CN-115598327-B - Method and device for predicting micro-water content in transformer oil based on improvement TRANSFORMER

CN115598327BCN 115598327 BCN115598327 BCN 115598327BCN-115598327-B

Abstract

The invention discloses a method and a device for predicting micro water content in Transformer oil based on an improved Transformer, which are used for obtaining a filtered micro water measurement signal and a temperature measurement signal, multiplying the filtered micro water measurement signal by (1+xi) to obtain a compensated micro water measurement signal, taking the compensated micro water measurement signal, the temperature measurement signal and weather digital information as training samples, training a neural network model by using the training samples to obtain a trained neural network model, acquiring the compensated micro water measurement signal in real time, displaying the compensated micro water measurement signal by a visual interface, inputting the real-time compensated micro water measurement signal, the future temperature measurement signal and the weather digital information into the trained neural network model to obtain the predicted Transformer water content in the future, and calculating the fault probability of the Transformer according to the predicted Transformer water content in the future. The invention provides support for monitoring the transformer in real time, finds potential faults of the transformer in time and provides basis for operation and maintenance of the transformer.

Inventors

PAN MING
LI FENGPAN
CAI LONG
YANG QINGFU
LIU YI
ZHU NING
YAO YANLIANG
ZHANG WENZE
GAO HAO
GU XIAOHU
LIU WEI
ZHANG XIAOTONG
WU BIN
SHEN JIANTAO
MA SINING

Assignees

上海置信电气有限公司
南京航空航天大学

Dates

Publication Date: 20260505
Application Date: 20220930

Claims (2)

