics are used to describe that when the feature enhancement module processes the comprehensive time window load characteristics, the first seasonal characteristics and the second seasonal characteristics extracted by the first convolutional layer and the second convolutional layer are the results of learning seasonal information.

[0077] The above-mentioned first activation function is used to describe the activation function located after the first convolutional layer, which is used to help the model learn more complex feature representations. The first activation function may include but is not limited to Gaussian Error Linear Unit activation function (Gaussian Error Linear Unit, referred to as GELU), rectified linear unit activation function (Rectified LinearUnit, referred to as ReLU), S-curve activation function Sigmoid, etc. The specific first activation function type can be selected according to the role of the activation function, which is not limited here.

[0078] The above-mentioned second activation function is used to describe the activation function located after the second convolutional layer, which is used to help the model learn more complex feature representations. The second activation function may include but is not limited to Gaussian Error Linear Unit (GELU), Rectified Linear Unit (ReLU), S-curve activation function Sigmoid, etc. The specific second activation function type can be selected according to the role of the activation function, which is not limited here.

[0079] The first trend convolution parameter is used to describe the feature obtained by performing a dot product operation between the first trend feature and the second trend feature after the first activation function activates the first trend feature.

[0080] The first seasonal convolution parameter is used to describe the feature obtained by performing a dot multiplication operation between the first seasonal feature and the second seasonal feature after the first activation function activates the first seasonal feature.

[0081] The second trend convolution parameter is used to describe the feature obtained by performing a dot multiplication operation on the second trend feature after the second activation function activates the second trend feature.

[0082] The above-mentioned second seasonal convolution parameter is used to describe the feature obtained by performing a dot multiplication operation on the second seasonal feature after the second activation function activates the second seasonal feature.

[0083] In an optional embodiment, the first trend feature, the first seasonal feature, the second trend feature, and the second seasonal feature extracted by the first convolution layer and the second convolution layer, and then the activation function is used to further determine the convolution parameters, so that the feature enhancement module can effectively learn the key information in the power load data. This multi-scale feature extraction method combines the learning of nonlinear activation functions and adaptive convolution parameters, significantly improving the performance of the model in processing power load forecasting tasks, enabling the model to more comprehensively understand the timing and periodicity characteristics of the power load, thereby generating more accurate load forecasting results.

[0084] In an optional embodiment, two convolutional layers of different sizes are used to extract features from the time series, and the convolutional layers are used again in an aggregated manner. The first convolutional layer uses a fine-grained convolution kernel that can capture the unique modality of the time series. The calculation formula is as follows:

[0085]

[0086] In the formula, σ(·) represents the first activation function, K (i) represents the input feature, f conv1 (·) and f conv2 (·) represent the first convolutional layer and the second convolutional layer respectively, and the symbol represents the dot product operation of the matrix, It indicates that the output of the first convolution layer may be the first trend convolution parameter and the first seasonal convolution parameter.

[0087] The second convolutional layer uses a rough convolution kernel, which is mainly used to capture the long-term correlation hidden in the time series. The calculation formula is as follows:

[0088]

[0089] In the formula, σ(·) represents the second activation function, K (i) represents the input feature, f conv1 (·) and f conv2 (·) represent the first convolutional layer and the second convolutional layer, respectively. represents the dot product operation of the matrix, The output of the second convolution layer may be a second trend convolution parameter and a second seasonal convolution parameter.

[0090] The feature outputs of the two convolutional layers are summed and input to the last convolutional layer for processing. The output is as follows:

[0091]

[0092] In the formula, f conv3 (.) represents the third convolutional layer, Represent the outputs of the first convolutional layer, the second convolutional layer, and the third convolutional layer respectively.

[0093] After the trend time series component and the seasonal time series component are extracted by the feature enhancement module, they are predicted using the fully connected layer, and finally the predicted power load data is obtained by summing up. The calculation formula is as follows:

[0094]

[0095] In the formula, and Respectively represent the output results of trend time series component and seasonal time series component, It is the predicted power load data after the two components are superimposed. trend and b trend are the weights and bias parameters of the fully connected layer of the trend time series component, W season and b season are the weights and bias parameters of the seasonal time series components.

[0096] Optionally, a time window load characteristic matrix is ​​established for the first load point and the second load point in the seasonal time series component using a time window embedding layer, including: constructing preset dimensional spatial characteristics of the seasonal time series component based on the first load point, the second load point and preset time window parameters, wherein the preset dimensional spatial characteristics are used to represent spatial characteristics in a preset dimension; and performing a linear transformation on the preset dimensional spatial characteristics to obtain a time window load characteristic matrix.

