US-12627145-B2 - Multi-energy integrated short-term load forecasting method and system

US12627145B2US 12627145 B2US12627145 B2US 12627145B2US-12627145-B2

Abstract

The disclosure provides a multi-energy integrated short-term load forecasting method and system, which relates to the technical field of load forecasting. In the disclosure, after classifying the acquired relevant data of multi-energy integrated short-term load forecasting, the data after sample classification is used to train the multi-energy integrated short-term load forecasting model. The model is composed of multiple layers of temporal convolutional networks having multi-head self-attention mechanism and rotary position embedding. Finally, the trained model is used to carry out the multi-energy integrated short-term load forecasting. The disclosure can fully mine the coupling feature between multi-energy loads, improve the accuracy of multi-energy integrated short-term load forecasting, and further improve the management level and service efficiency of integrated energy demand side.

Inventors

Kaile ZHOU
Rong Hu

Assignees

HEFEI UNIVERSITY OF TECHNOLOGY

Dates

Publication Date: 20260512
Application Date: 20230427
Priority Date: 20221030

Claims (4)

1 . An electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the programs may be configured to execute methods, comprising: performing pretreatment and correlation calculation on acquired relevant data of multi-energy integrated short-term load forecasting to determine input data of a multi-energy integrated short-term load forecasting model, and performing sample classification on the input data; wherein the relevant data of multi-energy integrated short-term load forecasting comprises multi-energy load historical data and external environment historical data corresponding to the multi-energy load historical data; acquiring a multi-energy integrated short-term load forecasting model constructed based on encoder-decoder, wherein the encoder comprises a plurality of layers of temporal convolutional networks having multi-head self-attention mechanism and rotary position embedding; and training the multi-energy integrated short-term load forecasting model based on the input data after the sample classification, and performing the multi-energy integrated short-term load forecasting by using the trained multi-energy integrated short-term load forecasting model; wherein the step of acquiring a multi-energy integrated short-term load forecasting model constructed based on encoder-decoder, wherein the encoder comprises a plurality of layers of temporal convolutional networks having multi-head self-attention mechanism and rotary position embedding comprises: S 21 . converting the input data of multi-energy integrated short-term load forecasting into a feature matrix; S 22 . constructing the encoder by a temporal convolutional network having self-attention mechanism and rotary position embedding, and based on the encoder, mining a coupling feature of the input data of multi-energy integrated short-term load forecasting; S 23 . eliminating a redundant feature in the coupling feature based on a distillation operation; S 24 . constructing a decoder, wherein the decoder is arranged for obtaining an output data of multi-energy integrated short-term load forecasting after the coupling feature from which the redundant feature has been eliminated is decoded by the decoder.
2 . The electronic device of claim 1 , wherein the step of performing pretreatment and correlation calculation on acquired data related to multi-energy integrated short-term load forecasting to determine input data of a multi-energy integrated short-term load forecasting model, and performing sample classification on the input data comprises: performing pretreatment on the multi-energy load historical data and the external environment historical data by a pretreating operation comprising a missing value filling and normalization processing; calculating a correlation coefficient of the described pretreated multi-energy load historical data sequence to determine the input data of the multi-energy integrated short-term load forecasting model; and dividing the pretreated and correlation-calculated input data into a training set, a validation set, and a test set.
3 . The electronic device of claim 2 , wherein the step of calculating a correlation coefficient of a sequence of the described pretreated multi-energy load historical data to determine the input data of the multi-energy integrated short-term load forecasting model comprises: performing correlation evaluation through Pearson correlation coefficient, comparing a correlation inside the multi-energy load historical data sequence and a correlation between the multi-energy load historical data and the external environment historical data, if the correlation inside the multi-energy load historical data sequence is greater than the correlation between the multi-energy load historical data and the external environment historical data, using the multi-energy load historical data sequence as an input only; on the contrary, if the correlation inside the multi-energy load historical data sequence is greater than 0.2 but less than the correlation between the multi-energy load historical data and the external environment historical data, then using the multi-energy load historical data sequence and the external environment historical data sequence simultaneously as inputs of the multi-energy integrated short-term load forecasting.
4 . The electronic device of claim 1 , wherein in S 22 , the step of constructing the encoder by a temporal convolutional network having self-attention mechanism and rotary position embedding, and based on the encoder, mining a coupling feature of the input data of multi-energy integrated short-term load forecasting comprises: S 221 . constructing an attention matrix based on the feature matrix, wherein the attention matrix comprises query matrix, key matrix and value matrix; S 222 . adding rotary position embedding to the query matrix and the key matrix; S 223 . calculating an attention score between the query matrix and the key matrix after adding rotary position embedding; S 224 . updating the query matrix according to a sparsity measure of the query matrix to obtain an attention query matrix; S 225 . executing an attention mechanism on the key matrix, the value matrix and the updated attention query matrix to mine the coupling feature of the input data of multi-energy integrated short-term load forecasting.

