CN-122020561-A - Power grid multisource data fusion investment benefit post-evaluation platform based on quantum mixing algorithm

CN122020561ACN 122020561 ACN122020561 ACN 122020561ACN-122020561-A

Abstract

The application provides a power grid multisource data fusion investment benefit post-evaluation platform based on a quantum hybrid algorithm, belongs to the technical field of power grid investment evaluation, overcomes the core defects of 'simple combination and lack of depth improvement' existing in combination of quantum and classical algorithms in the prior art, adopts a full-flow closed-loop architecture of 'quantum enhancement hybrid algorithm driving and three-dimensional data depth fusion', sequentially divides the whole architecture into five core levels of a quantized data preprocessing layer, a quantum-space diagram fusion layer, a quantum-reinforcement learning evaluation optimization layer, a quantum causal feedback layer and a visual output layer from bottom to top, and designs a quantum enhancement hybrid algorithm system with multiple algorithm cooperation aiming at the actual requirements of fusion evaluation of three types of core data of power grid operation, equipment state and financial investment, and the combination mode of quantum and classical algorithms is reconstructed from bottom logic, so that the power distribution network can be perfectly suitable for post-investment benefit evaluation.

Inventors

LIN XIA
AN CHENGWAN
Li Haogui
BIAN XIAOJUN
ZHAO XIA
SONG LIGONG
ZHANG YIZHE
ZHANG RONGXIAN
ZHANG JIANYIN

Assignees

国网山西省电力有限公司朔州供电分公司

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (10)