[0097] The above-mentioned preset time window parameters are used to describe the time window size λ required for constructing the preset dimensional spatial features based on the first load point. The size of the time window may include but is not limited to 6, 12, 18, 24, etc. The specific time window size needs to be determined according to the data sampling frequency, which is not limited here. The role of the preset time window parameters may include but is not limited to adjusting the model's ability to capture local and periodic features by adjusting the time window parameters.

[0098] In an optional embodiment, by determining the first load point and the preset time window parameters, the model can construct spatial features representing the load within a specific time scale when processing seasonal time series components, thereby linearly transforming these preset dimensional spatial features, mapping the information to a high-dimensional representation space that is more conducive to feature extraction and analysis, thereby obtaining a time window load feature matrix. The time window load feature matrix contains comprehensive information of the load point and its adjacent points, providing structured input for subsequent horizontal and vertical time window analysis.

[0099] In an optional embodiment, it is assumed that the decomposed representation of multiple power load time series is: The load value at each moment m, that is, the first load point, is used as the center, and the left and right lengths are λ to construct the time window feature. For the load values ​​at the first and last moments of the time series, the load values ​​at the first and last moments of the time series are copied when constructing the time window feature. The modified time series becomes

[0100] Take the load value x at time m m As the center, construct a feature vector v with a length of 2λ+1 m , the formula is as follows:

[0101] v m =[x m-λ ,…,x m-1 ,x m ,x m+1 ,…,x m+λ ]∈R 1×(2λ+1) ;

[0102] In the formula, v m represents the characteristic vector, λ represents the time window size, and m represents the load value at time m.

[0103] According to the above formula, the i-th seasonal component can be obtained The high-dimensional feature space of is as follows:

[0104] V (i) =[v1,v2,…,v M ] T ∈R M×(2λ+1) ;

[0105] Where V (i) is the i-th seasonal component The high-dimensional feature space of , T represents the transposition operation, and M represents M seasonal components.

[0106] Next, the time window feature space is transformed using a linear function. This processing method can obtain the global transient characteristics of the load in a higher-dimensional space. The linear transformation formula is as follows:

[0107] u m =v m W+b;

[0108] In the formula, u m v m The feature representation after linear transformation, W∈R (2λ+1)×D is the weight matrix, b∈R 1×D is the bias parameter, D is the dimension of the time window feature after linear transformation, and the symbol · represents the dot product of the matrix. All features are linearly transformed as follows:

[0109] U (i) =[u1,u2,…,u M ] T ∈R M×D ;

[0110] U (i) is the i-th seasonal component The representation form after linear transformation of the features is the comprehensive time window load features, and T represents the transposition operation.

[0111] Optionally, the horizontal time window analyzer includes: a local multilayer perceptron, a first normalization layer, and the horizontal time window analyzer is used to analyze the local correlation of the time window load characteristics at multiple moments to obtain the horizontal time window load characteristics at multiple moments, including: using the local multilayer perceptron to extract features of the time window load characteristics at multiple moments to obtain the first time window load characteristics at multiple moments; using the first normalization layer to normalize the first time window load characteristics at multiple moments to obtain the second time window load characteristics at multiple moments; transposing the second time window load characteristics at multiple moments to obtain the horizontal time window load characteristics at multiple moments.

[0112] The above-mentioned local multilayer perceptron is used to describe the network layer for extracting local features of time window load features. The role of the local multilayer perceptron may include but is not limited to extracting the first time window load features from the time window load features, that is, capturing the local correlation of load data within a specific time window, which helps the model understand short-term trends and periodic patterns and improve prediction accuracy.

[0113] The above-mentioned first time window load feature is used to describe the output after the local multilayer perceptron extracts the time window load feature. The first time window load feature represents the local feature representation obtained after the load data in the time window at each time point is processed by the neural network.

[0114] The above-mentioned first normalization layer is used to describe the network layer that normalizes the load characteristics of the first time window so that the data meets the standardization requirements. The types of the first normalization layer may include but are not limited to batch normalization (BatchNormalization, abbreviated as BN), layer normalization (Layer Normalization, abbreviated as LN), instance normalization (Instance Normalization, abbreviated as IN), etc. The specific first normalization layer needs to be determined according to the scenario requirements and model structure, and is not limited here.

