A method for predicting temperature rise of traction motor based on historical data

Abstract

The embodiment of the present invention discloses a method for predicting the temperature rise of a traction motor based on historical data, and the method comprises the following steps: acquiring the historical data of the traction motor, wherein the historical data of the traction motor is obtained by collecting the historical data of the traction motor within a period of time at the same frequency, and the historical data of the traction motor is stored as a field according to a type of sensor data; processing the historical data of the traction motor according to the required fields to obtain first temperature characteristic sequence data; performing time segmentation processing on the first temperature characteristic sequence data according to a preset time interval threshold to obtain second temperature characteristic sequence data; sampling the second temperature rise characteristic sequence data according to a set time window width and a sliding step size to obtain multiple groups of temperature characteristic data; dividing the multiple groups of temperature characteristic data to obtain training sample data and test sample data; and training and testing a neural network model using the training data and the test data to generate a target neural network model.

Description

Technical Field

[0001] The present invention relates to a method for predicting the temperature rise of a traction motor, and in particular to a method for predicting the temperature rise of a traction motor based on historical data. Background Art

[0002] With the innovation and development of social economy, equipment has become more efficient and intelligent, but the maintenance of equipment itself faces huge challenges. There are two traditional equipment maintenance methods, one is the post-failure maintenance method, and the other is the preventive maintenance method of periodic shutdown maintenance. These two traditional equipment maintenance methods will not only reduce productivity, but also cause certain economic losses to manufacturers, especially unplanned production interruptions.

[0003] In recent years, predictive maintenance has gradually gained favor among enterprises. It is developed from preventive maintenance and is a higher-level maintenance method. It is based on the equipment status and the predicted development trend of the equipment status. After information collection, processing and comprehensive analysis, maintenance cycles and maintenance items are purposefully arranged. "Repairs should be made when necessary and repairs must be done well." Compared with traditional maintenance methods, it has obvious advantages.

[0004] Motor predictive maintenance is to conduct regular (or continuous) status monitoring and fault diagnosis on the main parts of the motor when the motor is running, determine the state of the motor, predict the future development trend of the motor state, and formulate a predictive maintenance plan in advance based on the state development trend and possible failure mode of the motor to determine the time, content, method and necessary technical and material support for the motor to be repaired. Motor predictive maintenance integrates motor state monitoring, fault diagnosis, state prediction, maintenance decision support and maintenance activities, and is an emerging maintenance method.

[0005] Currently, in predictive maintenance of motors, there are usually two methods for motor temperature monitoring: one is physical modeling, and the other is data statistical analysis. The general problem with the physical modeling method is that the principle is complex and cannot be accurately modeled. Although the data statistical analysis method does not need to analyze the specific principle of the motor, it still has the problem of not being able to obtain universal and accurate results. Summary of the invention

[0006] The present invention provides a method for predicting the temperature rise of a traction motor based on historical data, which effectively solves the problems of difficulty in data analysis and insufficient prediction accuracy, and improves the accuracy of predicting the temperature rise of the traction motor.

[0007] An embodiment of the present invention provides a method for predicting the temperature rise of a traction motor based on historical data, comprising the following steps:

[0008] Acquire the historical data of the traction motor. The historical data of the traction motor is acquired by collecting the historical data of the traction motor within a period of time at the same frequency. The historical data of the traction motor includes: network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature. The historical data of the traction motor is stored as one field according to one type of sensor data;

[0009] Processing the traction motor historical data according to the required fields to obtain first temperature characteristic sequence data;

[0010] According to a preset time interval threshold, the first temperature characteristic sequence data is subjected to time segmentation processing to obtain second temperature characteristic sequence data;

[0011] The second temperature rise characteristic sequence data is sampled according to the set time window width and sliding step size to obtain multiple groups of temperature characteristic data;

[0012] Divide multiple groups of temperature feature data to obtain training sample data and test sample data;

[0013] Use the training data and test data to train and test the neural network model to generate the target neural network model.

[0014] Optionally, the same frequency is 10Hz.

[0015] Optionally, pandas is used to process the historical data of the traction locomotive.

[0016] Optionally, the required fields include at least one of: network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature.

[0017] Optionally, a transformation operation is performed on the multiple groups of temperature characteristic data, and the transformation operation includes: data standardization and EMD.

