Regional electrical load long-range prediction method considering multi-source distributed novel energy storage

Abstract

The invention discloses a regional electrical load long-range prediction method considering multi-source distributed novel energy storage, and the method comprises the steps: constructing an electrical load prediction model for a power system, and the electrical load prediction model comprises a coding part and a decoding part; acquiring sample data which is time sequence data of the electric load data; encoding the time series data into encoding output features in the encoding part; and the decoding part is used for decoding the sample data. The prediction result is the sum of the output component characteristic # imgabs0 # and the trend component characteristic # imgabs1 # of the frequency adaptive decomposition module. According to the method, a power grid manager can be effectively helped to recognize the potential supply and demand imbalance problem in advance in planning and operation, a corresponding coping strategy is formulated, the flexibility and adaptability of a power grid are improved, and decision support is provided for dispatching optimization and operation management of the power grid.

Description

Technical Field

[0001] The present invention relates to a power load forecasting technology, in particular to a regional power load long-term forecasting method taking into account multi-source distributed new energy storage. Background Art

[0002] In the context of responding to climate change and promoting energy transformation on a global scale, countries have elevated the development and utilization of new energy to the strategic level of energy policy. The gradual depletion of fossil energy and the increasingly serious environmental burden it brings have made energy structure optimization a key issue for global sustainable development. Clean energy such as wind and solar energy are gradually replacing traditional fossil energy due to their inexhaustible characteristics, becoming an important force in the adjustment of the global energy structure. The transformation of the energy structure not only directly contributes to the effective control of carbon emissions, but also provides a strong guarantee for global energy security by improving energy utilization efficiency and promoting green economic growth.

[0003] Against the backdrop of this global energy revolution, Liaoning Province, as an energy powerhouse in Northeast China, has a natural advantage in developing new energy on a large scale, thanks to its abundant natural resources, especially in wind and solar energy. Liaoning Province's vast land area and diverse climatic conditions provide a unique foundation for the large-scale development of wind and solar energy. At the same time, the national and local governments have also introduced a number of policies, from financial subsidies, technical research and development support to market mechanism innovation, to encourage and promote the development and application of new energy in a multi-pronged manner. These policies have not only injected strong impetus into the development of Liaoning Province's new energy industry, but also laid a solid foundation for the region's strategic position in the national energy structure adjustment.

[0004] However, the development of new energy is not smooth sailing. Although new energy such as wind and solar energy has the advantages of being clean and renewable, its intermittent and volatile characteristics also bring unprecedented challenges to the dispatch and operation of the power system. The power system must maintain a dynamic balance between real-time power generation and power consumption. The instability of new energy makes this balance difficult to maintain, thus posing a huge risk to the safe operation of the power grid. Therefore, how to ensure the stability of the power system while developing new energy has become a core issue that Liaoning Province must solve in the process of promoting the development of new energy.

[0005] To achieve this goal, accurate electricity load forecasting has become a crucial link. The effective use of new energy, especially the rational planning of its installed capacity and layout, must be based on accurate predictions of future electricity loads. The future development scale and layout of new energy in Liaoning Province will not only be affected by policies and resource conditions, but also depend to a large extent on the scientific prediction of future electricity loads. This kind of prediction can not only provide data support for the planning and dispatching of the power system, but also guide the rational layout and large-scale development of new energy projects, ensuring that new energy can be efficiently absorbed and applied in the future power system.

[0006] Therefore, how to scientifically predict the future electricity load in Liaoning Province has become a key link in ensuring the effective implementation of Liaoning Province's new energy development strategy. Therefore, it is necessary to develop an algorithm that can accurately predict electricity load through in-depth analysis of residents' electricity consumption habits and accurate grasp of long-term electricity consumption data. Summary of the invention

[0007] The purpose of the present invention is to provide an accurate and efficient electricity load forecasting algorithm, and to construct an electricity load forecasting model based on it, so as to provide support for the scientific planning of the future new energy installed capacity.