[0115] The above-mentioned second time window load characteristics are used to describe the results after the first time window load characteristics are standardized by the first standardization layer. The second time window load characteristics have a uniform scale and distribution. The role of the second time window load characteristics includes but is not limited to retaining the feature information extracted by the local multi-layer perceptron, ensuring the statistical consistency of the data, and providing standardized input for subsequent analysis and prediction tasks.

[0116] In an optional embodiment, the horizontal time window analyzer extracts the first time window load characteristics within each time window through a local multilayer perceptron, obtains the second time window load characteristics through the first normalization layer, and finally generates the horizontal time window load characteristics through a transposition operation. This series of steps together constitutes a complete process for capturing the local correlation of power loads, providing strong support for power load forecasting.

[0117] In an optional embodiment, the process of the horizontal time window analyzer performing local correlation analysis on the time window load characteristics at multiple moments is as follows:

[0118] First, at time m∈[1,M], the time window load features are extracted through a local multilayer perceptron, and the calculation formula is as follows:

[0119]

[0120] In the formula, Ω (i) ∈R M×D is the output of the local multilayer perceptron, i.e., the load characteristics of the first time window, f LocMLP (·) is the expression of the local multilayer perceptron, is the time window load characteristic of the feature matrix at time m.

[0121] The output of the local multilayer perceptron Ω (i) It will be normalized by the first normalization layer, which is calculated as follows:

[0122]

[0123] In the formula, represents the output of the horizontal time window analyzer, i.e., the load characteristics of the second time window, f norm (·) represents the function expression of the first normalization layer. The load characteristics of the second time window The horizontal time window load characteristics are obtained by matrix transposition. The calculation formula is as follows:

[0124]

[0125] Where, L (i) represents the load characteristics of the horizontal time window, represents the load characteristics of the second time window, and T represents the transposition operation.

[0126] Optionally, the longitudinal time window analyzer includes: a global multilayer perceptron, a second normalization layer, and the longitudinal time window analyzer is used to extract the transverse time window load characteristics of multiple moments to obtain the comprehensive time window load characteristics of multiple moments, including: using the global multilayer perceptron to extract the transverse time window load characteristics of multiple moments to obtain the third time window load characteristics; using the second normalization layer to normalize the third time window load characteristics of multiple moments to obtain the fourth time window load characteristics of multiple moments; transposing the fourth time window load characteristics of multiple moments to obtain the comprehensive time window load characteristics of multiple moments.

[0127] The above-mentioned global multilayer perceptron is used to describe the network layer used to process and extract global patterns and correlations of long sequence data features. The role of the global multilayer perceptron can include but is not limited to capturing and understanding the hidden global correlations of load data between different time windows, and then understanding the macro trends and periodic patterns of power load.

[0128] The above-mentioned third time window load feature is used to describe the output of the global multi-layer perceptron after feature extraction of the horizontal time window load feature. The third time window load feature can be used to represent the comprehensive feature representation of load data in different time windows, including long-range dependencies and global pattern information.

[0129] The above-mentioned second normalization layer is used to describe the network layer that normalizes the load characteristics of the third time window so that the data meets the standardization requirements. The types of the second normalization layer may include but are not limited to batch normalization (BatchNormalization, abbreviated as BN), layer normalization (Layer Normalization, abbreviated as LN), instance normalization (Instance Normalization, abbreviated as IN), etc. The specific second normalization layer needs to be determined according to the scenario requirements and model structure, and is not limited here.

[0130] The above-mentioned fourth time window load characteristics are used to describe the output after the second standardization layer processes the third time window load characteristics. The role of the fourth time window load characteristics may include but is not limited to providing standardized and structured input for subsequent prediction and analysis steps to ensure that the model can extract effective information from long-range dependencies.

[0131] In an optional embodiment, the longitudinal time window analyzer is composed of a global multi-layer perceptron and a second normalization layer. In the feature extraction process, the global multi-layer perceptron is first used to perform deep feature extraction on the transverse time window load features of multiple moments generated by the transverse time window analyzer to generate the third time window load features. This process can reveal the global correlation between the features of different time windows, thereby capturing the long-term trend and periodic changes of the power load. Subsequently, the second normalization layer normalizes the third time window load features and converts them into the fourth time window load features, ensuring the consistency and comparability of these features under different time windows, and improving the training stability and prediction accuracy of the model.