Description

CROSS REFERENCE TO RELATED APPLICATION This patent application claims the benefit and priority of Chinese Patent Application No. CN 202211341149.7 filed on Oct. 30, 2022, the disclosure of which is incorporated by reference herein in its entirety as part of the present application. TECHNICAL FIELD The present disclosure relates to the technical field of load forecasting, and more specifically, to a multi-energy integrated short-term load forecasting method and system. BACKGROUND ART In order to meet the demand of green and low carbon, integrated energy systems (IESs) have gradually replaced the traditional independent planning and operation of single energy systems. At present, IESs are no longer limited to a single type of load, and it is needed to comprehensively consider multiple energy systems and develop multi-energy load forecasting methods. In IESs, all kinds of energy supply each other through complex coupling mechanism, which makes the factors affecting the demand of all kinds of loads in IESs more complex, and the balance between supply and demand of various kinds of energy becomes more difficult. The load forecasting of IESs is an important support for the operation management and optimal scheduling of IESs. It is of great significance and practical application value to carry out accurate multi-energy load forecasting of IESs. At present, the load forecasting methods of IESs are mostly single-energy load forecasting. However, with the continuous development of the energy industry, the energy demand is affected by many factors (such as the actual environmental factors on the energy consumption side). Because the single-energy load forecasting method cannot fully mine the coupling feature between different loads, it cannot be directly applied to IESs multi-energy load forecasting. However, the existing multi-energy load short-term forecasting methods are mainly aimed at the load data itself, and lack of comprehensive consideration of the actual environmental factors on the energy consumption side, which cannot achieve accurate forecasting. In addition, the existing IESs load forecasting methods generally optimize the model parameters and structure, and still have some limitations in dealing with different forms of energy coupling problems. Because the coupling feature between different loads are ignored, these methods cannot be directly applied to multi-energy integrated short-term load forecasting. At the same time, due to the different sequence length between short-term load forecasting and ultra-short-term load forecasting, the current relatively mature ultra-short-term load forecasting methods (such as recurrent neural network, and long short-term memory) cannot be directly applied to short-term forecasting. It can be seen that the existing technology cannot directly forecast the multi-energy integrated load in the short term, let alone accurately forecast. SUMMARY (1) Technical Problems to be Solved In view of the shortcomings of the existing technology, the disclosure provides a multi-energy integrated short-term load forecasting method and system, which solves the problem that the existing technology cannot carry out short-term accurate forecasting on multi-energy integrated load. (2) Technical Solutions In order to achieve the above purpose, the following technical solutions of the present disclosure are adopted. In a first aspect, the disclosure first provides a multi-energy integrated short-term load forecasting method, including: performing pretreatment and correlation calculation on acquired relevant data of multi-energy integrated short-term load forecasting to determine input data of a multi-energy integrated short-term load forecasting model, and performing sample classification on the input data; wherein the relevant data of multi-energy integrated short-term load forecasting includes multi-energy load historical data and external environment historical data corresponding to the multi-energy load historical data;acquiring a multi-energy integrated short-term load forecasting model constructed based on encoder-decoder, wherein the encoder includes multiple layers of temporal convolutional networks having multi-head self-attention mechanism and rotary position embedding; andtraining the multi-energy integrated short-term load forecasting model based on the input data after the sample classification, and performing the multi-energy integrated short-term load forecasting by using the trained multi-energy integrated short-term load forecasting model. Preferably, the step of performing pretreatment and correlation calculation on acquired data related to multi-energy integrated short-term load forecasting to determine input data of a multi-energy integrated short-term load forecasting model, and performing sample classification on the input data includes: performing pretreatment on the multi-energy load historical data and the external environment historical data by a pretreating operation