CN-121980445-A - Comprehensive energy system fault diagnosis method and system

CN121980445ACN 121980445 ACN121980445 ACN 121980445ACN-121980445-A

Abstract

The invention provides a fault diagnosis method and a fault diagnosis system for a comprehensive energy system, which belong to the technical field of fault diagnosis for the comprehensive energy system and comprise the steps of acquiring a data set of a new built comprehensive energy system of a target domain and a data set of an existing comprehensive energy system of the source domain; carrying out correlation analysis on each source domain operation data set and each target domain operation data set, calculating a correlation coefficient, reserving a source domain larger than a set threshold value to obtain B source domains, obtaining operation data sets and fault sample data sets corresponding to the B source domains, optimizing outer migration weights, initializing particles, calculating superimposed source domain operation data and superimposed fault sample data for weight vectors of each particle, calculating the correlation coefficient of the superimposed source domain operation data and the target domain operation data set, inputting the superimposed source domain fault sample data into an inner-layer Glow-ECNN model to obtain current fault diagnosis accuracy, and outputting nested circulation convergence and model.

Inventors

WEI ZHEN
ZHAO XIANGJUN
WANG YANG
LI YUANFU
LI YIJIA
DUAN PEI
SHI LIGUO

Assignees

国网山东省电力公司青岛供电公司

Dates

Publication Date: 20260505
Application Date: 20251229

Claims (10)

1. The fault diagnosis method of the comprehensive energy system is characterized by comprising the following steps of: Acquiring a data set of a new integrated energy system built in a target domain and a data set of an existing integrated energy system in a source domain; performing correlation analysis on each source domain operation data set and each target domain data set, calculating a correlation coefficient, reserving source domains larger than a set threshold value to obtain B source domains, and obtaining operation data sets and fault sample data sets corresponding to the B source domains; initializing particles, and calculating superimposed source domain operation data and superimposed fault sample data for the weight vector of each particle; calculating a correlation coefficient between the superimposed source domain operation data and the target domain operation data set to serve as a first fitness value; Inputting the superimposed source domain fault sample data into an inner layer Glow-ECNN model to obtain the current fault diagnosis accuracy as a second fitness value; And (3) nested cyclic convergence and model output, namely judging that the correlation and the fault accuracy are relatively optimal based on the first fitness value and the second fitness value, and if so, outputting a source domain integrated energy system operation data set with optimal weight and a fault diagnosis model with optimal network parameters for fault diagnosis of the target domain integrated energy system.
2. The comprehensive energy system fault diagnosis method as claimed in claim 1, wherein the superimposed source domain fault sample data is input into an inner layer Glow-ECNN model to obtain the current fault diagnosis accuracy, the inner layer Glow-ECNN model is formed by connecting a Glow module and a ECNN module in series, the former is responsible for probabilistic modeling and enhancement of fault characteristics, and the latter is responsible for characteristic extraction and fault classification.
3. The method for diagnosing faults of an integrated energy system as claimed in claim 1, further comprising inner layer model super parameter optimization, specifically comprising: Defining a hyper-parameter solution space; initializing an ant colony algorithm; generating a super-parameter combination; Dividing the superimposed fault sample data output by the outer layer into a training set and a testing set, initializing a Glow-ECNN model by using the current super-parameter combination, and calculating fault diagnosis accuracy Acc on the testing set after training to serve as an adaptability value of the super-parameter combination; and sequencing all the super-parameter combinations according to Acc, reserving the optimal combination, increasing the concentration of the pheromone for the corresponding paths, and volatilizing the pheromone for the rest paths according to rho 1 . If the current iteration number reaches a set value or the variation of the highest Acc of the continuous set iterations is smaller than a set threshold, feeding back the current highest Acc to the outer layer.
4. The method for diagnosing faults of an integrated energy system as claimed in claim 1, wherein defining a hyper-parametric solution space for determining a range of hyper-parameters to be optimized comprises: (1) ECNN, after random initialization, back-propagation fine tuning by a model, wherein the ant colony algorithm optimizes the initial weight range to be [ -0.1,0.1]; (2) Learning rate l r ; (3) The network layer number L; (4) The number of neurons in each layer N i .
5. The method for diagnosing faults of an integrated energy system as claimed in claim 1, wherein the step of nested cyclic convergence and model output is as follows: Taking the Acc fed back by the inner layer as a second fitness value of the outer layer, and continuing the outer layer iteration until the outer layer migration weight optimization converges; Selecting a weight vector with the optimal balance between the first fitness value and the second fitness value and a corresponding inner-layer optimal super-parameter combination from the outer-layer final Pareto optimal solution set; and training a Glow-ECNN model by using an optimal super-parameter combination to obtain a fault diagnosis model aiming at a newly built comprehensive energy system, and deploying the fault diagnosis model to a system monitoring platform to realize real-time fault diagnosis.
6. The method for diagnosing a fault in an integrated energy system as claimed in claim 1, wherein the step of optimizing the outer migration weight further comprises: The method comprises the steps of updating Pareto optimal solutions, namely, carrying out rapid non-dominant sorting on all particles, dividing dominant grades, calculating the crowding degree of each non-dominant solution, reserving solutions with high crowding degree, and updating a global Pareto optimal solution set G best ; updating the particle position, namely updating the speed and the position of each particle according to a speed and position updating formula, and adjusting the position beyond the constraint range; And (3) outer layer iteration judgment, namely if the current iteration number reaches a set value or the global Pareto optimal solution set is continuously set for iteration times without obvious change, entering an inner layer model super-parameter optimization step.
7. A comprehensive energy system fault diagnosis system, comprising: The data set acquisition module is configured to acquire a data set of a new building comprehensive energy system of a target domain and a data set of an existing comprehensive energy system of a source domain; The correlation analysis module is configured to perform correlation analysis on each source domain operation data set and each target domain data set, calculate a correlation coefficient, reserve source domains larger than a set threshold value to obtain B source domains, and obtain operation data sets and fault sample data sets corresponding to the B source domains; the outer migration weight optimization module is configured to initialize particles, calculate superimposed source domain operation data and superimposed fault sample data for weight vectors of each particle; calculating a correlation coefficient between the superimposed source domain operation data and the target domain operation data set to serve as a first fitness value; Inputting the superimposed source domain fault sample data into an inner layer Glow-ECNN model to obtain the current fault diagnosis accuracy as a second fitness value; And the nested loop convergence and model output module is configured to judge that the correlation and the fault accuracy are relatively optimal based on the first fitness value and the second fitness value, and if so, output a source domain integrated energy system operation data set with optimal weight and a fault diagnosis model with optimal network parameters for fault diagnosis of the target domain integrated energy system.
8. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 6.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any of the preceding claims 1-6 when the program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, performs the steps of the method of any of the preceding claims 1-6.