[0132] In an optional embodiment, the process of the longitudinal time window analyzer performing global correlation analysis on the transverse time window load characteristics at multiple moments is as follows:

[0133] For each dimension d∈[1,2,...,D] of the horizontal time window load feature, after being processed by the global multilayer perceptron, its output calculation formula is:

[0134]

[0135] In the formula, represents the load characteristics of the horizontal time window, f GLoMLP (.) is the expression of the global multilayer perceptron, Represents the output of the global multilayer perceptron, i.e., the load characteristics of the third time window.

[0136] After the above transformations are performed on the features of each dimension, the output Γ of the global multilayer perceptron is obtained. (i) ∈R D×M Then the second normalization layer is used for normalization, and the calculation formula is as follows:

[0137]

[0138] In the formula, represents the load characteristics of the fourth time window, f norm (·) represents the function expression of the second normalization layer, Γ d Represents the load characteristics of the third time window.

[0139] Output of the longitudinal time window analyzer After matrix transposition, it will be used as the output of the longitudinal time window analyzer. The output calculation formula is as follows:

[0140]

[0141] In the formula, represents the load characteristics of the fourth time window, G (i)Represents the comprehensive time window load characteristics.

[0142] In an optional embodiment, Figure 2 is a flow chart of an optional method for predicting power load in a power system according to an embodiment of the present invention. Figure 2 As shown, the method first decomposes the input historical power load data into a power load time series; obtains a seasonal time series component and a trend time series component, and then inputs the seasonal time series component into a feature extraction module including a time window embedding layer, a horizontal time window analyzer, and a vertical time window analyzer for feature extraction, and then performs a dimensionality reduction operation on the extracted comprehensive time window features and inputs them into a feature enhancement module; at the same time, the trend time series component is input into the feature enhancement module, and the results of the feature enhancement of the two time series components are summed after passing through a fully connected layer to obtain an output result, namely, the predicted power load data.

[0143] According to another aspect of an embodiment of the present invention, a power load prediction device for a power system is also provided. The device can execute the power load prediction method for a power system provided in the above embodiment. The specific implementation method and preferred application scenario are the same as the above embodiment and will not be repeated here.

[0144] Figure 3 is a schematic diagram of a power load prediction device for a power system according to an embodiment of the present invention. Figure 3 As shown, the device includes: an acquisition module 302, used to acquire historical power load data of the power system, the historical power load data including multiple power load time series, wherein different power load time series are used to represent time series corresponding to different variables affecting the power load; a prediction module 304, used to input multiple power load time series into a power load prediction model, and use the power load prediction model to predict multiple power load time series of the power system in a future time period to obtain predicted power load data, wherein the predicted power load data is used to represent the predicted power load data of the power system in the future time period.

[0145] Optionally, the power load forecasting model includes: a time series decomposition module, a feature extraction module, and a feature enhancement module. The forecasting module includes: a decomposition unit, which is used to use the time series decomposition module to decompose multiple power load time series to obtain trend time series components and seasonal time series components, wherein the seasonal time series components are components that show periodic fluctuations in multiple power load time series, and the trend time series components are components that show long-term stable change trends in multiple power load time series; a feature extraction unit, which is used to use the feature extraction module to extract features from the seasonal time series components to obtain comprehensive time window load features; a feature enhancement unit, which is used to use the feature enhancement module to perform feature enhancement on the comprehensive time window load features and trend time series components to obtain predicted power load data.

[0146] Optionally, the feature extraction module includes: a time window embedding layer, a horizontal time window analyzer, a vertical time window analyzer, and a feature extraction unit, including: an establishment subunit, used to use the time window embedding layer to establish a time window load feature matrix for the first load point and the second load point in the seasonal time series component, wherein the second load point is used to represent the load point in the time window adjacent to the first load point, and the time window load feature matrix is ​​used to represent the time window load characteristics at multiple moments; a first feature extraction subunit, used to use the horizontal time window analyzer to extract the time window load characteristics at multiple moments to obtain the horizontal time window load characteristics at multiple moments; a second feature extraction subunit, used to use the vertical time window analyzer to extract the horizontal time window load characteristics at multiple moments to obtain the comprehensive time window load characteristics at multiple moments.

[0147] Optionally, the feature enhancement unit includes: a first feature enhancement subunit, used to use the feature enhancement module to perform feature enhancement on the comprehensive time window load characteristics to obtain a seasonal time series component output result; a second feature enhancement subunit, used to use the feature enhancement module to perform feature enhancement on the trend time series component to obtain a trend time series component output result; a determination subunit, used to determine the predicted power load data based on the trend time series component output result and the seasonal time series component output result.