[0018] Optionally, the neural network model is a deep neural network model, and the deep neural network model includes:

[0019] The first feature extraction module, the second feature extraction module, the third feature extraction module and the fourth feature extraction module connected in sequence are used to extract features from the training data in sequence; wherein,

[0020] The first feature extraction module, the second feature extraction module, and the third feature extraction module are respectively composed of:

[0021] It consists of a fully connected unit FC, a linear rectified linear unit ReLU, a batch normalization unit BN, and a partially discarded neuron unit Dropout;

[0022] The fourth feature extraction module is composed of a fully connected unit FC, a batch normalization unit BN and a partial discard neuron unit Dropout. The fourth feature extraction module is used to output temperature prediction data.

[0023] As can be seen from the above content, a motor temperature rise prediction method based on historical data provided by an embodiment of the present invention includes: obtaining traction motor historical data, the traction motor historical data is obtained by collecting traction motor historical data within a period of time at the same frequency, the traction motor historical data includes: network current, U-phase current instantaneous value, W-phase current instantaneous value, U-phase current effective value, W-phase current effective value, motor speed, motor temperature, and the traction motor historical data is stored as a field according to a type of sensor data; the traction motor historical data is processed according to the required field to obtain first temperature characteristic sequence data; according to a preset time interval threshold, the first temperature characteristic sequence data is time segmented to obtain second temperature characteristic sequence data; the second temperature rise characteristic sequence data is sampled according to the set time window width and sliding step size to obtain multiple groups of temperature characteristic data; the multiple groups of temperature characteristic data are divided to obtain training sample data and test sample data; the neural network model is trained and tested using the training data and the test data to generate a target neural network model. .

[0024] By applying the embodiment of the present invention, based on the historical data of the traction motor, the temperature of the traction motor in the future time period is predicted. Based on the powerful feature characterization capability of the neural network, there is no need to perform complex modeling analysis, which effectively solves the problems of difficult data analysis and insufficient prediction accuracy. Of course, any product or method implementing the present invention does not necessarily need to achieve all the advantages described above at the same time.

[0025] The innovative features of the embodiments of the present invention include:

[0026] 1. Based on the historical data of the traction motor, the temperature data of the traction motor in the future time period is predicted, and the relationship between the historical sensor data of the traction motor and the temperature of the traction motor in the future time period is effectively constructed, which is one of the innovations of the embodiment of the present invention.

[0027] 2. Based on the neural network method, a traction motor temperature rise prediction model is constructed, and the historical data of the traction motor is used to train and fit the neural network model. For different traction motors, there is no need to conduct specific analysis of complex principles. Problems are solved through data rather than knowledge. In addition, based on the powerful feature representation ability of the neural network, the problems of difficult data analysis and insufficient prediction accuracy are effectively solved. This is one of the innovations of the embodiments of the present invention.

[0028] 3. The first three feature extraction modules connected in sequence in the neural network model are respectively composed of: a fully connected unit FC, a linear rectified linear unit ReLU, a batch normalization unit BN and a partial discarding neuron unit Dropout, which are used to effectively extract features from the training sample set in sequence. This design method can effectively improve the prediction accuracy of temperature data and is one of the innovations of the embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative work.

[0030] Figure 1 A schematic flow chart of a method for predicting motor temperature rise based on historical data is provided for an embodiment of the present invention;

[0031] Figure 2 A structural diagram of a deep neural network model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] It should be noted that the terms "including" and "having" and any variations thereof in the embodiments of the present invention and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0034] The present invention provides a method for predicting motor temperature rise based on historical data. The embodiments of the present invention are described in detail below.

[0035] Figure 1 A flow chart of a method for predicting motor temperature rise based on historical data provided by an embodiment of the present invention. The method may include the following steps:

[0036] S101: Acquire the historical data of the traction motor. The historical data of the traction motor is obtained by collecting the historical data of the traction motor within a period of time at the same frequency. The historical data of the traction motor includes: network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature. The historical data of the traction motor is stored as one field according to one type of sensor data.

[0037] Among them, for the train traction system, the coverage of sensors is already very comprehensive. The current, voltage, motor speed, motor temperature and other data of each branch in the system are measured and recorded by corresponding professional sensors, and a huge amount of data will be generated every moment.