[0008] To achieve the above object, the present invention provides the following solutions:

[0009] A method for forecasting regional power load prospects taking into account multi-source distributed new energy storage includes the following steps:

[0010] Constructing a power load forecasting model for a power system, the power load forecasting model includes a feature encoding part and a feature decoding part; the feature encoding part includes a multi-layer encoder, and the feature decoding part includes a multi-layer decoder;

[0011] Get sample data The sample data It is the time series data of electricity load data;

[0012] Where H is the number of input data and W is the dimension of input data;

[0013] In the feature encoding part, the time series data is encoded into output features;

[0014] The output features of the encoder at layer l

[0015] Where l∈{1,2,…,N} represents the lth layer of the encoder output, and N is the total number of encoder layers; Encoder(·) is defined as:

[0016]

[0018] In the formula, represents the i-th layer decomposition component in the l-th layer encoder, represents the output of the l-1th layer encoder, AFF(·) represents the adaptive frequency domain filtering module, and FADecomp(·) represents the frequency adaptive decomposition module;

[0019] The steps in the feature decoding part are:

[0020] In the formula, is the stationary component characteristic, is the seasonal component characteristic, is the trend component feature, l∈{1,2,…,M} represents the lth layer of the decoder output, M is the total number of decoder layers, and decoder(·) is

[0021]

[0022] In the formula, They represent the stationary component features, seasonal component features, and trend component features of the i-th layer in the l-th layer decoder, respectively. represents the output features of the frequency adaptive decomposition module of the i-th layer of the l-th layer decoder, represents the output component features and trend component features of the l-1 layer decoder, AFF att (·) represents the adaptive frequency domain normalization attention module,

[0023] W l,i ,i∈{1,2,3} represents the trend feature of the i-th layer The weight matrix of

[0024] The prediction result is obtained, which is the output component feature of the frequency adaptive decomposition module of the feature decoding part. and trend component characteristics The sum of is defined as:

[0025]

[0026] Among them, W X for The projection matrix of , Y is the output result of the power load forecasting model.

[0027] In the above technical solution, the present invention has the following beneficial effects:

[0028] By combining the adaptive frequency domain filtering module (AFF) for frequency domain filtering, it can not only effectively capture the global dependencies of the time series, but also adaptively extract components of different frequencies from the input data. First, the input data is converted from the time domain to the frequency domain through Fourier transform or other frequency conversion techniques. The purpose of this step is to decompose the time series into components of different frequencies, each of which represents the changes in different time scales in the input data. In the frequency domain, different frequency components have different importance for the prediction task. Adaptive frequency domain feature extraction can automatically identify and enhance those frequency components that are most relevant to the prediction task by dynamically adjusting the parameters of the filter. Through adaptive filtering, the model can selectively amplify or suppress certain frequency components to ensure that the model captures key features more accurately. Adaptive frequency domain feature extraction is not only a simple frequency selection, but also includes weighted processing of different frequency components. By assigning different weights to each frequency component, the model can aggregate useful information in the frequency domain to form a more predictive global feature representation. This weighted fusion ensures that the model can capture short-term changes without ignoring long-term trends.

[0029] The present invention uses the adaptive frequency domain filter module (AFF) to replace the self-attention mechanism in Transformer, which has higher computational efficiency. Since the computational complexity of the self-attention mechanism is O(n 2 ), when processing long sequences, each input needs to be calculated pairwise with all other positions, which will result in high computational costs, especially for long time series data. AFF reduces the computational complexity to a linear or near-linear complexity of O(nlog(n)) through frequency domain operations, greatly reducing the demand for computing resources. For large-scale power load forecasting tasks, the use of AFF can significantly improve computational efficiency. In addition, AFF can more flexibly separate short-term fluctuations (high-frequency parts) and long-term trends (low-frequency parts) by performing adaptive frequency filtering in the frequency domain, thereby providing the model with more refined feature decomposition and processing capabilities. This enables AFF to capture more accurate features in time series forecasting, especially when dealing with time series with complex periodicity and trends.

[0030] The frequency adaptive decomposition module can decompose the stationary features, seasonal features and trend features in the power load data. Power load data usually has obvious periodic fluctuations (such as day and night peaks and valleys, weekends and weekdays) and long-term trends (such as seasonal changes and annual growth in power demand). There may also be high-frequency fluctuations such as those caused by sudden events, weather fluctuations or economic activities. The frequency adaptive decomposition module can separate the stationary and non-stationary features, separate the unimportant high-frequency noise in the preprocessing, and retain the important periodic fluctuations and long-term trend information. After removing the stationary factors, these components are further decomposed into seasonal and trend items through the expert decomposition module, which can more accurately capture the long-term power demand trend and short-term fluctuation changes, and enable the model to better identify the potential growth trend and change pattern of future power load. In addition, power load data may show significant differences in different regions, different time periods and different seasons. The introduction of the frequency adaptive decomposition module allows the model to dynamically decompose features in different scenarios, which can improve the adaptability and generalization ability of the model to diversified load data.

