System and method for evaluating state of vehicle-mounted voltage transformer of electric locomotive and motor train unit

Abstract

The invention discloses a state evaluation system and method for a vehicle-mounted voltage transformer of an electric locomotive and a motor train unit. The system comprises a data acquisition layer, a state evaluation layer and a data storage layer, the method specifically comprises the following steps that: a data acquisition layer acquires and sends the characteristic information amount of the vehicle-mounted voltage transformer in working condition operation to a state evaluation layer; the state evaluation layer is used for calculating the characteristic information amount of the vehicle-mounted voltage transformer, evaluating the health state of the vehicle-mounted voltage transformer and comparing and predicting the future state in combination with the current influence factors; and the data storage layer is used for storing data required by the training model, an evaluation prediction result and characteristic information amount. According to the invention, on-line monitoring of information characteristic quantities such as secondary side voltage, leakage current and local discharge capacity of the vehicle-mounted voltage transformer is utilized, state evaluation and future prediction of the transformer are realized, the operation and maintenance efficiency of equipment can be effectively improved, and the influence of faults on train and personnel safety is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power metering, and in particular to a system and method for evaluating the state of an onboard voltage transformer of an electric locomotive or a multiple-unit train. Background Art

[0002] Electromagnetic voltage transformers are used in EMU locomotives because of their high precision and large capacity. Their disadvantage is that the core is easily magnetically saturated and the linearity will be reduced when overloaded. The high-speed rail field is the most prone to 27.5kV voltage transformer failures, accounting for more than 70%. The on-board voltage transformer is installed on the roof, located between the contact network and the circuit breaker. It always maintains connection with the contact network during the operation of the locomotive. Most of the failures are caused by repeated overvoltage shocks when the locomotive is over-phased. At present, the failure of the roof voltage transformer is mostly manifested as the transformer bursting. When the accident occurs, the traction network has overvoltage, and the outside of the bursting equipment and the primary coil are burned and blackened, while the secondary side is intact. The main causes of explosion accidents are as follows: (1) Defects inside the transformer cause local discharge during operation, resulting in a short circuit between coil turns, which ultimately leads to an increase in current and internal overheating and explosion; (2) When the voltage transformer is subjected to abnormal overvoltage in the contact network, local discharge occurs, causing the insulation between coil turns to burn; (3) When the transformer is subjected to ferromagnetic resonance overvoltage, it becomes magnetically saturated and the internal energy cannot be released, resulting in an explosion.

[0003] At present, voltage transformers are widely used in electric locomotives and EMUs, but the status assessment of voltage transformers themselves has not been well studied and solved. The application of online monitoring systems on voltage transformers in power systems has matured, but the online monitoring technology of on-board voltage transformers on electric locomotives is still in its early stages and requires further in-depth research. Existing technologies mainly monitor the secondary voltage of voltage transformers, without comprehensively considering the impact of multiple information characteristics on the operating status of voltage transformers. At the same time, the current work focuses on real-time monitoring of its status, and the prediction of the future status of voltage transformers needs to be improved. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a system and method for evaluating the state of an on-board voltage transformer of an electric locomotive and an EMU, the system comprising a data acquisition layer, a state evaluation layer and a data storage layer; the method is specifically as follows: the data acquisition layer acquires and sends to the state evaluation layer the characteristic information of the on-board voltage transformer under operating conditions; the state evaluation layer is used to calculate and process the characteristic information of the on-board voltage transformer, to evaluate the health state of the on-board voltage transformer and to predict the future state by comparing with the current influencing factors; the data storage layer realizes the storage of the data required for the training model as well as the evaluation prediction results and the characteristic information.

[0005] In order to solve the above technical problems, the technical solution provided by the present invention is:

[0006] Electric locomotive and EMU on-board voltage transformer status assessment system, the system includes a data acquisition layer, a data storage layer and a status assessment layer;

[0007] The data acquisition layer includes: a voltage sensor, a current sensor, an ultrasonic partial discharge sensor, a data acquisition module, a data transmission module 1 and a power supply module 1, wherein the voltage sensor, the current sensor and the ultrasonic partial discharge sensor are respectively connected to the vehicle-mounted voltage transformer and the data acquisition module;

[0008] The voltage sensor is used to monitor the secondary side voltage of the on-board voltage transformer;

[0009] The current sensor is used to monitor the leakage current of the on-board voltage transformer;

[0010] The ultrasonic partial discharge sensor is used to determine whether partial discharge occurs in the voltage transformer by detecting the size of the electrical signal, and locate the partial discharge by using the propagation principle of ultrasonic waves in the air;

[0011] The data acquisition layer collects characteristic information of leakage current, secondary voltage value and partial discharge of the voltage transformer through the data acquisition module, and sends it to the state assessment layer through the data transmission module 1;