1. The method for predicting the micro-water content in the Transformer oil based on the improved Transformer is characterized by comprising the following steps of: Step 1, acquiring a micro-water measurement signal and a temperature measurement signal in transformer oil, and performing sliding filtering treatment on the micro-water measurement signal and the temperature measurement signal to obtain a filtered micro-water measurement signal and a filtered temperature measurement signal; Step 2, under the condition that the water content in the transformer oil is unchanged, acquiring micro-water measurement signals under different set temperature influence conditions, fitting a linear equation of the micro-water measurement signals and the set temperature by adopting a linear least square method according to the micro-water measurement signals and the set temperature, and acquiring the slope xi of the linear equation as a compensation coefficient; Step 3, obtaining weather information, encoding the weather information into weather digital information, taking the compensated micro water measurement signal, the temperature measurement signal and the weather digital information as training samples, and training the neural network model by using the training samples to obtain a trained neural network model; step 4, acquiring the compensated micro water measurement signal in real time, and displaying the compensated micro water measurement signal by using a visual interface for real-time monitoring of the transformer; Step 5, acquiring a real-time compensated micro-water measurement signal, a future temperature measurement signal and weather digital information, and inputting the real-time compensated micro-water measurement signal, the future temperature measurement signal and the weather digital information into a trained neural network model to obtain a future predicted transformer water content; Step 6, calculating the fault probability of the transformer according to the predicted water content of the transformer in the future, sending out alarm information when the fault probability is greater than a threshold value, and arranging overhaul and maintenance; the sliding filter processing includes: every time a sampling value is obtained, N sampling values obtained by continuous sampling are placed in an array according to a first-in first-out principle and are arranged side by side, the largest two sampling values and the smallest two sampling values are removed, arithmetic average processing is carried out on the remaining N-4 sampling values, and data after filtering processing is obtained, wherein N is a fixed number, and the sampling values comprise a micro water measurement signal and a temperature measurement signal; training the neural network model by using the training sample to obtain a trained neural network model, comprising: step 3.1, acquiring compensated micro-water measurement signals, temperature measurement signals and weather digital information, and performing Z-Score standardization processing on the compensated micro-water measurement signals, temperature measurement signals and weather digital information to obtain standardized compensated micro-water measurement signals, temperature measurement signals and weather digital information; step 3.2, taking the standardized compensated micro water measurement signals, the temperature measurement signals and the weather digital information as training samples, and dividing the training samples into a training set, a verification set and a test set; step 3.3, inputting the training set into a transducer+SOM+LSTM model to obtain network parameters of the transducer+SOM+LSTM model; Step 3.4, inputting the verification set into a transducer+SOM+LSTM model to obtain an adjustment value of the network parameter; Substituting the adjustment value of the network parameter into a transducer+SOM+LSTM model, inputting a test set into the transducer+SOM+LSTM model to obtain a test predicted value, calculating an MAPE evaluation index according to the test predicted value, and taking the transducer+SOM+LSTM model with the MAPE evaluation index reaching the requirement as a trained neural network model; The neural network model adopts a transducer+SOM+LSTM model, the transducer+SOM+LSTM model comprises a transducer model, a decoder of the transducer model is replaced by a structure of the SOM neural network in series with the LSTM neural network, wherein Encoder of the transducer model in the transducer+SOM+LSTM model is used for extracting characteristics in data, and the structure of the SOM neural network in series with the LSTM neural network is used for clustering and predicting the characteristics.
2. The device for predicting the micro-water content in the Transformer oil based on the improved Transformer is characterized by comprising the following modules: the signal preprocessing module is used for acquiring a micro-water measurement signal and a temperature measurement signal in the transformer oil, and performing sliding filtering processing on the micro-water measurement signal and the temperature measurement signal to obtain a filtered micro-water measurement signal and a filtered temperature measurement signal; The compensation micro-water signal module is used for acquiring micro-water measurement signals under different set temperature influence conditions under the condition that the water content in the transformer oil is unchanged, fitting a linear equation of the micro-water measurement signals and the set temperature by adopting a linear least square method according to the micro-water measurement signals and the set temperature, and acquiring the slope xi of the linear equation as a compensation coefficient; The neural network model training module is used for acquiring weather information, encoding the weather information into weather digital information, taking the compensated micro water measurement signal, the temperature measurement signal and the weather digital information as training samples, and training the neural network model by using the training samples to obtain a trained neural network model; The real-time monitoring module is used for acquiring the compensated micro-water measurement signal in real time, displaying the compensated micro-water measurement signal by using a visual interface and monitoring the transformer in real time; the transformer water content prediction module is used for acquiring real-time compensated micro-water measurement signals, future temperature measurement signals and weather digital information, inputting the real-time compensated micro-water measurement signals, the future temperature measurement signals and the weather digital information into the trained neural network model, and acquiring future predicted transformer water content; The alarming module is used for calculating the fault probability of the transformer according to the predicted water content of the transformer in the future, sending out alarming information when the fault probability is greater than a threshold value, and arranging overhaul and maintenance; the sliding filter processing includes: every time a sampling value is obtained, N sampling values obtained by continuous sampling are placed in an array according to a first-in first-out principle and are arranged side by side, the largest two sampling values and the smallest two sampling values are removed, arithmetic average processing is carried out on the remaining N-4 sampling values, and data after filtering processing is obtained, wherein N is a fixed number, and the sampling values comprise a micro water measurement signal and a temperature measurement signal; training the neural network model by using the training sample to obtain a trained neural network model, comprising: step 3.1, acquiring compensated micro-water measurement signals, temperature measurement signals and weather digital information, and performing Z-Score standardization processing on the compensated micro-water measurement signals, temperature measurement signals and weather digital information to obtain standardized compensated micro-water measurement signals, temperature measurement signals and weather digital information; step 3.2, taking the standardized compensated micro water measurement signals, the temperature measurement signals and the weather digital information as training samples, and dividing the training samples into a training set, a verification set and a test set; step 3.3, inputting the training set into a transducer+SOM+LSTM model to obtain network parameters of the transducer+SOM+LSTM model; Step 3.4, inputting the verification set into a transducer+SOM+LSTM model to obtain an adjustment value of the network parameter; Substituting the adjustment value of the network parameter into a transducer+SOM+LSTM model, inputting a test set into the transducer+SOM+LSTM model to obtain a test predicted value, calculating an MAPE evaluation index according to the test predicted value, and taking the transducer+SOM+LSTM model with the MAPE evaluation index reaching the requirement as a trained neural network model; The neural network model adopts a transducer+SOM+LSTM model, the transducer+SOM+LSTM model comprises a transducer model, a decoder of the transducer model is replaced by a structure of the SOM neural network in series with the LSTM neural network, wherein Encoder of the transducer model in the transducer+SOM+LSTM model is used for extracting characteristics in data, and the structure of the SOM neural network in series with the LSTM neural network is used for clustering and predicting the characteristics.