[0148] Optionally, the feature enhancement module includes: a first convolution layer, a second convolution layer, a first feature enhancement subunit, and a second feature enhancement subunit, including: using the first convolution layer to extract features of the trend time series component and the comprehensive time window load feature, respectively, to obtain a first trend feature and a first seasonal feature; using the second convolution layer to extract features of the trend time series component and the comprehensive time window load feature, respectively, to obtain a second trend feature and a second seasonal feature, wherein the convolution kernel size of the second convolution layer is larger than the convolution kernel size of the first convolution layer; determining the first trend convolution parameter based on the first activation function, the first trend feature, and the second trend feature, and determining the first seasonal convolution parameter based on the first activation function, the first seasonal feature, and the second seasonal feature; determining the second trend convolution parameter based on the second activation function, the second trend feature, and the first trend feature, and determining the second seasonal convolution parameter based on the second activation function, the second seasonal feature, and the first seasonal feature; obtaining the trend time series component output result based on the first trend convolution parameter and the second trend convolution parameter; obtaining the seasonal time series component output result based on the first seasonal convolution parameter and the second seasonal convolution parameter.

[0149] Optionally, a subunit is established, including: constructing preset dimensional spatial characteristics of seasonal time series components based on a first load point, a second load point and a preset time window parameter, wherein the preset dimensional spatial characteristics are used to represent spatial characteristics in a preset dimension; and performing a linear transformation on the preset dimensional spatial characteristics to obtain a time window load characteristic matrix.

[0150] Optionally, the horizontal time window analyzer includes: a local multilayer perceptron, a first normalization layer, and a first feature extraction subunit, including: using the local multilayer perceptron to extract features of time window load characteristics at multiple moments to obtain first time window load characteristics at multiple moments; using the first normalization layer to normalize the first time window load characteristics at multiple moments to obtain second time window load characteristics at multiple moments; transposing the second time window load characteristics at multiple moments to obtain horizontal time window load characteristics at multiple moments.

[0151] Optionally, the longitudinal time window analyzer includes: a global multilayer perceptron, a second normalization layer, and a second feature extraction subunit, including: using the global multilayer perceptron to extract features of the transverse time window load characteristics at multiple moments to obtain third time window load characteristics; using the second normalization layer to normalize the third time window load characteristics at multiple moments to obtain fourth time window load characteristics at multiple moments; transposing the fourth time window load characteristics at multiple moments to obtain comprehensive time window load characteristics at multiple moments.

[0152] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0153] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0154] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0155] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0156] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0157] 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.

[0158] 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.

[0159] In addition, each functional unit in each embodiment of the present invention 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.

[0160] 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 invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of 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 invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0161] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting power load in a power system, characterized in that: include: Acquire historical power load data of the power system, wherein the historical power load data includes a plurality of power load time series, wherein different power load time series are used to represent time series corresponding to different variables affecting the power load; The multiple power load time series are input into a power load prediction model, and the power load prediction model is used to predict the multiple power load time series of the power system in the future time period to obtain predicted power load data, wherein the predicted power load data is used to represent the predicted power load data of the power system in the future time period.

2. The method for predicting power load of a power system according to claim 1, characterized in that: The power load forecasting model includes: a time series decomposition module, a feature extraction module, and a feature enhancement module. The multiple power load time series are input into the power load forecasting model, and the power load forecasting model is used to forecast the multiple power load time series of the power system in the future time period to obtain target forecast power load data, including: Decomposing the multiple power load time series by using the time series decomposition module to obtain trend time series components and seasonal time series components, wherein the seasonal time series components are components showing periodic fluctuations in the multiple power load time series, and the trend time series components are components showing long-term stable change trends in the multiple power load time series; Using the feature extraction module to extract features from the seasonal time series components to obtain comprehensive time window load features; The feature enhancement module is used to enhance the features of the comprehensive time window load feature and the trend time series component to obtain the predicted power load data.

3. The method for predicting power load of a power system according to claim 2, characterized in that: The feature extraction module includes: a time window embedding layer, a horizontal time window analyzer, and a vertical time window analyzer. The feature extraction module is used to extract features of the seasonal time series components to obtain comprehensive time window load features, including: Using the time window embedding layer to establish a time window load characteristic matrix for the first load point and the second load point in the seasonal time series component, wherein the second load point is used to represent the load point in the time window adjacent to the first load point, and the time window load characteristic matrix is ​​used to represent the time window load characteristics at multiple moments; The horizontal time window analyzer is used to extract the time window load characteristics of the multiple moments to obtain the horizontal time window load characteristics of the multiple moments; The longitudinal time window analyzer is used to extract the transverse time window load characteristics of the multiple moments to obtain the comprehensive time window load characteristics of the multiple moments.