[0012] The data storage layer includes a data storage module, a data transmission module 2 and a power supply module 2;

[0013] The data storage module is used to store the original state data of the voltage transformer information characteristic quantity under the working condition sent by the data transmission module 1 and the data updated by the state evaluation layer;

[0014] The data transmission module 2 is used to send the data stored in the data storage module to the status assessment layer;

[0015] The state assessment layer includes a model training module, a state assessment module, a prediction module, a data transmission module 3, an early warning alarm module and a power supply module 3; the state assessment layer is used to assess the current health state of the voltage transformer and predict its future health state;

[0016] The model training module is used to train the health status assessment model based on the feature vector set sent by the data transmission module 2;

[0017] The state evaluation module is used to evaluate the current health state of the voltage transformer under the influence of current factors using the model trained by the model training module;

[0018] The early warning alarm module is used to receive the evaluation result of the state evaluation module, and send out an early warning signal when the characteristic electrical quantity of the voltage transformer is abnormal, and send out an alarm signal when a voltage transformer fails;

[0019] The prediction module is used to predict the future health state of the voltage transformer under the influence of the characteristic information of the leakage current, the secondary side voltage value and the partial discharge amount by using the model trained by the model training module;

[0020] The data transmission module three is used to transmit the evaluation results of the state evaluation module and the prediction results of the prediction module to the data storage module.

[0021] The method for evaluating the state of the on-board voltage transformer of an electric locomotive or EMU is based on the above-mentioned on-board voltage transformer state evaluation system of an electric locomotive or EMU, and comprises the following steps:

[0022] S1. The data acquisition layer collects the secondary voltage U, leakage current I, phase θ and partial discharge characteristic information P of the on-board voltage transformer under the working condition monitored by the voltage sensor, current sensor and ultrasonic partial discharge sensor through the data acquisition module;

[0023] S2. The characteristic information collected by S1 is sent to the data storage module of the data storage layer through the data transmission module 1 for storage, and then sent to the training module of the status evaluation layer through the data transmission module 2 to establish the health status evaluation LSTM-1 model and perform training;

[0024] S3. The characteristic information collected by S1 is sent to the state evaluation module of the state evaluation layer through the data transmission module 1, and the health status evaluation LSTM-1 model trained by S2 is used to evaluate the health status of the on-board voltage transformer under the influence of the characteristic information of the leakage current, the secondary side voltage value and the partial discharge amount, and the evaluation result of the current health status of the on-board voltage transformer is obtained, and sent to the data storage module for storage through the data transmission module 3;

[0025] The evaluation result of S4.S3 is transmitted to the alarm module for processing, and a warning signal is issued when the characteristic electrical quantity of the voltage transformer is abnormal, and an alarm signal is issued when a voltage transformer fails;

[0026] S5. The historical data of the characteristic information of leakage current, secondary voltage value and partial discharge amount stored in the data storage module are sent to the prediction module of the state evaluation layer through the data transmission module 2, and the health state evaluation LSTM-2 model is established and trained; the future health state of the voltage transformer under the influence of the characteristic information of leakage current, secondary voltage value and partial discharge amount is predicted by using the trained LSTM-2 model, and the prediction result of the future health state of the on-board voltage transformer is obtained;

[0027] S6. The data transmission module 3 of the state evaluation layer transmits the characteristic information of the evaluation result of S3 and the prediction result of S5 to the data storage module of the data storage layer for storage.

[0028] Further,

[0029] The specific steps of S2 using the characteristic information data stored in the data storage module to establish the health status assessment LSTM-1 model and perform training are as follows:

[0030] S2.1 Assume that the health assessment coefficient is H. By calculating and analyzing the health assessment coefficient, a quantitative determination of the health level is obtained. The calculation formula is:

[0031] H n =aU n +bI n +cθ n +dt n +eP n +f (1)

[0033] Where 1≤n≤k and it is an integer, H is the overall health assessment coefficient; U is the secondary voltage; I is the leakage current, the phase is θ, P is the partial discharge amount; f is the interference correction constant;

[0034] According to the proportion of different signal characteristic values ​​in the running status evaluation, different weights a, b, c, d and e are assigned to them;

[0035] Set the health critical standard to x, where x is an empirical parameter obtained from actual engineering or set according to actual conditions; when H>x, the operation state is fault operation; when H<x, the operation state is normal operation;

[0036] S2.2 takes the measured data as the training set and trains it using the mean square error loss function;

[0037]

[0038] Among them, H i is the actual value, is the evaluation value;

[0039] S2.3 divides time t into M1 intervals as large time intervals; each large time interval is divided into M2 intervals as small time intervals; the data of the large time interval is used as long-term memory data, and the data of the small time interval is used as short-term memory data;

[0040] S2.4 establishes a long short-term memory neural network model LSTM-1 to evaluate the operating status of the voltage transformer and eliminate false alarms caused by temporary voltage fluctuations under working conditions;

[0041] Under the S2.5 operating condition, the secondary voltage of the voltage transformer is relatively stable, and the secondary voltage is closely related to the operating time. In the establishment of the long-term and short-term memory neural network, the weight of the long-term memory data is strengthened to obtain the evaluation value based on the change of a large time interval.