[0031] Adaptive frequency domain normalized attention module. For medium- and long-term electricity load forecasting tasks, short-term fluctuations and noise in time domain data may affect the model's capture of global trends. By performing cross-attention calculations entirely in the frequency domain, the model focuses on the main frequency features and reduces sensitivity to time domain noise and short-term instability. Frequency-domain-based cross-attention allows the model to operate directly on frequency components without increasing complexity, eliminating the complexity of processing local dependencies in the time domain and focusing on long-term dependencies and global patterns of time series. In addition, frequency features usually directly reflect the periodic patterns of time series, such as daily and seasonal changes. Placing attention calculations in the frequency domain can more accurately identify and amplify these periodic features, making the model more adaptable and expressive when processing data with strong periodicity and trends (such as climate data and electricity load data). BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0033] Figure 1 A schematic diagram of the overall flow of the prediction method provided by the present invention;

[0034] Figure 2 The overall architecture diagram of the power load forecasting model provided by the present invention;

[0035] Figure 3A structural diagram of the adaptive frequency domain filtering module provided by the present invention;

[0036] Figure 4 A structural diagram of a frequency adaptive decomposition module provided by the present invention;

[0037] Figure 5 Structural diagram of the adaptive frequency domain normalized attention module provided by the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] Figure 1-5 The present invention provides a method for forecasting regional power load prospects taking into account multi-source distributed new energy storage in an embodiment, comprising the following steps:

[0041] Step 1: construct an electricity load forecasting model for the power system. The electricity load forecasting model includes a feature encoding part and a feature decoding part; the feature encoding part includes multiple layers of encoders, and the feature decoding part includes multiple layers of decoders.

[0042] The power system load prediction model FANet proposed in the present invention consists of two major parts: feature encoding (Encoder) and feature decoding (Decoder). Both parts contain multi-layer structures, namely, multi-layer encoders and decoders. Each layer of encoder mainly includes an adaptive frequency filtering module and a two-layer frequency adaptive decomposition module; each layer of decoder includes an adaptive frequency domain filtering module, a three-layer frequency adaptive decomposition module and a frequency adaptive enhanced attention module. The frequency adaptive enhanced attention module is located between the frequency adaptive decomposition module of one layer and the frequency adaptive decomposition modules of the other two layers.

[0043] Step 2: Get sample data Sample Data It is the time series data of electricity load data.

[0044] Among them, H is the number of input data and W is the dimension of the input data.

[0045] Step 3: Encode the time series data into output features in the feature encoding part.

[0046] The output features of the encoder at layer l

[0047] Where l∈{1,2,…,N} represents the lth layer of the encoder output, N is the total number of encoder layers, H and W are The dimension of; when l = 1, is the input of the encoder of the first layer, is the input time series data, H and W are Dimension.

[0048] Encoder(·) is defined as:

[0050] The output of the layer encoder, AFF(·) represents the adaptive frequency domain filtering module, and FADecomp(·) represents the frequency adaptive decomposition module; is the output feature of the encoder at layer l.

[0051] The steps in the feature decoding part are:

[0052] In the formula, is the stationary component characteristic, is the seasonal component characteristic, is the trend component feature, l∈{1,2,…,M} represents the lth layer of the decoder output, M is the total number of decoder layers, and decoder(·) is defined as follows:

[0054] characteristics, seasonal component characteristics, trend component characteristics, represents the output features of the frequency adaptive decomposition module of the i-th layer of the l-th layer decoder, Represents the output features and trend features of the l-1 layer decoder, AFF att (·) represents the adaptive frequency domain normalization attention module, W l,i ,i∈{1,2,3} represents the trend feature of the i-th layer The weight matrix of .

[0055] The prediction result is the output component feature of the frequency adaptive decomposition module of the feature decoding part and trend component characteristics The sum of is defined as:

[0056]

[0057] Among them, W X for The projection matrix of , Y is the output result of the power load forecasting model.