[0038] The embodiment of the present invention adopts a neural network model to predict the temperature rise of the traction motor, and the training and testing of the model are based on the actual collected data. Various relevant historical data of the temperature rise of the traction motor are collected at a uniform frequency. There is no strict limit on the collection interval and the number of historical data collected here, but it can be understood that various historical data of the traction motor under different conditions should be collected as much as possible, and the number of samples should not be too small. In this embodiment, the network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature of the traction motor are sampled at a sampling frequency of 10Hz. The sampled data is recorded in a time series, and each type of sensor data is stored as a field.

[0039] Before each data preprocessing, the collected data is resampled at the same frequency and stored in a tdms (Technical Data Managerment Streaming) or csv (Comma-Separated Values) file. The high-frequency sampling data is stored in a tmds file, and the low-frequency sampling data is stored in a csv file. Using different frequency data to train the neural network model can further optimize the target neural network model.

[0040] S102: Processing the traction motor historical data according to required fields to obtain first temperature characteristic sequence data.

[0041] In this embodiment, pandas is used to process the historical data of the traction locomotive. Pandas is a data analysis package of Python, that is, a data analysis and processing tool that needs to read files, process data (display certain row data or indexes, operations, statistical operations), and write files to complete data processing. The above tdms or csv file is read into pandas, the required fields are set, the required field columns are selected, and the first temperature characteristic sequence data is obtained.

[0042] The required fields include network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature. All or part of the fields in the traction motor historical data include at least one of the above fields. In this embodiment, all of the above fields are selected as the required fields.

[0043] S103: performing time segment processing on the first temperature characteristic sequence data according to a preset time interval threshold to obtain second temperature characteristic sequence data.

[0044] When the adjacent time intervals exceed the preset time interval, they are considered to be data of different time periods. The data of different time periods are assumed to be too long to be suitable for using the data of a certain time period to predict the temperature rise data of the traction motor outside the time period.

[0045] In this embodiment, after the first temperature characteristic sequence data is segmented according to the preset time interval, the time segment information of the first temperature characteristic sequence data can be obtained, that is, the second temperature characteristic sequence data. The time segment information can be used to sample the second temperature characteristic sequence data through a sliding window in the following text, and multiple groups of temperature characteristic data can be sampled through the sliding window as sample data for model training.

[0046] S104: Sampling the second temperature rise characteristic sequence data according to the set time window width and sliding step size to obtain multiple groups of temperature characteristic data.

[0047] In this embodiment, in order to obtain more sample points, the preset window sliding step size can be 1, that is, corresponding to 0.1 seconds, and the time window width can be 320 steps to obtain a feature matrix. Among them, the setting of the window sliding step size and the time window width can be designed according to actual conditions, and this embodiment does not specifically limit this.

[0048] The elements in the same row of the feature matrix represent data at the same time point. For example, the first row represents the temperature feature data collected in 0.1 seconds. The elements in each column represent each type of temperature feature data. In the present embodiment, the elements in each column are network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature.

[0049] Furthermore, a transformation operation is performed on the feature matrix, and the transformation operation includes: data standardization and EMD.

[0050] In order to make the model prediction results more accurate, the EMD (Empirical Mode Decomposition) method is used to further transform the feature matrix. This method can decompose the data into high and low frequency parts, so that the distribution of different frequencies can be predicted separately, which can make the model prediction results more accurate, especially in areas with large high-frequency components.

[0051] After frequency decomposition, the data is normalized, that is, data standardization, which maps all data in the feature matrix to values ​​in the range of 0-1 that conform to the standard normal distribution, so as to effectively improve the prediction effect of the model.

[0052] The transformed feature matrix is ​​sorted in chronological order, and the temperature feature data in the first time period is selected as the basic data, and the temperature data in the second time period is selected as the data to be predicted; the basic data and the data to be predicted are used to train the neural network model.

[0053] In this embodiment, the temperature feature data of the first 160 steps are selected from the feature matrix as the neural network model input data features, and the temperature data of the last 160 steps are selected as the neural network model output data labels. The fields of the input data include network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature, and the field of the output data is motor temperature.

[0054] S105: Divide the multiple groups of temperature characteristic data to obtain training sample data and test sample data;

[0055] S106: Use the training data and the test data to train and test the neural network model to generate a target neural network model.