[0042] S2.6. Under working conditions, the changes in leakage current and partial discharge reflect the status of the voltage transformer in real time, and their fluctuations are less affected by time. In the establishment of the long-short-term memory neural network, the weight of short-term memory data is strengthened to obtain the evaluation value based on the changes in small time intervals;

[0043] S2.7 The collected actual data is divided into training data and verification data, of which 70% is used for training and 30% is used for verification;

[0044] S2.8 divide the weight matrix into large time intervals and small time intervals;

[0045] S2.9 Train and verify the long short-term memory neural network model LSTM-1.

[0046] Further,

[0047] The state evaluation module in S3 uses the information feature weighted comparison method to weight the secondary side voltage, leakage current and partial discharge respectively to obtain the state evaluation coefficient H, and then compares it with the current actual state coefficient after ratio calculation with the judgment threshold, so as to evaluate the state of the voltage transformer. The specific steps are as follows:

[0048] S3.1 Assume that the root mean square error calculation result within the M2 hour time interval is as follows:

[0049]

[0050] Among them, H i represents the i-th actual measurement value; represents the i-th evaluation result value; n represents a total of n sampling points in the small time interval;

[0051] The conditions for determining magnetic field fluctuations are:

[0052] |RMSE-RMSE train |>ε (4)

[0053] Among them, RMSE train It represents the average value of RMSE of the two small time intervals measured in the training set; ε represents the accuracy, which is set according to engineering experience;

[0054] S3.2 When voltage fluctuates |RMSE-RMSE train When |>ε is satisfied, the time interval is expanded to the large time interval that includes the small time interval, and its MAE value is calculated. The MAE calculation formula is as follows:

[0055]

[0056] Among them, H i represents the i-th actual measurement value; represents the i-th actual result value; N represents that there are N sampling points in this large time interval;

[0057] The conditions for non-periodic fluctuation of magnetic field are:

[0058] |MAE-MAE train |<δ (6)

[0059] Among them, MAE train It represents the MAE value of adjacent large time intervals measured in the training set; δ represents the accuracy, which is set according to engineering experience;

[0060] When H>x, the operation state is fault operation, and when H<x, the operation state is normal operation.

[0061] Further,

[0062] The early warning alarm module in S4 has three output states: ① normal state: no alarm signal is issued, and the indicator light is green; ② abnormal state: a warning signal is issued, and the indicator light is yellow; ③ fault state: a fault alarm signal is issued, and the indicator light is red.

[0063] Further,

[0064] The specific steps of S5 are:

[0065] S5.1 Assume that the health assessment prediction coefficient is J. By calculating and analyzing the health assessment coefficient, a quantitative determination of health status is obtained. The calculation formula is:

[0066]

[0067] Where, 1≤n≤k and is an integer, J is the overall health assessment coefficient; U1—U n is the historical data of the secondary side voltage; I1—I n is the leakage current history data, θ1—θ n is the phase history data, P1-P n is the historical data of partial discharge; F is the interference correction constant;

[0068] Different weights A, B, C, D and E are assigned to different signal feature values ​​according to their proportion in the running status evaluation;

[0069] Set the health critical standard to X, where X is set based on the empirical parameters obtained from actual engineering or based on actual conditions; when J>X, the operation state is fault operation; when J<X, the operation state is normal operation;

[0070] S5.2 takes the measured data as the training set and trains it using the mean square error loss function;

[0071]

[0072] Among them, J i is the actual value, is the predicted value;

[0073] S5.3 divides time t into M1 intervals as large time intervals; each large time interval is divided into M2 intervals as small time intervals; the data of the large time interval is used as long-term memory data, and the data of the small time interval is used as short-term memory data;

[0074] S5.4 establishes a long short-term memory neural network model LSTM-2 to predict the health status of the voltage transformer;

[0075] Under the S5.5 operating condition, the secondary voltage of the voltage transformer is relatively stable, and the secondary voltage is closely related to the operating time. In the establishment of the long-term and short-term memory neural network, the weight of the long-term memory data is strengthened to obtain the predicted value based on the change of a large time interval;

[0076] S5.6. Under the operating condition, the changes of leakage current and partial discharge reflect the status of voltage transformer in real time, and their fluctuations are less affected by time. In the establishment of long short-term memory neural network, the weight of short-term memory data is strengthened to obtain the evaluation prediction value based on the change of small time interval;

[0077] S5.7 The collected actual data is divided into training data and verification data, of which 70% is used for training and 30% is used for verification;

[0078] S5.8 divide the weight matrix into large time intervals and small time intervals;

[0079] S5.9 trains the long short-term memory neural network model LSTM-2 and verifies the prediction results.

