Method and system for predicting corona onset voltage of motor stator winding, electronic equipment and computer readable storage medium

Abstract

The invention discloses a motor stator winding corona onset voltage prediction method, a motor stator winding corona onset voltage prediction system, electronic equipment and a computer readable storage medium, and belongs to the technical field of motors. The size of each element in a data set obtained after processing is linearly transformed to 0-1, data dimensions can be removed, the situation that data distribution is in a flat state due to the fact that part of sample data is too large or too small is avoided, the feature recognition degree of the data is improved, data feature extraction of the convolutional neural network is facilitated, and the accuracy of data feature extraction is improved. And the calculation speed and accuracy of the prediction model are improved.

Description

Technical Field

[0001] This invention relates to the field of motor technology, and specifically to a method, system, electronic device, and computer-readable storage medium for predicting the corona initiation voltage of a motor stator winding. Background Technology

[0002] Corona discharge itself has a low initiation voltage and may occur at multiple weak points in the insulation of the stator windings, such as the stator slot, slot outlet, and the "R" angle of the conductor bars. If left unchecked for a long time, it may cause insulation layer damage and reduce insulation performance, leading to serious consequences such as winding short circuits and insulation breakdown. Therefore, it is necessary to achieve reliable prediction of corona discharge in motor stator windings to ensure the safe and stable operation of the motor.

[0003] Secondly, temperature, humidity, and air pressure all affect the corona initiation voltage, and since all three are atmospheric factors, they also have complex interrelationships. Specifically, temperature, as an atmospheric parameter, affects both air humidity and air pressure, indirectly causing changes in the corona initiation voltage. Furthermore, temperature affects the negative temperature coefficient insulation material on the stator winding surface, directly leading to changes in the corona initiation voltage. The combined effects of temperature, humidity, and air pressure make predicting the corona initiation voltage complex and difficult, necessitating an effective method for reliable prediction. This invention proposes an effective prediction method to reliably predict the corona initiation voltage of the stator windings of large motors operating in high-altitude areas.

[0004] It should be noted that, considering the uncertain impact of corona discharge on the environment, the corona discharge prediction mentioned in this invention is only applicable before the stator winding corona discharge occurs, and is not applicable to the retrospective estimation of the corona initiation voltage after the corona discharge. Summary of the Invention

[0005] This invention aims to solve the problems existing in the prior art and provides a method, system, electronic device, and computer-readable storage medium for predicting the corona initiation voltage of motor stator windings. The method preprocesses the corona initiation voltage at temperature, humidity, air pressure, stator slots, stator slot outlets, and "R" corners. After processing, the size of each element in the resulting dataset is linearly transformed to between 0 and 1, which can remove the data dimensions and avoid some sample data being too large or too small, resulting in a flat data distribution. This improves the feature recognition of the data, which is beneficial for the data feature extraction of convolutional neural networks, and improves the calculation speed and accuracy of the prediction model.

[0006] The objective of this invention is achieved through the following technical solution: In a first aspect, a method for predicting the corona initiation voltage of a motor stator winding includes the following steps: Step 1: Obtain multidimensional data from the motor stator winding environment; Step 2: Based on the preset standardization function and partitioning ratio, preprocess the multidimensional data and divide it into training set and prediction set; Step 3: Input the resulting training set into the convolutional neural network and train the convolutional neural network. Step 4: Input the prediction set into the prediction model to obtain the predicted corona initiation voltage of the target motor stator winding.

[0007] Preferably, in step one, the multidimensional data includes temperature, humidity, air pressure, and the corona voltage at the stator slot, stator outlet, and "R" corner corresponding to the environmental factors.

[0008] Preferably, in step two, the temperature, humidity, air pressure, and the corona voltage at the stator slot, stator outlet, and "R" corner corresponding to the environmental factors are preprocessed, and the size of each element in the resulting dataset is linearly transformed to between 0 and 1.

[0009] Preferably, in step one, the multidimensional data includes environmental factors, corona initiation voltage, and corona initiation voltage of the same motor electronic winding at different time points within a preset time period to predict the possible corona discharge of the motor stator winding in time sequence; or environmental factors, corona initiation voltage, and corona initiation voltage of the same motor electronic winding located in different environments at the same time to predict the corona initiation voltage of the stator windings of multiple motors.