[0058] Specifically, the encoder includes an adaptive frequency domain filtering module and two layers of frequency adaptive decomposition modules with the same structure.

[0059] In order to achieve more lightweight and efficient global information capture, the present invention proposes a power load data feature extraction module that uses adaptive frequency filtering to fuse hybrid features. It captures features at different frequencies by performing frequency domain analysis on the sequence, and can handle long-term and short-term dependencies in historical load data.

[0060] The steps in the adaptive frequency domain filtering module include:

[0061] Step 1: Build input data The input data is the time series power load data, where H is the number of sample data and W is the dimension of the sample data.

[0062] Step 2: Perform a fast Fourier transform on the input data to convert the data from the time domain to the frequency domain, and express it as follows:

[0063]

[0064] In the formula, represents the frequency domain features after Fourier transformation, h, w represent the number and dimension index of input data, u, v represent the frequency domain features X after Fourier transformation F The number and dimension index of , i = h × w;

[0065] Step 3: Use a learnable frequency domain filter The frequency domain characteristics of the dot product input are expressed as follows:

[0066]

[0067] In the formula, It is a learnable frequency domain filter with the same shape as the frequency domain feature; It consists of a linear layer with a 1×1 convolutional layer, a ReLU activation function, and another 1×1 convolutional linear layer; X F 'For the learned frequency domain filter The result after dot multiplication of the input frequency domain features;

[0068] Step 4: Convert the filtered frequency domain features back to spatial domain features through inverse fast Fourier transform:

[0069]

[0070] is also equivalent to:

[0071]

[0072] Mathematically, it is equivalent to using large-kernel dynamic convolution as the weight of the feature mixer to obtain the output result of the adaptive frequency domain filtering module. is the spatial feature after fast Fourier transformation.

[0073] The present invention uses an adaptive frequency domain filtering module to perform frequency domain analysis to capture the global dependencies between different elements in the input time series power load data. The model can automatically adapt to the characteristics of different frequencies and can efficiently process signals or data with different periodicities, especially the smooth transition of long-term and short-term dependencies.

[0074] In order to more flexibly decompose the stationary and non-stationary information of the input features and retain the seasonal characteristics in the time series, the present invention proposes a feature decomposition module based on frequency adaptation. This module can dynamically process the complex frequency characteristics in the data, such as short-term fluctuations and abnormal behaviors, by adaptively normalizing the frequency domain, and by adaptively adjusting the frequency bandwidth, allowing the model to make the optimal decomposition in different scenarios. The steps in the frequency adaptive decomposition module include:

[0075] Step 1: Transform the spatial features after fast Fourier transformation The input of the adaptive frequency domain filter module is added as the input of the frequency adaptive decomposition module.

[0076] in,

[0077]

[0078] Step 2: Input to the first-layer frequency adaptive decomposition module Perform frequency residual learning (FRL): convert the input features into the frequency domain through discrete Fourier transform (DFT), and identify the K frequency components with the largest amplitude in each input feature and set them as non-stationary components

[0079]

[0080] In the formula, frequency residual learning is performed by one-dimensional discrete Fourier transform for each input feature X t Implemented and represented as DFT(·); TopK(·) selects the frequency set with the largest amplitude in the first K and calculates it using the Amp(·) function; Filter represents the frequency set from Z t Filter out K t Frequency operation, in order to reduce the impact of non-stationary signals, frequency residual learning uses IDFT (·) to restore the first K frequency components to non-stationary components

[0081] Step 3: Calculate seasonal components Calculating trend components From X t Remove Calculate the stationary component X res :

[0082]

[0083] Where F(·) is a set of average pooling filters, softmax(L(x non )) is the weight of the trend component feature, For the stable component, is the non-stationary component, For seasonal components, is the trend component;

[0084] Step 4: Use the prediction backbone model g with a normalization layer θ Get the stationary component characteristics after adaptive decomposition It can be expressed as:

[0085]

[0086] The normalization layer makes the prediction backbone model g θ Can focus on the stationary component of the input X res Dynamic changes in

[0087] Step 5: Use a feedforward neural network to analyze the seasonal component of the input Perform a nonlinear transformation:

[0088]

[0089] Where FF(·) represents a feedforward neural network, is the seasonal component feature after learning by the feedforward neural network;

[0090] Step 6: Trend component Directly recorded as trend feature

[0091] Step 7: Pass the stationary features of the prediction model and seasonal characteristics after nonlinear transformation Weighted fusion is performed to obtain the output of the first-layer frequency adaptive decomposition module:

[0092]

[0093] in, is the output of the first-layer frequency adaptive decomposition module in the encoder.