[0080] The beneficial effects of the present invention are:

[0081] The system and method for evaluating the state of the on-board voltage transformer for power locomotives and EMUs provided by the present invention comprehensively consider the influence of various information characteristics on the operating state of the voltage transformer, utilize the online monitoring of information characteristics such as the secondary side voltage, leakage current and partial discharge of the on-board voltage transformer, and simultaneously realize the state evaluation and future prediction of the transformer, which can effectively improve the operation and maintenance efficiency of the equipment and avoid failures that affect the safety of trains and personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0083] Figure 1 Schematic diagram of the voltage transformer health status assessment system of the present invention

[0084] Figure 2 The flow chart of the voltage transformer state evaluation method of the present invention is DETAILED DESCRIPTION

[0085] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the following embodiments are provided only for the purpose of illustration and are not intended to limit the scope of the present invention. Those skilled in the art may make various modifications and substitutions to the present invention without departing from the purpose and spirit of the present invention.

[0086] like Figure 1 As shown, the present invention is a schematic diagram of the on-board voltage transformer state evaluation system of the electric locomotive and EMU of the present invention, and the system includes: a data acquisition layer, a state evaluation layer, and a data storage layer.

[0087] (I) Data acquisition layer, including: voltage sensor module, current sensor module, ultrasonic partial discharge sensor module, data acquisition module, data transmission module 1 and power supply module, used to send characteristic information of the operating environment and reaction working status of the voltage transformer to the state evaluation layer, leakage current, secondary side voltage value and partial discharge amount as reference characteristic values ​​to provide judgment basis for the state evaluation layer, and transmit the collected information to the data storage layer through the data transmission module 1 for training the model.

[0088] The voltage sensor, current sensor and ultrasonic partial discharge sensor are connected to the voltage transformer and the data acquisition module respectively. The data acquisition module collects the status information of the voltage transformer based on the edge sensing devices such as the voltage sensor, current sensor and ultrasonic partial discharge sensor. Among them, the voltage sensor monitors the secondary side voltage of the voltage transformer by reducing the voltage signal generated by the voltage transformer to make it conform to the input range of the voltage and current acquisition card; the current sensor is used to realize the monitoring of the leakage current of the vehicle-mounted voltage transformer; the ultrasonic partial discharge sensor is used to determine whether the voltage transformer has partial discharge by detecting the size of the electrical signal, and locate the partial discharge by using the principle of ultrasonic propagation in the air.

[0089] The data transmission module 1 is used to upload the collected real-time status data to the data status evaluation layer for processing through wireless communication transmission methods such as LoRa or 5G.

[0090] The power supply module is used to supply power to the voltage sensor, current sensor, ultrasonic partial discharge sensor and data acquisition module.

[0091] (II) The data storage layer includes a data storage module, a data transmission module 2 and a power supply module 2. The data storage module is used to store the original state data of the voltage transformer information characteristic quantity under the working condition sent by the data transmission module 1 and the data updated by the state evaluation layer; the data transmission module 2 is used to send the data stored in the data storage module to the state evaluation layer. The power supply module 2 is used to realize the power supply of the data storage module and the data transmission module 2.

[0092] (III) The status assessment layer includes a model training module, a status assessment module, a prediction module, a data transmission module 3, an early warning alarm module and a power supply module 3; the status assessment layer processes and predicts information feature quantities and reasonably allocates long-term and short-term dependencies between information feature quantities to assess the current health status of the voltage transformer, diagnose and warn of possible equipment failures, and predict the future health status of the equipment.

[0093] The model training module is used to train the health status assessment model by calling a large amount of historical data based on the feature vector set sent by the data transmission module 2; the status assessment module is used to evaluate the current health status of the voltage transformer under the influence of current factors using the model trained by the model training module; the early warning alarm module is used to receive the assessment results of the status assessment module, issue a warning signal when the characteristic electrical quantity of the voltage transformer is abnormal, and issue an alarm signal when the voltage transformer fails; the prediction module is used to predict the future health status of the voltage transformer under the influence of the characteristic information of the leakage current, secondary voltage value and partial discharge using the model trained by the model training module; the data transmission module 3 is used to transmit the assessment results of the status assessment module and the prediction results of the prediction module to the data storage module. The power supply module 3 is used to realize the power supply of the model training module, the prediction module, the status assessment module, the data transmission module and the early warning alarm module.

