CN-122020441-A - Method for identifying abnormal mode of thermal energy storage system

CN122020441ACN 122020441 ACN122020441 ACN 122020441ACN-122020441-A

Abstract

The invention provides a method for identifying abnormal modes of a thermal energy storage system, which belongs to the technical field of thermal energy storage systems, and comprises the steps of collecting temperature distribution data, pressure fluctuation data and heat flux density data in the operation process of the thermal energy storage system, carrying out multi-scale wavelet decomposition on the temperature data to extract high-frequency and low-frequency components, carrying out time domain and frequency domain analysis on the pressure data to extract peak frequency and amplitude attenuation coefficients, and carrying out gradient calculation and topology analysis on the heat flow data to extract singular point distribution characteristics, fusing and normalizing the three types of characteristics to establish a system running state characteristic library, applying a trained model to the system running state characteristic library to extract deep abnormal characteristics, outputting an abnormal mode classification result and a confidence score, and judging the abnormal state when the confidence score exceeds a preset threshold value, thereby solving the technical problem of insufficient accuracy of identifying the abnormal mode of the heat energy storage system.

Inventors

LIU YUDI
ZHANG LEI
ZHANG CHENXI
WANG ZEZHONG
ZHU CHANG
WEI FEI
BAI DINGRONG

Assignees

鄂尔多斯实验室
清华大学

Dates

Publication Date: 20260512
Application Date: 20251209

Claims (10)

1. A thermal energy storage system abnormal pattern recognition method is characterized by comprising the steps of collecting system temperature distribution data, pressure fluctuation data and heat flux density data, conducting multi-scale wavelet decomposition on the temperature distribution data to establish a temperature characteristic vector set, conducting analysis on the pressure fluctuation data to establish a pressure characteristic vector set, conducting processing on the heat flux density data to establish a heat flux characteristic vector set, conducting feature fusion on the temperature characteristic vector set, the pressure characteristic vector set and the heat flux characteristic vector set to establish a system operation state feature library, extracting historical abnormal pattern samples from similar thermal energy storage system data sources, conducting field feature alignment to establish a migration learning source data set, constructing a thermal energy storage system mathematical simulation model, injecting abnormal disturbance parameters to generate a simulation abnormal sample set, conducting data enhancement processing to obtain an expansion training sample set, mixing the migration learning source data set with the expansion training sample set, establishing a mixed training data set, constructing an abnormal pattern recognition network model, conducting training by means of the mixed training data set, inputting the trained abnormal pattern recognition network model to the system operation state feature library, outputting an abnormal pattern classification result and an abnormal confidence score, when the abnormal confidence score is larger than a preset threshold, extracting abnormal condition parameters, acquiring a model based on the abnormal condition score, and conducting adjustment of the abnormal condition feature type based on the abnormal condition, and the abnormal condition feature score, conducting adjustment on the abnormal condition training based on the abnormal condition parameters, and conducting the abnormal condition training based on the abnormal condition parameters.
2. The method according to claim 1, wherein in the step of establishing the temperature feature vector set, a high frequency component and a low frequency component of the temperature distribution data are extracted, the high frequency component reflecting a rapid change characteristic of the temperature, the low frequency component reflecting a slow change trend of the temperature.
3. The method according to claim 2, wherein in the step of creating the pressure feature vector set, time domain analysis and frequency domain analysis are performed on the pressure fluctuation data, and a peak frequency and an amplitude attenuation coefficient of the pressure fluctuation data are extracted, wherein the peak frequency refers to a frequency value corresponding to the maximum amplitude of the pressure fluctuation data in a frequency spectrum after fourier transformation, and the amplitude attenuation coefficient refers to an attenuation rate of the amplitude of the pressure fluctuation data with time in a time domain.
4. A method according to claim 3, wherein in the step of establishing the heat flux feature vector set, gradient calculation is performed on the heat flux density data to obtain a heat flux density gradient field, topology analysis is performed on the heat flux density gradient field, and singular point distribution features of the heat flux density gradient field are extracted.
5. The method of claim 4, wherein in the step of establishing a system operation state feature library, a temperature feature vector set, a pressure feature vector set and a heat flow feature vector set are subjected to feature fusion to obtain fusion feature vectors, and the fusion feature vectors are normalized, wherein the feature fusion refers to vector splicing after aligning the temperature feature vector set, the pressure feature vector set and the heat flow feature vector set according to time stamps to form the fusion feature vectors containing multiple physical field information.
6. The method of claim 5, wherein the similar thermal energy storage system data source refers to a historical operating database of other thermal energy storage systems having similarities to the current thermal energy storage system in structural type, capacity level, and operating conditions.
7. The method of claim 6, wherein in the step of creating the migration learning source data set, domain feature alignment is performed on the historical abnormal pattern samples, wherein the domain feature alignment refers to mapping the feature distribution of the historical abnormal pattern samples to a feature space similar to the current thermal energy storage system by feature transformation by calculating the distribution difference between the historical abnormal pattern samples of the similar thermal energy storage system data sources and the current thermal energy storage system operation data in the feature space and measuring the distance between the two data distributions by using a maximum mean difference criterion.
8. The method according to claim 7, wherein in the step of generating the simulated abnormal sample set, abnormal disturbance parameters are injected into a mathematical simulation model of the thermal energy storage system, wherein the mathematical simulation model of the thermal energy storage system refers to a mathematical description model of the thermal energy storage system established based on a thermodynamic equation and a heat transfer equation, the mathematical simulation model of the thermal energy storage system simulates temperature field distribution, pressure field distribution and heat flow field distribution of the thermal energy storage system under different working conditions through a numerical solution method, and the abnormal disturbance parameters comprise a heat storage medium leakage rate parameter, a heat exchanger scaling thickness parameter, a circulating pump efficiency reduction rate parameter and an insulation layer aging coefficient parameter.
9. The method of claim 8, wherein the data enhancement process comprises performing random noise injection, time series slice reorganization and feature dimension shuffling on the simulated abnormal sample set, wherein the random noise injection refers to superimposing random noise subject to normal distribution on feature values of the simulated abnormal samples.
10. The method of claim 9, wherein the abnormal pattern recognition network model is structured such that an input layer receives the fused feature vector, a first convolution layer extracts local features, a second convolution layer extracts global features, a multi-headed attention mechanism layer performs weighted aggregation on the features, a recurrent neural network layer captures timing dependencies, and a full connection layer outputs an abnormal pattern classification result and an abnormal confidence score.