Description

Comprehensive energy system fault diagnosis method and system Technical Field The invention belongs to the technical field of comprehensive energy system fault diagnosis, and particularly relates to a comprehensive energy system fault diagnosis method and system. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. The comprehensive energy system integrates various energy forms such as electricity, heat, cold, gas and the like, improves the energy utilization efficiency through multi-energy complementation and collaborative optimization, and becomes a core component of a novel power system and an energy internet. However, the system has complex structure, various equipment types and changeable operation conditions, equipment faults are easy to occur, and if the equipment faults are not diagnosed and processed in time, chain reaction can be caused, so that large-area energy supply is interrupted, and huge economic loss is caused. The deep learning model has strong data feature extraction and pattern recognition capability, and is widely applied to the field of fault diagnosis. However, accurate training of the model depends on a large amount of marked complete fault sample data, and a newly constructed comprehensive energy system can only accumulate continuous operation monitoring data such as voltage, current, temperature, pressure and the like at the initial stage of operation, and the fault sample data is extremely deficient, namely, on one hand, the normal operation of the system can be destroyed by artificial manufacturing faults, the risk is high, the cost is high, on the other hand, naturally occurring faults have randomness, a large-scale sample set is difficult to form in a short time, so that the problem of data starvation of the traditional deep learning model is prominent, and a fault diagnosis model cannot be effectively constructed. The migration learning technology provides a thought for solving the problems, and the core of the technology is to migrate the knowledge of the existing source domain system with sufficient data, namely the existing comprehensive energy system, to the target domain system, namely the new comprehensive energy system, so as to make up the defect of insufficient data of the target domain. However, the application of the existing transfer learning in the fault diagnosis of the comprehensive energy system has the following defects: 1. the source domain system screening lacks quantification standard, only selects similar systems through experience, and data with weak relevance with the target domain is easy to introduce, so that migration effectiveness is reduced. 2. The source domain data is not given differential weight, and all source domain data by default have the same contribution, so that the value of the high-correlation source domain cannot be highlighted. 3. The migration process and model optimization are mutually independent, a closed loop is not formed, the quality of migration data can influence the model training effect, and the diagnosis feedback of the model can guide the optimization of the migration data, but the correlation is not utilized by the existing method, so that the overall diagnosis precision is limited. Disclosure of Invention In order to overcome the defects of the prior art, the invention provides a fault diagnosis method and a fault diagnosis system for a comprehensive energy system, which solve the problem of missing a new fault sample of the comprehensive energy system by quantitatively screening a source domain, optimizing migration weights by multiple targets, intelligently optimizing super parameters and iterating inner and outer layers of loops, and simultaneously improve the fault diagnosis precision and efficiency and meet the requirement of stable operation of the system. To achieve the above object, one or more embodiments of the present invention provide the following technical solutions: in a first aspect, a method for diagnosing faults of an integrated energy system is disclosed, comprising: Acquiring a data set of a new integrated energy system built in a target domain and a data set of an existing integrated energy system in a source domain; performing correlation analysis on each source domain operation data set and each target domain data set, calculating a correlation coefficient, reserving source domains larger than a set threshold value to obtain B source domains, and obtaining operation data sets and fault sample data sets corresponding to the B source domains; initializing particles, and calculating superimposed source domain operation data and superimposed fault sample data for the weight vector of each particle; calculating a correlation coefficient between the superimposed source domain operation data and the target domain operation data set to serve as a first fitness value; Inputting the superimposed sourc