[0094] like Figure 2 As shown, the present invention also provides a flow chart of a method for evaluating the state of an on-board voltage transformer of an electric locomotive and an EMU, comprising the following steps:

[0095] S1. The data acquisition layer collects the secondary voltage U, leakage current I, phase θ and partial discharge characteristic information P of the on-board voltage transformer under the working condition monitored by the voltage sensor, current sensor and ultrasonic partial discharge sensor through the data acquisition module;

[0096] S2. The characteristic information collected by S1 is sent to the data storage module of the data storage layer through the data transmission module 1 for storage, and then sent to the training module of the status evaluation layer through the data transmission module 2 to establish the health status evaluation LSTM-1 model and perform training;

[0097] S3. The characteristic information collected by S1 is sent to the state evaluation module of the state evaluation layer through the data transmission module 1, and the health status evaluation LSTM-1 model trained by S2 is used to evaluate the health status of the on-board voltage transformer under the influence of the characteristic information of the leakage current, the secondary side voltage value and the partial discharge amount, and the evaluation result of the current health status of the on-board voltage transformer is obtained, and sent to the data storage module for storage through the data transmission module 3;

[0098] The evaluation result of S4.S3 is transmitted to the alarm module for processing, and a warning signal is issued when the characteristic electrical quantity of the voltage transformer is abnormal, and an alarm signal is issued when a voltage transformer fails;

[0099] S5. The historical data of the characteristic information of leakage current, secondary voltage value and partial discharge amount stored in the data storage module are sent to the prediction module of the state evaluation layer through the data transmission module 2, and the health state evaluation LSTM-2 model is established and trained; the future health state of the voltage transformer under the influence of the characteristic information of leakage current, secondary voltage value and partial discharge amount is predicted by using the trained LSTM-2 model, and the prediction result of the future health state of the on-board voltage transformer is obtained;

[0100] S6. The data transmission module 3 of the state evaluation layer transmits the characteristic information of the evaluation result of S3 and the prediction result of S5 to the data storage module of the data storage layer for storage.

[0101] Among them, S2 uses the characteristic information data stored in the data storage module to establish the health status assessment LSTM-1 model and conduct training in the following specific steps:

[0102] S2.1 Assume that the health assessment coefficient is H. By calculating and analyzing the health assessment coefficient, a quantitative determination of the health level is obtained. The calculation formula is:

[0103] H n =aU n +bI n +cθ n +dt n +eP n +f (1)

[0105] Where 1≤n≤k and it is an integer, H is the overall health assessment coefficient; U is the secondary voltage; I is the leakage current, the phase is θ, P is the partial discharge amount; f is the interference correction constant;

[0106] According to the proportion of different signal characteristic values ​​in the running status evaluation, different weights a, b, c, d and e are assigned to them;

[0107] Set the health critical standard to x, where x is an empirical parameter obtained from actual engineering or set according to actual conditions; when H>x, the operation state is fault operation; when H<x, the operation state is normal operation;

[0108] S2.2 takes the measured data as the training set and trains it using the mean square error loss function;

[0109]

[0110] Among them, H i is the actual value, is the evaluation value;

[0111] S2.3 divides time t into M1 intervals as large time intervals; each large time interval is divided into M2 intervals as small time intervals; the data of the large time interval is used as long-term memory data, and the data of the small time interval is used as short-term memory data;

[0112] S2.4 establishes a long short-term memory neural network model LSTM-1 to evaluate the operating status of the voltage transformer and eliminate false alarms caused by temporary voltage fluctuations under working conditions;

[0113] Under the S2.5 operating condition, the secondary voltage of the voltage transformer is relatively stable, and the secondary voltage is closely related to the operating time. In the establishment of the long-term and short-term memory neural network, the weight of the long-term memory data is strengthened to obtain the evaluation value based on the change of a large time interval.

[0114] S2.6. Under working conditions, the changes in leakage current and partial discharge reflect the status of the voltage transformer in real time, and their fluctuations are less affected by time. In the establishment of the long-short-term memory neural network, the weight of short-term memory data is strengthened to obtain the evaluation value based on the changes in small time intervals;

[0115] S2.7 The collected actual data is divided into training data and verification data, of which 70% is used for training and 30% is used for verification;

[0116] S2.8 divide the weight matrix into large time intervals and small time intervals;

[0117] S2.9 Train and verify the long short-term memory neural network model LSTM-1.