1. The utility model provides an evaluation platform behind electric wire netting multisource data fusion investment benefit based on quantum hybrid algorithm which characterized in that includes: The quantized data preprocessing layer is used for introducing a quantum generation countermeasure network Q-GAN to realize quantum-principal component countermeasure dimension reduction on the basis of quantum principal component analysis QPCA dimension reduction and outputting robust dimension reduction data subjected to anti-interference treatment; The quantum-space diagram fusion layer is used for carrying out quantum entanglement on the multi-source data of the power grid by constructing a quantum knowledge graph GQ-KG of the power grid, realizing semantic association and knowledge fusion of the multi-source data of the power grid, reconstructing a convolution kernel of a space diagram neural network ST-GNN by introducing quantum diagram convolution operation to obtain a quantum enhancement space diagram neural network QE-ST-GNN, and converting quantum characteristics of the quantum knowledge graph GQ-KG of the power grid into classical fusion characteristics by the quantum enhancement space diagram neural network QE-ST-GNN; The quantum-reinforcement learning evaluation optimization layer builds an evaluation-feedback-optimization dynamic closed loop of the power grid investment benefit through a quantum reinforcement learning model QE-DRL, and realizes the efficient solution of multi-objective optimization and the dynamic adjustment of an evaluation strategy; the quantum causal traceability feedback layer is used for carrying out deep fusion on quantum correlation analysis, causal inference and antagonism fusion technology to obtain a quantum causal antagonism fusion Q-CAF method, so as to realize accurate positioning and quantification attribution of evaluation abnormality; And the visual output layer is used for realizing classical conversion and multidimensional visual display of the quantum evaluation result.
2. The quantum mixing algorithm-based power grid multisource data fusion investment benefit post-evaluation platform as claimed in claim 1 is characterized in that the implementation steps of the quantized data preprocessing layer are as follows: data integration, namely carrying out standardized integration on three types of core data, namely power grid operation data, equipment state data and financial data, and constructing and forming a three-dimensional data matrix Wherein R is a real number domain, and n represents the total sample amount/data record number of the multi-source data of the power grid; Quantum coding, namely mapping the integrated three-dimensional data into quantum states by adopting an angle/phase coding mode: the angle parameters of the quantum states are jointly determined by the normalized values of the three-dimensional data, so that the efficient conversion from classical data to the quantum states is realized; antagonistic dimension reduction by extraction of principal component quantum states from quantum states by quantum principal component analysis QPCA And inputting the main component quantum state into a quantum generation countermeasure network Q-GAN to perform characteristic defense training, and finally outputting robust dimension reduction data subjected to anti-interference processing.
3. The grid multisource data fusion investment benefit post-evaluation platform based on the quantum mixing algorithm is characterized in that a quantum generation countermeasure network Q-GAN comprises a generator and a discriminator, the generator simulates distribution characteristics of grid data through characteristics of quantum superposition states, a grid data sample with strong anti-interference capability is generated, the discriminator is constructed by a quantum neural network QNN, and real grid data and noise-containing interference data are accurately distinguished by utilizing parallel advantages of quantum calculation.
4. The quantum mixing algorithm-based power grid multisource data fusion investment benefit post-evaluation platform disclosed by claim 1 is characterized in that the construction process of a power grid quantum knowledge graph GQ-KG is as follows: Firstly, mapping power grid operation data, equipment state data and financial data into three core nodes of a power grid quantum knowledge graph GQ-KG respectively, wherein the power grid operation data corresponds to a power grid topological node, the equipment state data corresponds to an electric equipment node and the financial data corresponds to a financial node; And then, establishing attribute association among different types of nodes through the strong association characteristic of quantum entanglement, so as to realize semantic association and knowledge fusion of multi-source data.
5. The quantum-time space diagram fusion post-investment benefit evaluation platform based on the quantum mixing algorithm is characterized by comprising the following implementation steps of: The space dimension feature extraction, namely constructing a quantum computing model through a quantum gate circuit, computing quantum entanglement strength between any two nodes in a power grid quantum knowledge graph GQ-KG, taking the quantum entanglement strength as space association weight between the nodes, and realizing accurate quantification of power grid data topology association features; the time dimension feature fusion is that a quantum phase encoding mode is adopted to map the time sequence feature of the power grid data into a quantum state phase parameter, and then the time dimension phase feature and the space dimension entanglement weight feature are subjected to depth fusion through quantum inverse Fourier transform IQFT, so that the integrated extraction of the time-space feature is realized; And the fusion characteristic output is that the quantum state of the power grid quantum knowledge graph is subjected to evolution calculation through the throughput sub-convolution unitary transformation, and the evolved quantum state is converted into classical fusion characteristic data through quantum measurement operation, so that the efficient conversion from quantum characteristics to classical characteristics is realized.
6. The grid multisource data fusion investment benefit post-evaluation platform based on the quantum mixing algorithm according to claim 1 is characterized in that the implementation steps of the quantum-reinforcement learning evaluation optimization layer are as follows: the quantized state space construction, namely carrying out joint quantum coding on the fusion characteristics output by the quantum-space diagram fusion layer and the multi-objective evaluation indexes of the power grid investment benefit, converting the fusion characteristics into a quantum superposition state, and constructing and forming a high-dimensional quantum state space: ; Wherein, the Is a high-dimensional quantum state space, For the quantum amplitude of the light, For evaluating state basis vectors, k is an evaluation state basis vector index of the high-dimensional quantum state space, and m is the total number of evaluation state basis vectors in the high-dimensional quantum state space; and (3) optimizing the quantum reinforcement strategy, namely reconstructing a deep reinforcement learning strategy network by adopting a quantum approximation optimization algorithm QAOA to obtain a quantum reinforcement learning model QE-DRL, and realizing the rapid search and optimization of the optimal evaluation strategy by utilizing the advantage of quantum parallel computing.
7. The grid multisource data fusion investment benefit post-evaluation platform based on the quantum mixing algorithm according to claim 6 is characterized in that the implementation steps of quantum strategy optimization are as follows: Constructing a weight-based linear weighted reward function, taking financial yield, equipment availability and carbon emission reduction as core evaluation indexes of economic, technical and social targets respectively, and determining the weight of each evaluation index through a quantum game balancing algorithm to realize scientific distribution of the multiple target weights; The quantum enhancement strategy updating is carried out by carrying out iterative optimization on the reconstructed strategy network parameters through a quantum variation algorithm, utilizing the advantages of quantum parallel computation, searching a plurality of strategy directions simultaneously, greatly improving the efficiency of strategy optimization and realizing the collaborative optimization of multiple targets of economy, technology and society; And the dynamic feedback optimization is to build a quantum-classical data interface, feed back a real-time evaluation result to a strategy network through the data interface, access power grid real-time operation data, and automatically and dynamically adjust evaluation parameters and strategies by the strategy network according to the evaluation result and the power grid real-time operation data to form a complete set of 'evaluation-feedback-optimization' dynamic closed loop.
8. The grid multisource data fusion investment benefit post-evaluation platform based on the quantum mixing algorithm according to claim 5 is characterized in that the implementation steps of the quantum-causal traceability feedback layer are as follows: The quantum causal graph construction, wherein a key causal link which has obvious influence on an investment benefit evaluation result is screened out by calculating quantum entanglement intensity among nodes based on a power grid quantum knowledge graph GQ-KG, a quantum causal graph of data characteristic-evaluation index-benefit result is constructed based on the key causal links, and the nodes of the quantum causal graph are quantized power grid data characteristic and evaluation index, and the sides are quantum causal association intensity among the nodes; Quantum enhancement causal inference, namely reconstructing causal inference logic by adopting a quantum Bayesian network QBN, accelerating the searching and analyzing process of causal links by utilizing the advantages of quantum parallel computation, and realizing accurate positioning and quantification attribution of evaluation anomalies And (3) optimizing feedback output, namely generating a targeted investment optimization suggestion according to an abnormal positioning result and quantized attribution data obtained by quantum causal inference and combining with the actual running condition of the power grid, pushing the optimization suggestion to a power grid investment decision-making department, and completing the whole evaluation-tracing-optimizing closed loop.
9. The power grid multisource data fusion investment benefit post-evaluation platform based on the quantum mixing algorithm according to claim 8, wherein the implementation steps of quantum enhancement causal inference are as follows: When the investment benefit evaluation result deviates from a preset threshold value, starting a quantum back propagation algorithm, starting from an evaluation result abnormal node, carrying out back tracing along a causal link of a quantum causal graph, and accurately positioning a core cause node which leads to evaluation abnormality to form a complete abnormality causal path; and the quantization attribution is that the inner product modulus square of each causal node quantum state and abnormal quantum state is calculated through quantum measurement operation, and is used as the contribution degree of each node to the evaluation of the abnormality, so that the accurate quantization of the abnormality cause is realized.
10. The quantum mixing algorithm-based power grid multisource data fusion investment benefit post-evaluation platform as claimed in claim 1, wherein the visual output layer can be further adapted to an interface of a power grid system.