[0094] Step 8: The steps in the second-layer frequency adaptive decomposition module are the same as those in the first layer, including

[0095]

[0096] in, is the output of the second-layer frequency adaptive decomposition module in the encoder.

[0097] The first or second frequency adaptive decomposition module of the encoder extracts and separates the stationary and non-stationary features (such as seasonal and trend changes) in the input sequence through frequency domain decomposition. Through adaptive frequency selection and weighted fusion, the model can dynamically adjust the attention of different frequency features, and enhance the modeling ability of these features through feedforward neural networks and backbone prediction models, so that the model can better handle non-stationary and stationary time series and improve the prediction accuracy of complex time series power load data. The frequency adaptive decomposition module of the decoder works in a similar way to the encoder and will not be repeated here.

[0098] Preferably, the decoder comprises a serially connected adaptive frequency domain filtering module, a first-layer frequency adaptive decomposition module, an adaptive frequency domain normalized attention module, a second-layer frequency adaptive decomposition module and a third-layer frequency adaptive decomposition module;

[0099] The first-layer frequency adaptive decomposition module, the second-layer frequency adaptive decomposition module and the third-layer frequency adaptive decomposition module have the same structure as the frequency adaptive decomposition module of the encoder, and the adaptive frequency domain filtering module of the decoder has the same structure as the adaptive frequency domain filtering module of the encoder.

[0100] The present invention uses the same adaptive frequency domain filter module as the encoder to perform feature decomposition and processing.

[0101] In the adaptive frequency domain filtering module, the input data of the decoder is After performing frequency domain conversion and mapping the input data from the time domain to the frequency domain, the output is:

[0102]

[0103] In the formula, when l=1,

[0104] The decoder's adaptive frequency domain filtering is not only used to preserve global trends and seasonal characteristics, but also effectively captures local change details. Through weighted fusion of frequency components, the decoder can dynamically adjust its focus on each frequency feature, thereby ensuring that the generated sequence accurately captures global features and reflects local dynamic changes in time series data, which is crucial for complex time series prediction.

[0105] The frequency adaptive decomposition module applied to the encoder is extended to the decoder. In the first layer of the frequency adaptive decomposition module, the frequency adaptive decomposition mechanism is used to adjust the time series power load characteristics output by the adaptive frequency domain filter module. Perform layer-by-layer decomposition, where t represents the frequency adaptive filter module of the decoder. Separate different frequency features, including high-frequency stationary component features Seasonal component characteristics of the intermediate frequency and low-frequency trend component characteristics As shown in the formula:

[0106]

[0107] The stationary component characteristics and seasonal component characteristics Weighted fusion to obtain the input features of the adaptive frequency domain normalized attention module

[0108]

[0109] Among them, the stationary component characteristics Can capture short-term fluctuations and seasonal component characteristics Reflects seasonal characteristics and trend component characteristics Corresponding to long-term trend characteristics.

[0110] Similar to the processing in the encoder, the frequency adaptive decomposition module in the decoder can adaptively adjust the processing of these features to ensure that the balance of long-term and short-term features is taken into account during the decoding generation process. And this design enables the decoder to not only process long-term trends when generating sequences, but also dynamically capture short-term changes, thereby more flexibly responding to non-stationary time series. Finally, by effectively fusing stationary features with seasonal features, it is ensured that the generated sequence contains both short-term fluctuation information and can capture long-term trend changes, thereby improving the accuracy of power load forecasting.