[0118] Among them, the state evaluation module in S3 uses the information feature weighted comparison method to weight the secondary side voltage, leakage current and partial discharge respectively to obtain the state evaluation coefficient H, and then calculates the ratio with the current actual state coefficient and compares it with the judgment threshold to evaluate the state of the voltage transformer. The specific steps are:

[0119] S3.1 Assume that the root mean square error calculation result within the M2 hour time interval is as follows:

[0120]

[0121] Among them, H i represents the i-th actual measurement value; represents the i-th evaluation result value; n represents a total of n sampling points in the small time interval;

[0122] The conditions for determining magnetic field fluctuations are:

[0123] |RMSE-RMSE train |>ε (4)

[0124] Among them, RMSE train It represents the average value of RMSE of the two small time intervals measured in the training set; ε represents the accuracy, which is set according to engineering experience;

[0125] S3.2 When voltage fluctuates |RMSE-RMSE train When |>ε is satisfied, the time interval is expanded to the large time interval that includes the small time interval, and its MAE value is calculated. The MAE calculation formula is as follows:

[0126]

[0127] Among them, H i represents the i-th actual measurement value; represents the i-th actual result value; N represents that there are N sampling points in this large time interval;

[0128] The conditions for non-periodic fluctuation of magnetic field are:

[0129] |MAE-MAE train |<δ (6)

[0130] Among them, MAE train It represents the MAE value of adjacent large time intervals measured in the training set; δ represents the accuracy, which is set according to engineering experience;

[0131] When H>x, the operation state is fault operation, and when H<x, the operation state is normal operation.

[0132] Among them, the early warning alarm module in S4 has three output states: ① normal state: no alarm signal is issued, the indicator light is green; ② abnormal state: a warning signal is issued, the indicator light is yellow; ③ fault state: a fault alarm signal is issued, and the indicator light is red.

[0133] Among them, the specific steps of S5 are:

[0134] S5.1 Assume that the health assessment prediction coefficient is J. By calculating and analyzing the health assessment coefficient, a quantitative determination of health status is obtained. The calculation formula is:

[0135]

[0136] Where, 1≤n≤k and is an integer, J is the overall health assessment coefficient; U1—U n is the historical data of the secondary side voltage; I1—I n is the leakage current history data, θ1—θ n is the phase history data, P1-P n is the historical data of partial discharge; F is the interference correction constant;

[0137] Different weights A, B, C, D and E are assigned to different signal feature values ​​according to their proportion in the running status evaluation;

[0138] Set the health critical standard to X, where X is set based on the empirical parameters obtained from actual engineering or based on actual conditions; when J>X, the operation state is fault operation; when J<X, the operation state is normal operation;

[0139] S5.2 takes the measured data as the training set and trains it using the mean square error loss function;

[0140]

[0141] Among them, J i is the actual value, is the predicted value;

[0142] S5.3 divides time t into M1 intervals as large time intervals; each large time interval is divided into M2 intervals as small time intervals; the data of the large time interval is used as long-term memory data, and the data of the small time interval is used as short-term memory data;

[0143] S5.4 establishes a long short-term memory neural network model LSTM-2 to predict the health status of the voltage transformer;

[0144] Under the S5.5 operating condition, the secondary voltage of the voltage transformer is relatively stable, and the secondary voltage is closely related to the operating time. In the establishment of the long-term and short-term memory neural network, the weight of the long-term memory data is strengthened to obtain the predicted value based on the change of a large time interval;

[0145] S5.6. Under the operating condition, the changes of leakage current and partial discharge reflect the status of voltage transformer in real time, and their fluctuations are less affected by time. In the establishment of long short-term memory neural network, the weight of short-term memory data is strengthened to obtain the evaluation prediction value based on the change of small time interval;

[0146] S5.7 The collected actual data is divided into training data and verification data, of which 70% is used for training and 30% is used for verification;

[0147] S5.8 divide the weight matrix into large time intervals and small time intervals;

[0148] S5.9 trains the long short-term memory neural network model LSTM-2 and verifies the prediction results.