[0010] Preferably, in step two, the multidimensional data obtained from the motor stator winding environment should be divided into two parts according to the prediction requirements: the first part is the environmental factor data from the target motor and does not include corona initiation voltage data, with each sample being a 3*1 data matrix; the second part is the data from the sample database and simulation experiment data, which includes both corona initiation voltage data and environmental factor data, with each sample being a 6*1 data matrix; when dividing the preprocessed multidimensional data, the first part is used as the prediction set and the second part is used as the training set.

[0011] Preferably, in step three, based on the possible ranges of learning rate and random discard rate, all possible parameter combinations that may optimize model performance are listed and the corresponding convolutional neural network model is constructed.

[0012] Preferably, step three includes the following steps: Step 31: Treat each ant as a convolutional neural network parameter configuration scheme, and treat all possible parameters of the convolutional neural network as nodes of the ant colony algorithm; initialize the ant colony, so that the ants are randomly distributed on the nodes, initialize the pheromone matrix so that the pheromone between each node is equal, set the maximum number of iterations, and start the iteration. Step 32: Ants choose paths based on probability, determined by pheromone concentration and the attractiveness of each path. The transfer probability calculation function for ants at nodes is shown below: Ants at the node Upward node The transition probability is defined as: ; In the formula: This indicates that the ant is located at the node. The set of path points that can be traversed in the next step; The relative importance of the trajectory reflects the role of the information accumulated by the ant during its movement. The relative importance of visibility reflects the degree to which heuristic factors are valued in ants' path selection during movement. For the ant located at the node Node The intensity of pheromones along the path, The attractiveness of each path to the ant can be defined as a performance function of the ant. t This indicates the current iteration number of the algorithm; The transition probability function For performance functions, the ant's position on the node Any reachable node It is defined as follows: ; In the formula, , and These are the true values ​​of the validation set; , and These are the predicted values ​​output by the prediction model; Step 33: After all the ants have formed a complete path, assign values ​​to the parameters of the convolutional neural network based on the movement paths of the ants. Step 34: Train the convolutional neural network using the training set; the convolutional neural network structure contains 6 layers connected end to end, namely the input layer, the first convolutional layer, the pooling layer, the second convolutional layer, the fully connected layer, and the regression layer; Step 35: The regression layer is responsible for outputting the prediction results of the convolutional neural network.

[0013] Preferably, in step 35, the input layer is responsible for inputting the input parameters into the model, and the input parameters are: ; In the formula, Representing the One input parameter; The first and second convolutional layers have the same structure and obtain data features through convolution. Their kernel size is 3*1 and their stride is 1. The output after this convolution is represented as follows: ; In the formula, Indicates the first One output unit For this convolutional layer, the input unit is... This represents the convolution kernel of the convolutional layer. For activation functions; The pooling layer is responsible for downsampling the output features of the convolutional layer to avoid the curse of dimensionality and overfitting. The downsampled output is: ; In the formula, For the pooling layer One output unit for The downsampling coefficients of the function, in addition The function uses average pooling. The fully connected layer is connected to all nodes of the second convolutional layer, and the output expression is: ; In the formula, For the fully connected layer One output unit For the first The convolutional layer neural unit and the first The connection weights between fully connected layer neural units, The first convolutional layer One output unit Indicates the first The threshold of a neural unit; Random dropout should also be performed after the fully connected layer, and its function expression is: ; In the formula, The probability that each neuron is preserved.

[0014] Preferably, the method further includes step 36, which involves calculating the performance function of each path in the current iteration and comparing it with the best performance function from the previous iteration to obtain the path with the highest performance function. Step 37, update the pheromones along the path. The formula for updating pheromone intensity is shown below: For ants at nodes Any reachable node It is defined as follows: ; In the formula, The dissipation coefficient of the pheromone. For the first Only one ant is on the path in this loop. The pheromone material left on the surface per unit length t Represents the number of iterations. t+k Representative at the t After the first iteration, and then after... k After the nth iteration, the total number of iterations is now; The formula is defined as follows: ; In the formula, The calculation formula is: ; in, F It is a constant; Q i For ants l In the path ij The performance function for moving upwards; Step 38: Verify whether the maximum number of iterations has been reached. If the optimal number of iterations has been reached, output the parameter combination represented by the optimal path. Step 39: Calculate the model's coefficient of determination R2, root mean square error (RMSE), and mean absolute percentage error (MAPE) using the true values ​​of the training set and the predicted values ​​output by the prediction model based on the training set. These are used as evaluation indicators of the model's predictive performance. The smaller the RMSE and mean absolute percentage error, and the closer the coefficient of determination is to 1, the better the model's predictive performance.