Description

Method and device for predicting micro-water content in transformer oil based on improvement TRANSFORMER Technical Field The invention relates to a method and a device for predicting micro-water content in Transformer oil based on an improved Transformer, and belongs to the technical field of monitoring of power equipment. Background The power transformer is used as an important pivot in the power transmission and distribution links and is widely distributed in the power grid. The oil immersed transformer has the characteristics of excellent insulating property and good economy, and occupies a large specific gravity. But due to the large number of them, the distance is far away, which is difficult to achieve and can effectively maintain safely. At present, the insulation performance of equipment is reduced due to the fact that the micro water content in transformer oil exceeds the standard, and the damage and downtime of the equipment due to the fact that partial discharge breakdown is further caused becomes one of the primary problems in the industry. When the micro water content of the transformer oil is increased, the moisture not only can reduce the breakdown voltage of the insulation system and increase the dielectric loss, but also can directly participate in the chemical degradation reaction of high polymer materials such as oilpaper fibers and the like to promote the degradation and aging of the materials, so that the degradation of various performances of the insulation system is accelerated, when the micro water content in the oil exceeds a certain threshold value, the insulation performance of the equipment is greatly reduced, and serious accidents such as insulation breakdown, equipment burning and the like can be caused. And the combination of micro water and organic acids in oil not only reduces the insulation capability of the transformer oil, but also causes the loss of arc extinguishing capability. According to our national relevant standards (GB/T7595-2000) it is prescribed that transformers of the voltage class of 35kV and below, the breakdown voltage of the oil in operation should not be lower than 30kV, whereas for transformers of the voltage class of 500kV, the breakdown voltage of the oil in operation should be 50kV. Through the water content in the monitoring transformer oil, not only can prevent that transformer oil insulating strength from reducing to dangerous level, but also can evaluate the whole insulating situation of transformer and make the judgement to the leakproofness of equipment according to the size of moisture content in the oil. Therefore, how to realize online detection of the micro water content of the transformer in the running state and predict the variation trend of the micro water content through an algorithm is a technical problem which needs to be solved by those skilled in the art. Disclosure of Invention The invention provides a method and a device for predicting the micro water content in Transformer oil based on an improved Transformer, and aims to overcome the defects of high micro water measurement cost and poor real-time property in the existing Transformer oil in the prior art. The technical scheme adopted by the invention is as follows: in a first aspect, a method for predicting micro-water content in transformer oil based on improvement TRANSFORMER includes the steps of: step 1, acquiring a micro-water measurement signal and a temperature measurement signal in transformer oil, and performing sliding filtering treatment on the micro-water measurement signal and the temperature measurement signal to obtain a filtered micro-water measurement signal and a filtered temperature measurement signal. And 2, under the condition that the water content in the transformer oil is unchanged, acquiring micro-water measurement signals under different set temperature influence conditions, fitting a linear equation of the micro-water measurement signals and the set temperature by adopting a linear least square method according to the micro-water measurement signals and the set temperature, and acquiring the slope xi of the linear equation as a compensation coefficient. Multiplying the filtered micro water measurement signal by (1+ζ) to obtain a compensated micro water measurement signal. And step 3, obtaining weather information, encoding the weather information into weather digital information, taking the compensated micro water measurement signal, the temperature measurement signal and the weather digital information as training samples, and training the neural network model by using the training samples to obtain a trained neural network model. And 4, acquiring the compensated micro water measurement signal in real time, and displaying the compensated micro water measurement signal by using a visual interface for real-time monitoring of the transformer. And 5, acquiring a real-time compensated micro-water measurement signal, a future temperature measurement signal