CN-121983951-A - Quantum hybrid network-based power load prediction method, system, equipment and medium

CN121983951ACN 121983951 ACN121983951 ACN 121983951ACN-121983951-A

Abstract

The invention is suitable for the field of power load prediction, and discloses a power load prediction method, a system, equipment and a medium based on a quantum hybrid network, wherein the method comprises the steps of acquiring original load time sequence data and decomposing the original load time sequence data into seasonal components and trending components; the method comprises the steps of constructing a hybrid prediction model, wherein the model comprises parallel optimization Transformer branches and quantum hybrid network branches, predicting seasonal components through the optimization Transformer branches to obtain seasonal prediction results, the optimization Transformer branches comprise a dimension-unchanged embedded DI and sparse attention transfomer mechanism, predicting trend components through the quantum hybrid network branches to obtain trend prediction results, the quantum hybrid network branches comprise parallel variable component sub-circuits VQC and multi-layer perceptron MLP, and fusing the seasonal prediction results with the trend prediction results to obtain electric load prediction values. The invention adopts double-branch processing and fusion, thereby realizing the improvement of the accuracy and stability of power load prediction while reducing the consumption of computing resources.

Inventors

Deng Yuedan
GAO ZHENGHAO
XU KUI
OU JIAXIANG
XIAO YANHONG
ZHENG YUANWEI

Assignees

贵州电网有限责任公司

Dates

Publication Date: 20260505
Application Date: 20251218

Claims (10)

1. A method for predicting power load based on a quantum hybrid network, comprising: acquiring original load time sequence data and decomposing the data into seasonal components and trending components; constructing a hybrid prediction model, wherein the model comprises parallel optimization transducer branches and quantum hybrid network branches; predicting seasonal components by optimizing a transducer branch to obtain seasonal prediction results, wherein the optimized transducer branch comprises a dimension-unchanged embedded DI and sparse attention transfomer mechanism; Predicting a trend component through a quantum hybrid network branch to obtain a trend prediction result, wherein the quantum hybrid network branch comprises a parallel variable component sub-circuit VQC and a multi-layer perceptron MLP; And fusing the seasonal prediction result and the trend prediction result to obtain a power load prediction value.
2. The quantum hybrid network-based power load prediction method of claim 1, wherein predicting the seasonal component by optimizing the transducer branch to obtain a seasonal prediction result comprises: Processing seasonal components through dimension-invariant embedding DI to obtain embedded features; Inputting the embedded features into an encoder and a decoder based on a sparse attention transducer mechanism to obtain output features of the decoder; the output characteristics of the decoder are input into a convolutional neural network, and seasonal prediction results are generated through dimension transformation.
3. The quantum hybrid network-based power load prediction method of claim 2, wherein the predicting the trending component through the quantum hybrid network branch to obtain the trending prediction result comprises: Preliminary feature extraction is carried out on the trend component through a linear layer; based on the result of the preliminary feature extraction, carrying out normalization processing to obtain multi-feature representation; Based on the multi-feature representation, the multi-feature representation is respectively output to a variable component sub-circuit VQC and a multi-layer sensing neural network MLP which are parallel to each other, so as to obtain the output features of the variable component sub-circuit VQC and the output features of the multi-layer sensing neural network MLP; And carrying out weighted combination on the output characteristics of the VQC and the MLP through a trainable linear weight layer to obtain a trend prediction result.
4. The quantum hybrid network-based power load prediction method of claim 2, wherein the embedding DI processes seasonal components through dimension invariance resulting in embedded features comprising: carrying out convolution operation on seasonal components by using a convolution neural network CNN, and mapping the seasonal components into a multichannel characteristic map; dividing the multi-channel feature map into a plurality of non-overlapping blocks along the time step dimension by performing block division operation to obtain block division input; and if the lengths of the seasonal components cannot be divided completely, zero filling is carried out, and the embedded features after the blocks are obtained.
5. A method of quantum hybrid network based power load prediction as claimed in claim 3, wherein the variable component sub-circuit VQC is constructed from a feature encoding layer, a trainable parameter and entanglement layer and a measurement layer comprising: Performing unitary transformation coding on the multi-feature representation through a feature coding layer; evolution of the encoded quantum state is carried out by adopting a plurality of layers of trainable parameters and entanglement layers; and measuring the evolved quantum state through a measuring layer to obtain the output characteristics of the variable component sub-circuit VQC.
6. The quantum hybrid network-based power load prediction method of claim 5, wherein the multi-layer sensory neural network MLP comprises: multiplying the multi-feature representation by a weight matrix and delivering to a hidden neuron layer; and applying bias to the value of the hidden neuron layer, and scaling by adopting an activation function to obtain the output characteristics of the multi-layer perception neural network MLP.
7. The quantum hybrid network-based power load prediction method of claim 6, wherein the weighting combination of the output characteristics of the VQC and the MLP through the trainable linear weight layer to obtain the trend prediction result comprises: Combining the output characteristics of the variable component sub-circuit VQC with the output characteristics of the multi-layer perception neural network MLP, adopting a two-to-one linear weight layer and utilizing a trainable weighting parameter to carry out weighted summation, and obtaining a trend prediction result.
8. A quantum hybrid network based electrical load prediction system employing the method of any one of claims 1-7, comprising: The decomposition module is used for acquiring the original load time sequence data and decomposing the original load time sequence data into seasonal components and trending components; The model construction module is used for constructing a hybrid prediction model, and the model comprises parallel optimization transducer branches and quantum hybrid network branches; the first prediction module is used for predicting seasonal components through an optimized converter branch to obtain seasonal prediction results, wherein the optimized converter branch comprises a dimension-unchanged embedded DI and sparse attention transfomer mechanism; The second prediction module is used for predicting a trend component through a quantum hybrid network branch to obtain a trend prediction result, wherein the quantum hybrid network branch comprises a parallel variable component sub-circuit VQC and a multi-layer perceptron MLP; And the fusion module is used for fusing the seasonal prediction result and the trend prediction result to obtain a power load prediction value.
9. An electronic device, comprising: A memory and a processor; The memory is for storing computer executable instructions for execution by the processor to implement the steps of the quantum hybrid network-based power load prediction method of any one of claims 1 to 7.
10. A computer readable storage medium, comprising computer executable instructions stored thereon, which when executed by a processor, implement the steps of the quantum hybrid network based power load prediction method of any of claims 1 to 7.