[0149] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0150] Finally, it should be noted that the above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The on-board voltage transformer status assessment system for electric locomotives and EMUs is characterized by: The system includes a data collection layer, a data storage layer and a status assessment layer; The data acquisition layer includes: a voltage sensor, a current sensor, an ultrasonic partial discharge sensor, a data acquisition module, a data transmission module 1 and a power supply module 1, wherein the voltage sensor, the current sensor and the ultrasonic partial discharge sensor are respectively connected to the vehicle-mounted voltage transformer and the data acquisition module; The voltage sensor is used to monitor the secondary side voltage of the on-board voltage transformer; The current sensor is used to monitor the leakage current of the on-board voltage transformer; The ultrasonic partial discharge sensor is used to determine whether partial discharge occurs in the voltage transformer by detecting the size of the electrical signal, and locate the partial discharge by using the propagation principle of ultrasonic waves in the air; The data acquisition layer collects characteristic information of leakage current, secondary voltage value and partial discharge of the voltage transformer through the data acquisition module, and sends it to the state assessment layer through the data transmission module 1; The data storage layer includes a data storage module, a data transmission module 2 and a power supply module 2; The data storage module is used to store the original state data of the voltage transformer information characteristic quantity under the working condition sent by the data transmission module 1 and the data updated by the state evaluation layer; The data transmission module 2 is used to send the data stored in the data storage module to the status assessment layer; The state assessment layer includes a model training module, a state assessment module, a prediction module, a data transmission module 3, an early warning alarm module and a power supply module 3; the state assessment layer is used to assess the current health state of the voltage transformer and predict its future health state; The model training module is used to train the health status assessment model based on the feature vector set sent by the data transmission module 2; The state evaluation module is used to evaluate the current health state of the voltage transformer under the influence of current factors using the model trained by the model training module; The early warning alarm module is used to receive the evaluation result of the state evaluation module, and send out an early warning signal when the characteristic electrical quantity of the voltage transformer is abnormal, and send out an alarm signal when a voltage transformer fails; The prediction module is used to predict the future health state of the voltage transformer under the influence of the characteristic information of the leakage current, the secondary side voltage value and the partial discharge amount by using the model trained by the model training module; The data transmission module three is used to transmit the evaluation results of the state evaluation module and the prediction results of the prediction module to the data storage module.

2. A method for evaluating the state of a voltage transformer on board an electric locomotive or EMU, based on the system for evaluating the state of a voltage transformer on board an electric locomotive or EMU according to claim 1, characterized in that: The following steps are involved: S1. The data acquisition layer collects the secondary voltage U, leakage current I, phase θ and partial discharge characteristic information P of the on-board voltage transformer under the working condition monitored by the voltage sensor, current sensor and ultrasonic partial discharge sensor through the data acquisition module; S2. The characteristic information collected by S1 is sent to the data storage module of the data storage layer through the data transmission module 1 for storage, and then sent to the training module of the status evaluation layer through the data transmission module 2 to establish the health status evaluation LSTM-1 model and perform training; S3. The characteristic information collected by S1 is sent to the state evaluation module of the state evaluation layer through the data transmission module 1, and the health status evaluation LSTM-1 model trained by S2 is used to evaluate the health status of the on-board voltage transformer under the influence of the characteristic information of the leakage current, the secondary side voltage value and the partial discharge amount, and the evaluation result of the current health status of the on-board voltage transformer is obtained, and sent to the data storage module for storage through the data transmission module 3; The evaluation result of S4.S3 is transmitted to the alarm module for processing, and a warning signal is issued when the characteristic electrical quantity of the voltage transformer is abnormal, and an alarm signal is issued when a voltage transformer fails; S5. The historical data of the characteristic information of leakage current, secondary voltage value and partial discharge amount stored in the data storage module are sent to the prediction module of the state evaluation layer through the data transmission module 2, and the health state evaluation LSTM-2 model is established and trained; the future health state of the voltage transformer under the influence of the characteristic information of leakage current, secondary voltage value and partial discharge amount is predicted by using the trained LSTM-2 model, and the prediction result of the future health state of the on-board voltage transformer is obtained; S6. The data transmission module 3 of the state evaluation layer transmits the characteristic information of the evaluation result of S3 and the prediction result of S5 to the data storage module of the data storage layer for storage.

3. The method for evaluating the state of the on-board voltage transformer of an electric locomotive and a multiple unit according to claim 2 is characterized in that: The specific steps of S2 using the characteristic information data stored in the data storage module to establish the health status assessment LSTM-1 model and perform training are as follows: S2.1 Assume that the health assessment coefficient is H. By calculating and analyzing the health assessment coefficient, a quantitative determination of the health level is obtained. The calculation formula is: H n =aU n +bI n +cθ n +dt n +eP n +f (1) Where 1≤n≤k and it is an integer, H is the overall health assessment coefficient; U is the secondary voltage; I is the leakage current, the phase is θ, P is the partial discharge amount; f is the interference correction constant; According to the proportion of different signal characteristic values ​​in the running status evaluation, different weights a, b, c, d and e are assigned to them; Set the health critical standard to x, where x is an empirical parameter obtained from actual engineering or set according to actual conditions; when H>x, the operation state is fault operation; when H<x, the operation state is normal operation; S2.2 takes the measured data as the training set and trains it using the mean square error loss function; Among them, H i is the actual value, is the evaluation value; S2.3 divides time t into M1 intervals as large time intervals; each large time interval is divided into M2 intervals as small time intervals; the data of the large time interval is used as long-term memory data, and the data of the small time interval is used as short-term memory data; S2.4 establishes a long short-term memory neural network model LSTM-1 to evaluate the operating status of the voltage transformer and eliminate false alarms caused by temporary voltage fluctuations under working conditions; Under the S2.5 operating condition, the secondary voltage of the voltage transformer is relatively stable, and the secondary voltage is closely related to the operating time. In the establishment of the long-term and short-term memory neural network, the weight of the long-term memory data is strengthened to obtain the evaluation value based on the change of a large time interval. S2.