[0056] Among them, the neural network model is a deep neural network model. Figure 2 A structural diagram of a deep neural network model provided by an embodiment of the present invention, such as Figure 2 As shown, the deep neural network model includes:

[0057] The first feature extraction module, the second feature extraction module, the third feature extraction module and the fourth feature extraction module connected in sequence are used to extract features from the training data in sequence, such as Figure 2 As shown, by continuously performing feature extraction, the dimensions of the input data are reduced to 128, 64, and 32 respectively, until the temperature prediction data is finally output.

[0058] Among them, the first feature extraction module, the second feature extraction module, and the third feature extraction module are respectively composed of:

[0059] It consists of a fully connected unit FC, a linear rectified linear unit ReLU, a batch normalization unit BN, and a partially discarded neuron unit Dropout;

[0060] The fourth feature extraction module is composed of a fully connected unit FC, a batch normalization unit BN and a partial discard neuron unit Dropout, and the fourth feature extraction module is used to output temperature prediction data.

[0061] During the training process, the neural network modeling will perform forward propagation and back propagation, continuously updating the parameters of the neural network model, so that the value of the loss function of the neural network model gradually decreases, until the value of the loss function reaches convergence, and the last updated parameters are used as the parameters of the neural network model. The model is applied to the temperature rise prediction of the traction motor and the prediction data is obtained.

[0062] The technical solution provided by the embodiment of the present invention predicts the temperature rise of the traction motor based on the neural network method, which can effectively solve the problems of difficulty in predictive data analysis and insufficient prediction accuracy in the prior art. By using the historical temperature data of the traction motor for model training, there is no need to perform complex modeling analysis, and the prediction of the traction motor temperature at future moments is realized, solving the problem of modeling difficulties.

[0063] The above-mentioned system and device embodiments correspond to the system embodiments and have the same technical effects as the method embodiments. For specific descriptions, please refer to the method embodiments. The device embodiments are obtained based on the method embodiments. For specific descriptions, please refer to the method embodiment part, which will not be repeated here. It can be understood by those of ordinary skill in the art that the accompanying drawings are only schematic diagrams of one embodiment, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present invention.

[0064] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the temperature rise of a traction motor based on historical data, characterized in that: The following steps are involved: Acquire the traction motor historical data, wherein the traction motor historical data is acquired by collecting the traction motor historical data within a period of time at the same frequency, and the traction motor historical data includes: network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature, and the traction motor historical data is stored as one field according to one type of sensor data; Processing the traction motor historical data according to the required fields to obtain first temperature characteristic sequence data; According to a preset time interval threshold, the first temperature characteristic sequence data is subjected to time segmentation processing to obtain second temperature characteristic sequence data; Sampling the second temperature characteristic sequence data according to the set time window width and sliding step size to obtain multiple groups of temperature characteristic data; Dividing the multiple groups of temperature characteristic data to obtain training sample data and test sample data; The training sample data and the test sample data are used to train and test the neural network model to generate a target neural network model.

2. The method according to claim 1, characterized in that Said same frequency is 10 Hz.

3. The method according to claim 1, characterized in that Pandas is used to process the traction motor historical data.

4. The method according to claim 1, characterized in that The required fields include: at least one of network current, instantaneous value of U-phase current, instantaneous value of W-phase current, effective value of U-phase current, effective value of W-phase current, motor speed, and motor temperature.

5. The method according to claim 1, characterized in that A transformation operation is performed on the multiple groups of temperature characteristic data, wherein the transformation operation includes: data standardization and EMD.

6. The method according to claim 1, characterized in that The neural network model is a deep neural network model, and the deep neural network model includes: The first feature extraction module, the second feature extraction module, the third feature extraction module and the fourth feature extraction module connected in sequence are used to extract features from the training sample data in sequence; wherein, The first feature extraction module, the second feature extraction module, and the third feature extraction module are respectively composed of: It consists of a fully connected unit FC, a linear rectified linear unit ReLU, a batch normalization unit BN, and a partially discarded neuron unit Dropout; The fourth feature extraction module is composed of a fully connected unit FC, a batch normalization unit BN and a partial discard neuron unit Dropout, and the fourth feature extraction module is used to output temperature prediction data.

Images