Description

Method for identifying abnormal mode of thermal energy storage system Technical Field The invention belongs to the technical field of heat energy storage systems, and particularly relates to a method for identifying abnormal modes of a heat energy storage system. Background The heat energy storage system is used as an important technical means for energy storage, is widely applied to the fields of electric power peak clipping and valley filling, regional heating, industrial waste heat recovery and the like, the traditional abnormal mode identification method mainly depends on single physical parameter monitoring or rule-based threshold judgment, the abnormal operation of the system is detected through setting a predefined rule such as an upper temperature limit, a lower temperature limit, a pressure alarm value, a flow abnormal range and the like, when the monitored parameters exceed the preset range, an alarm is triggered, a mature application mode is formed in a large-scale energy storage power station, a regional heating system and an industrial heat energy management platform, a part of advanced systems also adopt a single feature classifier based on machine learning, but the traditional parameter monitoring and shallow machine learning methods cannot effectively extract the deep abnormal discrimination features, so that the identification accuracy of abnormal modes such as local overheating, pressure oscillation, heat flow set and the like is insufficient, and the existing method is more prone to misjudgment or missed judgment especially when training data are limited or the abnormal mode features are similar. That is, the prior art has the technical problem that the abnormal mode identification accuracy of the thermal energy storage system is insufficient. Disclosure of Invention In view of the above, the invention provides a method for identifying abnormal modes of a thermal energy storage system, which can solve the technical problem of insufficient accuracy of identifying abnormal modes of the thermal energy storage system in the prior art. The invention provides a method for identifying abnormal modes of a thermal energy storage system, which comprises the following steps of collecting system temperature distribution data, pressure fluctuation data and heat flux density data, carrying out multi-scale wavelet decomposition on the temperature distribution data to establish a temperature characteristic vector set, analyzing the pressure fluctuation data to establish a pressure characteristic vector set, processing the heat flux density data to establish a heat flux characteristic vector set, carrying out characteristic fusion on the temperature characteristic vector set, the pressure characteristic vector set and the heat flux characteristic vector set to establish a system operation state characteristic library, extracting historical abnormal mode samples from a similar thermal energy storage system data source, carrying out field characteristic alignment to establish a migration learning source data set, constructing a thermal energy storage system mathematical simulation model, injecting abnormal disturbance parameters to generate a simulated abnormal sample set, carrying out data enhancement processing to obtain an extended training sample set, mixing the migration learning source data set with the extended training sample set to establish a mixed training data set, constructing an abnormal mode identification network model, carrying out training by utilizing the mixed training data set, inputting the training state characteristic library to complete abnormal mode identification network model, outputting an abnormal mode classification result and an abnormal confidence score, carrying out field characteristic alignment when the preset degree is larger than a preset degree, carrying out regulation and control on the abnormal operation state characteristic of the thermal energy storage system based on the abnormal state characteristic model, and regulating the abnormal operation characteristic parameters based on the abnormal operation characteristic model. The multi-scale wavelet decomposition refers to wavelet transformation of the temperature distribution data under different scales, and the temperature signal is decomposed into high-frequency components and low-frequency components with different frequency components through the localized characteristics of wavelet basis functions in a time domain and a frequency domain. In the step of establishing the pressure characteristic vector set, time domain analysis and frequency domain analysis are carried out on the pressure fluctuation data, and peak frequency and amplitude attenuation coefficient of the pressure fluctuation data are extracted, wherein the peak frequency refers to a frequency value corresponding to the maximum amplitude of the pressure fluctuation data in a frequency spectrum after Fourier transformation, and the amplitude attenuati