6. Under working conditions, the changes in leakage current and partial discharge reflect the status of the voltage transformer in real time, and their fluctuations are less affected by time. In the establishment of the long-short-term memory neural network, the weight of short-term memory data is strengthened to obtain the evaluation value based on the changes in small time intervals; S2.7 The collected actual data is divided into training data and verification data, of which 70% is used for training and 30% is used for verification; S2.8 divide the weight matrix into large time intervals and small time intervals; S2.9 Train and verify the long short-term memory neural network model LSTM-1.

4. The method for evaluating the state of the on-board voltage transformer of an electric locomotive and a multiple unit according to claim 3 is characterized in that: The state evaluation module in S3 uses the information feature weighted comparison method to weight the secondary side voltage, leakage current and partial discharge respectively to obtain the state evaluation coefficient H, and then compares it with the current actual state coefficient after ratio calculation with the judgment threshold, so as to evaluate the state of the voltage transformer. The specific steps are as follows: S3.1 Assume that the root mean square error calculation result within the M2 hour time interval is as follows: Among them, H i represents the i-th actual measurement value; represents the i-th evaluation result value; n represents a total of n sampling points in the small time interval; The conditions for determining magnetic field fluctuations are: |RMSE-RMSE train |>ε (4) Among them, RMSE train It represents the average value of RMSE of the two small time intervals measured in the training set; ε represents the accuracy, which is set according to engineering experience; S3.2 When voltage fluctuates |RMSE-RMSE train When |>ε is satisfied, the time interval is expanded to the large time interval that includes the small time interval, and its MAE value is calculated. The MAE calculation formula is as follows: Among them, H i represents the i-th actual measurement value; represents the i-th actual result value; N represents that there are N sampling points in this large time interval; The conditions for non-periodic fluctuation of magnetic field are: |MAE-MAE train |<δ (6) Among them, MAE train It represents the MAE value of adjacent large time intervals measured in the training set; δ represents the accuracy, which is set according to engineering experience; When H>x, the operation state is fault operation, and when H<x, the operation state is normal operation.

5. The method for evaluating the state of the on-board voltage transformer of an electric locomotive and a multiple unit according to claim 2 is characterized in that: The early warning alarm module in S4 has three output states: ① normal state: no alarm signal is issued, and the indicator light is green; ② abnormal state: a warning signal is issued, and the indicator light is yellow; ③ fault state: a fault alarm signal is issued, and the indicator light is red.

6. The method for evaluating the state of the on-board voltage transformer of an electric locomotive or EMU according to claim 2, characterized in that: The specific steps of S5 are: S5.1 Assume that the health assessment prediction coefficient is J. By calculating and analyzing the health assessment coefficient, a quantitative determination of health status is obtained. The calculation formula is: Where, 1≤n≤k and is an integer, J is the overall health assessment coefficient; U1—U n is the historical data of the secondary side voltage; I1—I n is the leakage current history data, θ1—θ n is the phase history data, P1-P n is the historical data of partial discharge; F is the interference correction constant; Different weights A, B, C, D and E are assigned to different signal feature values ​​according to their proportion in the running status evaluation; Set the health critical standard to X, where X is set based on the empirical parameters obtained from actual engineering or based on actual conditions; when J>X, the operation state is fault operation; when J<X, the operation state is normal operation; S5.2 takes the measured data as the training set and trains it using the mean square error loss function; Among them, J i is the actual value, is the predicted value; S5.3 divides time t into M1 intervals as large time intervals; each large time interval is divided into M2 intervals as small time intervals; the data of the large time interval is used as long-term memory data, and the data of the small time interval is used as short-term memory data; S5.4 establishes a long short-term memory neural network model LSTM-2 to predict the health status of the voltage transformer; Under the S5.5 operating condition, the secondary voltage of the voltage transformer is relatively stable, and the secondary voltage is closely related to the operating time. In the establishment of the long-term and short-term memory neural network, the weight of the long-term memory data is strengthened to obtain the predicted value based on the change of a large time interval; S5.

6. Under the operating condition, the changes of leakage current and partial discharge reflect the status of voltage transformer in real time, and their fluctuations are less affected by time. In the establishment of long short-term memory neural network, the weight of short-term memory data is strengthened to obtain the evaluation prediction value based on the change of small time interval; S5.7 The collected actual data is divided into training data and verification data, of which 70% is used for training and 30% is used for verification; S5.8 divide the weight matrix into large time intervals and small time intervals; S5.9 trains the long short-term memory neural network model LSTM-2 and verifies the prediction results.

Images