[0015] Preferably, in step 39, the coefficient of determination R 2 The formulas for defining the root mean square error (RMSE) and the mean absolute percentage error (MAPE) are as follows: ; ; ; In the formula, For predicted values, For the true value, This is the average value. This indicates the number of samples in the training set.

[0016] Secondly, a system for predicting the corona initiation voltage of a motor stator winding includes: The detection module is used to acquire environmental data of the stator winding of the target motor; The data processing module preprocesses the multidimensional data according to a preset standardization function and divides it into a training set and a prediction set. The model training module is used to input the formed training set into the convolutional neural network, train the convolutional neural network through the ant colony optimization algorithm, obtain the optimal combination of learning rate and random drop-off rate of the convolutional neural network, and form the final prediction model. The performance evaluation module is used to calculate the model's coefficient of determination, root mean square error, and mean absolute percentage error based on the true values ​​of the training set and the predicted values ​​output by the prediction model from the training set, thereby evaluating the model's predictive performance. The voltage prediction module is used to calculate the target motor stator winding environment dataset based on the optimized and trained prediction model, and output the prediction data.

[0017] Preferably, the data processing module includes: The first data processing submodule is used to standardize the feature sample data and the prediction target data according to the standardization equation in the above prediction method, and remove the dimensions. The second data processing submodule is used to sort the standardized data by category to obtain a dataset that conforms to the model input format corresponding to the detection data and sample data. The third data processing submodule is used to divide the sample feature dataset into a training set and a prediction set according to a preset ratio.

[0018] Preferably, the ant colony optimization module includes: The first ant colony optimization submodule is used to initialize the data and assign values ​​to the number of optimization nodes and the number of ants according to the number of training set groups and the number of model parameter groups. The second ant colony optimization submodule is used to calculate the probability based on the probability transfer formula of ants on nodes in the above prediction method, and guide the ants to transfer. The third ant colony optimization submodule is used to update the pheromone intensity on the path according to the pheromone intensity adjustment formula of the ants in the above prediction method. The fourth ant colony optimization submodule is used to determine whether the maximum number of iterations has been reached and to output the optimal parameters.

[0019] Thirdly, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described prediction method.

[0020] Fourthly, a computer-readable storage medium stores computer instructions that cause a computer to perform the aforementioned prediction method.

[0021] The beneficial effects of this technical solution are as follows: I. The present invention provides a method for predicting the corona initiation voltage of motor stator windings. The method preprocesses the temperature, humidity, air pressure, stator slots, stator slot outlets, and corona initiation voltage at the "R" corner. After processing, the size of each element in the resulting dataset is linearly transformed to between 0 and 1, which can remove the data dimensions and avoid some sample data being too large or too small, resulting in a flat data distribution. This improves the feature recognition of the data, which is beneficial for the data feature extraction of convolutional neural networks, and improves the calculation speed and accuracy of the prediction model. II. The present invention provides a method for predicting the corona initiation voltage of a motor stator winding. It performs prediction based on multi-dimensional data including temperature, humidity, and environmental conditions, taking into account the complex operating conditions of the motor at high altitudes, thus improving the accuracy of the prediction method. Furthermore, leveraging the automatic feature extraction capabilities of convolutional neural networks, this invention employs a convolutional neural network to solve for the complex relationships between multi-dimensional data, making the prediction method comprehensive and scientific. Finally, due to the excellent global optimization performance of the ant colony optimization algorithm, this invention uses the ant colony algorithm to optimize the parameter selection of the convolutional neural network to form the final prediction model, further improving the accuracy of the prediction method.