[0111] The steps in the adaptive frequency domain normalization attention module are as follows: Step 1: Use the regular transformation expression to extract feature representations that are more suitable for frequency domain processing. The input is q comes from the decoder, k and v come from the encoder,

[0112]

[0113] In the formula, w q ,w k ,w v are the three learnable weights corresponding to q, k, and v, and q is the l-th layer feature of the decoder After w q The weighted result, k, v are the l-th layer features of the encoder respectively After w k ,w v Weighted results;

[0114] Step 2: Use Fourier transform to transform q, k, v to capture input features and The periodic patterns and the frequency information of long-term and short-term characteristics in are expressed as follows:

[0115]

[0116] Step 3: Obtain the output features of the adaptive frequency domain normalized attention module through learnable frequency domain filters and inverse Fourier transform

[0117]

[0118] In the formula, It consists of a linear layer followed by a 1×1 convolutional layer, a ReLU activation function, and another 1×1 convolutional linear layer.

[0119] Dynamically capture input features through attention mechanism in frequency domain and The model can adaptively adjust the model's attention to different frequency features to ensure effective modeling of global trends and local fluctuations, thereby improving prediction accuracy.

[0120] The steps in the second-layer frequency adaptive decomposition module are:

[0121] The steps in the second and third frequency adaptive decomposition modules are:

[0122] In the last two layers of the decoder’s frequency adaptive decomposition module, the input of the adaptive frequency domain normalized attention module is With output The weighted value is used as the input of the second-layer frequency adaptive decomposition module in the decoder and is recorded as It is expressed as follows:

[0123]

[0124] After the last two layers of the decoder frequency adaptive decomposition module, the output features are

[0125] Preferably, the method further includes optimizing the network model, the steps of which include:

[0126] To train the power load prediction model, the initialized model parameters are first set, and then the power load prediction model parameters are optimized through the loss function; in order to train a model that can accurately predict the power load, the present invention first sets the initialized model parameters, and then optimizes the model parameters through the loss function.

[0127] There are two optimization goals, one is the mean square loss It is used to measure the gap between the model's predicted value and the true value, expressed as follows:

[0128]

[0129] In the formula, is the output result of the power load forecasting model, y i ∈Y H×W is the real power load data, i={1,2,...,n}, representing the i-th sample;

[0130] Another optimization objective is the regularization loss Used to prevent model overfitting. By introducing an additional constraint term △ in the model's loss function, the complexity of the model is controlled, encouraging the model to learn simpler and smoother parameters, thereby improving the generalization ability of the model. The regularization loss is recorded as:

[0131]

[0132] Among them, the constraint λ is the regularization strength, ω is the regularization coefficient matrix;

[0133] In order to optimize the stationary features Seasonal characteristics In the learning process, we will guide the loss For model optimization, it is defined as:

[0134]

[0135] Prediction with prior guidance can be considered as a multi-task optimization problem, where Ensure θ Accurately predict stationary frequency characteristics Ensure that the feedforward neural network FF(·) accurately predicts seasonal characteristics Ensure that the power load forecasting model is optimized along the overall forecast accuracy. Seasonal characteristics The subscript t indicates that the encoder and decoder are common, which can be understood as t = {en, de}

[0136] By combining regularization loss and mean square error (MSE), the present invention achieves accurate prediction of power load. Mean square loss is used to measure the error between the model's predicted value and the actual value, helping the model to minimize the prediction deviation, especially for data with a long time span, it can effectively reduce the prediction error. By minimizing the square difference between the predicted result and the true value, the model can ensure accuracy in regression tasks. Regularization loss controls the complexity of the model by introducing additional constraints in the loss function, avoiding overfitting of the model to noise or specific data features during training. Regularization effectively improves the generalization ability of the model, ensuring that the model not only performs well on training data, but also maintains a high prediction accuracy on unseen test data.

[0137] The combination of these two loss functions ensures the accurate fitting of the model to the historical power load data on the one hand, and enhances the adaptability of the model to future changes through regularization, especially to the periodic fluctuations and sudden changes that may exist in the load data. The optimization strategy of the present invention enables the model to strike a balance between accuracy and stability when facing complex and nonlinear power load data, thereby improving the ability to predict future power loads and providing important support for the intelligent dispatching and energy planning of the power system.