4. The method for predicting power load of a power system according to claim 2, characterized in that: The feature enhancement module is used to enhance the features of the comprehensive time window load features and the trend time series components to obtain the predicted power load data, including: Using the feature enhancement module to enhance the comprehensive time window load feature to obtain a seasonal time series component output result; Using the feature enhancement module to perform feature enhancement on the trend time series component to obtain an output result of the trend time series component; The predicted power load data is determined based on the trend time series component output result and the seasonal time series component output result.

5. The method for predicting power load of a power system according to claim 4, characterized in that: The feature enhancement module includes: a first convolution layer and a second convolution layer. The feature enhancement module is used to perform feature enhancement on the comprehensive time window load feature to obtain a seasonal time series component output result. The feature enhancement module is used to perform feature enhancement on the trend time series component to obtain the trend time series component output result, including: Using the first convolutional layer to extract features from the trend time series component and the comprehensive time window load feature, respectively, to obtain a first trend feature and a first seasonal feature; The second convolution layer is used to extract features of the trend time series component and the comprehensive time window load feature to obtain a second trend feature and a second seasonal feature, wherein the convolution kernel size of the second convolution layer is larger than the convolution kernel size of the first convolution layer; Determine a first trend convolution parameter based on a first activation function, the first trend feature, and the second trend feature, and determine a first seasonal convolution parameter based on the first activation function, the first seasonal feature, and the second seasonal feature; Determine a second trend convolution parameter based on a second activation function, the second trend feature, and the first trend feature, and determine a second seasonal convolution parameter based on the second activation function, the second seasonal feature, and the first seasonal feature; Based on the first trend convolution parameter and the second trend convolution parameter, obtaining the trend time series component output result; Based on the first seasonal convolution parameter and the second seasonal convolution parameter, the seasonal time series component output result is obtained.

6. The method for predicting power load of a power system according to claim 3, characterized in that: The time window embedding layer is used to establish a time window load characteristic matrix for the first load point and the second load point in the seasonal time series component, including: Constructing a preset dimensional spatial feature of the seasonal time series component based on the first load point, the second load point and a preset time window parameter, wherein the preset dimensional spatial feature is used to represent a spatial feature in a preset dimension; The preset dimensional space features are linearly transformed to obtain the time window load feature matrix.

7. The method for predicting power load of a power system according to claim 3, characterized in that: The horizontal time window analyzer includes: a local multilayer perceptron, a first normalization layer, and uses the horizontal time window analyzer to analyze the local correlation of the time window load characteristics at multiple moments to obtain the horizontal time window load characteristics at multiple moments, including: Using the local multilayer perceptron to extract the time window load characteristics of the multiple moments, to obtain the first time window load characteristics of the multiple moments; Normalizing the first time window load characteristics at the multiple moments using the first normalization layer to obtain the second time window load characteristics at the multiple moments; The second time window load characteristics at the multiple moments are transposed to obtain the horizontal time window load characteristics at the multiple moments.

8. The method for predicting power load of a power system according to claim 3, characterized in that: The longitudinal time window analyzer includes: a global multi-layer perceptron, a second normalization layer, and the longitudinal time window analyzer is used to extract the transverse time window load characteristics of the multiple moments to obtain the comprehensive time window load characteristics of the multiple moments, including: Using the global multi-layer perceptron to extract the load characteristics of the horizontal time windows at the multiple moments, to obtain the load characteristics of the third time window; Using the second normalization layer to normalize the third time window load characteristics at the multiple moments, to obtain fourth time window load characteristics at the multiple moments; The fourth time window load characteristics of the multiple moments are transposed to obtain the comprehensive time window load characteristics of the multiple moments.

9. A power load forecasting device for a power system, characterized in that: include: An acquisition module, used for acquiring historical power load data of the power system, wherein the historical power load data includes a plurality of power load time series, wherein different power load time series are used to represent time series corresponding to different variables affecting the power load; A prediction module is used to input the multiple power load time series into a power load prediction model, and use the power load prediction model to predict the multiple power load time series of the power system in the future time period to obtain predicted power load data, wherein the predicted power load data is used to represent the predicted power load data of the power system in the future time period.

10. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 8 when running.

Images