Description

Quantum hybrid network-based power load prediction method, system, equipment and medium Technical Field The invention relates to the field of power load prediction, in particular to a power load prediction method, a system, equipment and a medium based on a quantum hybrid network. Background The power load prediction is a core link of power system planning and operation management, and directly influences the running stability of a power grid and the balance of power supply and demand. Therefore, efficient prediction of power load is an important basis for ensuring operation of the power system. However, in actual situations, the change rule of the power load data is very complex, and the power load data has the characteristics of obvious seasonal fluctuation and long-term growth trend. However, in the traditional prediction method, it is often difficult to accurately capture the relationship between the long-term trend and the seasonal fluctuation and the relationship between the short-term fluctuation and the long-term electricity consumption scale. These two core features cannot meet the requirements of the existing energy fine management. Therefore, on the premise of ensuring the prediction precision, the prediction efficiency and stability are required to be further improved. Based on the requirement, the seasonal fluctuation feature and the long-term trend feature in the load data are separately processed, and a more appropriate technical means is selected for modeling aiming at the characteristics of each feature, so that the prediction accuracy is integrally improved. By means of the strategy, the prediction performance is remarkably improved, the use efficiency of computing resources is optimized, unnecessary computing power consumption is avoided, the running cost of the whole prediction system is finally reduced, and fine management and efficient scheduling of the power system are better supported. Disclosure of Invention The present invention has been made in view of the above-mentioned problems of limited accuracy, prediction efficiency and stability existing in the prior art. Therefore, the invention provides a quantum hybrid network-based power load prediction method, a quantum hybrid network-based power load prediction system, quantum hybrid network-based power load prediction equipment and a quantum hybrid network-based power load prediction medium, which solve the problems that when the conventional power load prediction technology processes multivariable time series data, complex seasonal period fluctuation and long-term trend evolution are difficult to be simultaneously considered, and simultaneously solve the problems that the computation complexity of the conventional transducer full-attention mechanism is increased in a quadratic manner when the conventional transducer full-attention mechanism processes long-term series data, so that the consumption of computing resources is huge, the reasoning speed is low, and the real-time scheduling requirement of a power grid cannot be met. In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the present invention provides a method for predicting power load based on a quantum hybrid network, comprising: acquiring original load time sequence data and decomposing the data into seasonal components and trending components; constructing a hybrid prediction model, wherein the model comprises parallel optimization transducer branches and quantum hybrid network branches; predicting seasonal components by optimizing a transducer branch to obtain seasonal prediction results, wherein the optimized transducer branch comprises a dimension-unchanged embedded DI and sparse attention transfomer mechanism; Predicting a trend component through a quantum hybrid network branch to obtain a trend prediction result, wherein the quantum hybrid network branch comprises a parallel variable component sub-circuit VQC and a multi-layer perceptron MLP; And fusing the seasonal prediction result and the trend prediction result to obtain a power load prediction value. The invention relates to a power load prediction method based on a quantum hybrid network, which comprises the following steps of: Processing seasonal components through dimension-invariant embedding DI to obtain embedded features; Inputting the embedded features into an encoder and a decoder based on a sparse attention transducer mechanism to obtain output features of the decoder; the output characteristics of the decoder are input into a convolutional neural network, and seasonal prediction results are generated through dimension transformation. The invention relates to a quantum hybrid network-based power load prediction method, which comprises the following steps of: Preliminary feature extraction is carried out on the trend component through a linear layer; based on the result of the preliminary feature extraction, carrying out normalization proces