[0138] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for forecasting regional power load prospects taking into account multi-source distributed new energy storage, characterized in that: The steps include: Constructing a power load forecasting model for a power system, the power load forecasting model includes a feature encoding part and a feature decoding part; the feature encoding part includes a multi-layer encoder, and the feature decoding part includes a multi-layer decoder; Get sample data The sample data It is the time series data of electricity load data; Where H is the number of input data and W is the dimension of input data; In the feature encoding part, the time series data is encoded into output features; The output features of the encoder at layer l Where l∈{1,2,…,N} represents the lth layer of the encoder output, and N is the total number of encoder layers; Encoder(·) is defined as: In the formula, represents the i-th layer decomposition component in the l-th layer encoder, represents the output of the l-1th layer encoder, AFF(·) represents the adaptive frequency domain filtering module, and FADecomp(·) represents the frequency adaptive decomposition module; The steps in the feature decoding part are: In the formula, is the stationary component characteristic, is the seasonal component characteristic, is the trend component feature, l∈{1,2,…,M} represents the lth layer of the decoder output, M is the total number of decoder layers, and decoder(·) is defined as: In the formula, They represent the stationary component features, seasonal component features, and trend component features of the i-th layer in the l-th layer decoder, respectively. represents the output features of the frequency adaptive decomposition module of the i-th layer of the l-th layer decoder, represents the output component features and trend component features of the l-1 layer decoder, AFF att (·) represents the adaptive frequency domain normalization attention module, W l,i ,i∈{1,2,3} represents the trend feature of the i-th layer The weight matrix of The prediction result is obtained, which is the output component feature of the frequency adaptive decomposition module of the feature decoding part. and trend component characteristics The sum of is defined as: Among them, W X for The projection matrix of , Y is the output result of the power load forecasting model.

2. The method for predicting regional power load prospects taking into account multi-source distributed new energy storage according to claim 1 is characterized in that: The encoder comprises in sequence an adaptive frequency domain filtering module and a frequency adaptive decomposition module with two layers of the same structure; The steps in the adaptive frequency domain filtering module include: Step 1: Build input data Step 2: Perform a fast Fourier transform on the input data to convert the data from the time domain to the frequency domain, and express it as follows: In the formula, represents the frequency domain features after Fourier transformation, h, w represent the number and dimension index of input data, u, v represent the frequency domain features X after Fourier transformation F The number and dimension index of , i = h × w; Step 3: Use a learnable frequency domain filter The frequency domain characteristics of the dot product input are expressed as follows: In the formula, It is a learnable frequency domain filter with the same shape as the frequency domain feature; It consists of a linear layer with a 1×1 convolutional layer, a ReLU activation function, and another 1×1 convolutional linear layer; X F 'For the learned frequency domain filter The result after dot multiplication of the input frequency domain features; Step 4: Convert the filtered frequency domain features back to spatial domain features through inverse fast Fourier transform: is also equivalent to: Mathematically, it is equivalent to using large-kernel dynamic convolution as the weight of the feature mixer to obtain the output result of the adaptive frequency domain filtering module. is the spatial feature after fast Fourier transformation; The steps in the frequency adaptive decomposition module include: Step 1: Transform the spatial features into The input of the adaptive frequency domain filter module is added as the input of the first layer frequency adaptive decomposition module. in, Step 2: Input to the first-layer frequency adaptive decomposition module Perform frequency residual learning (FRL): The input features are converted to the frequency domain through discrete Fourier transform (DFT), and the K frequency components with the largest amplitude in each input feature are identified and set as non-stationary components In the formula, frequency residual learning is performed by one-dimensional discrete Fourier transform for each input feature X t Implemented and represented as DFT(·); TopK(·) selects the frequency set with the largest amplitude in the first K and calculates it with the Amp(·) function; Filter represents the frequency set from Z t Filter out K t Frequency operation, in order to reduce the impact of non-stationary signals, frequency residual learning uses IDFT (·) to restore the first K frequency components to non-stationary components Step 3: Calculate seasonal components Calculating trend components From X t Remove Calculate the stationary component X res : Where F(·) is a set of average pooling filters, softmax(L(x non )) is the weight of the trend component feature, For the stable component, is the non-stationary component, For seasonal components, is the trend component; Step 4: Use the prediction backbone model g with a normalization layer θ Get the stationary component characteristics after adaptive decomposition It can be expressed as: The normalization layer makes the prediction backbone model g θ Can focus on the stationary component of the input X res Dynamic changes in Step 5: Use a feedforward neural network to analyze the seasonal component of the input Perform a nonlinear transformation: Where FF(·) represents a feedforward neural network, is the seasonal component feature after learning by the feedforward neural network; Step 6: Trend component Directly recorded as trend feature Step 7: Pass the stationary features of the prediction model and seasonal characteristics after nonlinear transformation Weighted fusion is performed to obtain the output of the first-layer frequency adaptive decomposition module: in, is the output of the first-layer frequency adaptive decomposition module in the encoder. Step 8: The steps in the second-layer frequency adaptive decomposition module are the same as those in the first layer, including: in, is the output of the second-layer frequency adaptive decomposition module in the encoder.