Description

Power grid multisource data fusion investment benefit post-evaluation platform based on quantum mixing algorithm Technical Field The application relates to the technical field of power grid investment assessment, in particular to a power grid multisource data fusion investment benefit post-assessment platform based on a quantum mixing algorithm. Background The current evaluation work after the investment benefit of the power grid mainly relies on a classical single algorithm to process multi-source heterogeneous data, has obvious bottleneck in data processing calculation force and evaluation result precision, and is difficult to meet the evaluation requirement of the mass multi-source data of the power grid. Although intelligent algorithms such as space-time diagram neural networks and reinforcement learning have been developed in the field of multi-source data fusion, advantages are shown in feature extraction and data fusion, the performance upper limit of a classical algorithm is not broken through by combining a quantum computing technology, the quantum computing technology realizes preliminary application of a hybrid algorithm in the field of energy finance subdivision such as power dispatching optimization and finance combination optimization, and the feasibility of quantum computing in the field is verified, but a complete technical scheme of deep coupling of 'quantum computing and multi-class intelligent algorithms' and full-dimension assessment of investment benefits of an adaptive distribution network has not been formed. In particular, the prior art has the following drawbacks: 1. The hybrid algorithm lacks depth fusion, wherein the combination of quanta and classical algorithms in the prior art only stays in a surface layer application mode of quantum coding and traditional algorithm, the algorithm is not reconstructed and optimized from a core logic layer, the natural advantages of quantum calculation in parallel operation and high-dimensional space processing can not be fully exerted, and the practical effect of algorithm fusion is limited. 2. The space-time correlation capturing is insufficient, wherein the power grid operation data has obvious space-time distribution characteristics, the equipment state data has complex topological correlation characteristics, the financial data presents obvious time sequence change characteristics, the multi-dimensional correlation of three types of data is extremely complex, the traditional classical algorithm is difficult to accurately model and effectively capture the complex multi-dimensional correlation, fusion characteristic distortion is easy to cause, and the accuracy of a subsequent evaluation result is influenced. 3. The dynamic optimization capability is weak, most of the existing power grid investment benefit post-evaluation technologies only can finish static output of evaluation results, lack of dynamic feedback and optimization adjustment mechanisms based on power grid real-time operation data, cannot adapt to dynamic changes of power grid operation states and investment benefits in real time, have insufficient timeliness and practicality of the evaluation results, and are difficult to form dynamic guidance on investment decisions. 4. The robustness and the interpretability are unbalanced, namely, when the robustness of the existing data fusion algorithm is insufficient and evaluation result deviation is easy to occur in the face of external interference factors such as noise, new energy output fluctuation and the like generated in the power grid data acquisition process, meanwhile, the quantum algorithm has certain black box characteristics, the combined hybrid algorithm further aggravates the difficult problem of the interpretability of the evaluation result, and the forming reason and influence factors of the evaluation result are difficult to be clearly evaluated. 5. The multi-objective optimization efficiency is low, the power grid investment benefit evaluation relates to a plurality of core objectives such as economic benefit, reliable technology, social environment protection and the like, is a typical multi-objective optimization problem, is extremely easy to fall into a local optimal solution when the traditional quantum algorithm solves the multi-objective optimization problem, is difficult to obtain a global optimal solution, has high calculation complexity, takes too long solving time and cannot meet the high-efficiency evaluation requirement of actual engineering. Disclosure of Invention In order to overcome the core defects of 'simple combination and lack of depth improvement' existing in the combination of quanta and classical algorithms in the prior art, and the practical requirements of fusion evaluation of three types of core data including power grid operation, equipment state and financial investment, the application provides a power grid multisource data fusion investment benefit post-evaluation platform based on a quanta