3. The method for predicting regional power load prospects taking into account multi-source distributed new energy storage according to claim 2 is characterized in that: The decoder sequentially comprises an adaptive frequency domain filtering module, a first-layer frequency adaptive decomposition module, an adaptive frequency domain normalized attention module, a second-layer frequency adaptive decomposition module and a third-layer frequency adaptive decomposition module; The first-layer frequency adaptive decomposition module, the second-layer frequency adaptive decomposition module, and the third-layer frequency adaptive decomposition module have the same structure as the frequency adaptive decomposition module of the encoder, and the adaptive frequency domain filtering module of the decoder has the same structure as the adaptive frequency domain filtering module of the encoder; In the adaptive frequency domain filtering module, the input data of the decoder is After performing frequency domain conversion and mapping the input data from the time domain to the frequency domain, the output is: In the first-layer frequency adaptive decomposition module, the frequency adaptive decomposition mechanism is used to adjust the time series power load characteristics output by the adaptive frequency domain filtering module. Perform decomposition operations to separate different frequency features, including high-frequency stationary component features Seasonal component characteristics of the intermediate frequency and low-frequency trend component characteristics As shown in the formula: The stationary component characteristics and seasonal component characteristics Weighted fusion to obtain the input features of the adaptive frequency domain normalized attention module The steps in the adaptive frequency domain normalization attention module are: Step 1: Use the regular transformation expression to extract the feature representation that is more suitable for frequency domain processing. The input is q comes from the decoder, k and v come from the encoder, In the formula, w q ,w k ,w v are the three learnable weights corresponding to q, k, and v, and q is the l-th layer feature of the decoder After w q The weighted result, k, v are the l-th layer features of the encoder respectively After w k ,w v Weighted results; Step 2: Use Fourier transform to transform q, k, v to capture input features and The periodic patterns and the frequency information of long-term and short-term characteristics in are expressed as follows: Step 3: Obtain the output features of the adaptive frequency domain normalized attention module through learnable frequency domain filters and inverse Fourier transform In the formula, It consists of a linear layer followed by a 1×1 convolutional layer, a ReLU activation function, and another 1×1 convolutional linear layer. The steps in the second and third frequency adaptive decomposition modules are: In the last two layers of the decoder’s frequency adaptive decomposition module, the input of the adaptive frequency domain normalized attention module is With output The weighted value is used as the input of the second-layer frequency adaptive decomposition module in the decoder and is recorded as It is expressed as follows: After the last two layers of the decoder frequency adaptive decomposition module, the output features are 4. The method for predicting regional power load prospects taking into account multi-source distributed new energy storage according to claim 1 is characterized in that: It also includes the optimization of the network model, the steps include: To train the power load forecasting model, the model parameters are initialized and then optimized through the loss function. There are two optimization goals, one is the mean square loss It is used to measure the gap between the model's predicted value and the true value, expressed as follows: In the formula, is the output result of the power load forecasting model, y i ∈Y H×W is the real power load data, i={1,2,...,n}; Another optimization objective is the regularization loss It is used to prevent the model from overfitting; by introducing an additional constraint term △ into the loss function of the power load forecasting model to control the complexity of the model, the regularization loss is recorded as: Among them, the constraint λ is the regularization strength, ω is the regularization coefficient matrix; In order to optimize the stationary features Seasonal characteristics In the learning process, we will guide the loss For model optimization, it is defined as: Prediction with prior guidance can be considered as a multi-task optimization problem, where Ensure θ Accurately predict stationary frequency characteristics Ensure that the feedforward neural network FF(·) accurately predicts seasonal characteristics Ensure that the power load forecasting model is optimized along the overall forecasting accuracy.

Images