[0022] III. This invention provides a method for predicting the corona initiation voltage of a motor stator winding. The convolutional neural network improved using the ant colony optimization algorithm exhibits stronger generalization ability and robustness, effectively avoiding overfitting and demonstrating strong generalization capability. This allows it to reduce reliance on prior knowledge while maintaining reliable predictive performance when facing new environmental data outside the training set. Traditional physical modeling requires deep knowledge of discharge theory and numerous assumptions. The CNN-ACO method, however, is more like a data-driven "black box" model. It learns patterns through data, reducing reliance on prior knowledge of complex physical mechanisms, making it particularly suitable for complex phenomena whose mechanisms are not yet fully understood. Therefore, this method is far more efficient than manual trial and error or grid search, obtaining a high-performance predictive model at an acceptable time cost, significantly improving R&D and prediction efficiency in the long run. Attached Figure Description

[0023] Figure 1 This is a flowchart of the prediction method in this invention; Figure 2 This is a schematic diagram of the connection of the prediction system in this invention; Figure 3 This is a schematic diagram of the connection of the electronic device in this invention. Detailed Implementation The present invention will be further described in detail below with reference to embodiments, but the implementation of the present invention is not limited thereto.

[0024] Example 1 like Figure 1As shown, in a first aspect, this embodiment provides a method for predicting the corona initiation voltage of a motor stator winding, comprising the following steps: Step 1: Obtain multidimensional data from the motor stator winding environment; Step 2: Based on the preset standardization function and partitioning ratio, preprocess the multidimensional data and divide it into training set and prediction set; Step 3: Input the resulting training set into the convolutional neural network and train the convolutional neural network. Step 4: Input the prediction set into the prediction model to obtain the predicted corona initiation voltage of the target motor stator winding.

[0025] Example 2 This embodiment provides a method for predicting the corona initiation voltage of a motor stator winding, including the following steps: Step 1: Obtain multidimensional data from the motor stator winding environment; Step 2: Based on the preset standardization function and partitioning ratio, preprocess the multidimensional data and divide it into training set and prediction set; Step 3: Input the resulting training set into the convolutional neural network and train the convolutional neural network. Step 4: Input the prediction set into the prediction model to obtain the predicted corona initiation voltage of the target motor stator winding.

[0026] In step one, the multidimensional data includes temperature, humidity, air pressure, and the corona voltage at the stator slot, stator outlet, and "R" corner corresponding to the environmental factors.

[0027] In step two, the temperature, humidity, air pressure, and the corona voltage at the stator slot, stator outlet, and "R" corner corresponding to these environmental factors are preprocessed. After processing, the size of each element in the resulting dataset is linearly transformed to between 0 and 1.

[0028] In step one, the multidimensional data includes environmental factors, corona initiation voltage, and corona initiation voltage of the same motor electronic winding at different time points within a preset time period to predict the possible corona discharge of the motor stator winding in time sequence; or environmental factors, corona initiation voltage, and corona initiation voltage of the same motor electronic winding located in different environments at the same time to predict the corona initiation voltage of the stator windings of multiple motors.

[0029] In step two, the multidimensional data obtained from the motor stator winding environment should be divided into two parts according to the prediction requirements: the first part is the environmental factor data from the target motor and does not include corona initiation voltage data, with each sample being a 3*1 data matrix; the second part is the data from the sample database and simulation experiment data, which includes both corona initiation voltage data and environmental factor data, with each sample being a 6*1 data matrix. When dividing the preprocessed multidimensional data, the first part is used as the prediction set and the second part is used as the training set.

[0030] In step three, based on the possible values ​​of the learning rate and random discard rate, all possible parameter combinations that may optimize model performance are listed and the corresponding convolutional neural network model is constructed.

[0031] Step three includes the following steps: Step 31: Treat each ant as a convolutional neural network parameter configuration scheme, and treat all possible parameters of the convolutional neural network as nodes of the ant colony algorithm; initialize the ant colony, so that the ants are randomly distributed on the nodes, initialize the pheromone matrix so that the pheromone between each node is equal, set the maximum number of iterations, and start the iteration. Step 32: Ants choose paths based on probability, determined by pheromone concentration and the attractiveness of each path. The transfer probability calculation function for ants at nodes is shown below: Ants at the node Upward node The transition probability is defined as: ; In the formula: This indicates that the ant is located at the node. The set of path points that can be traversed in the next step; The relative importance of the trajectory reflects the role of the information accumulated by the ant during its movement. The relative importance of visibility reflects the degree to which heuristic factors are valued in ants' path selection during movement. For the ant located at the node Node The intensity of pheromones along the path, The attractiveness of each path to the ant can be defined as a performance function of the ant. t This indicates the current iteration number of the algorithm; The transition probability function For performance functions, the ant's position on the node Any reachable node It is defined as follows: ; In the formula, , and These are the true values ​​of the validation set; , and These are the predicted values ​​output by the prediction model; Step 33: After all the ants have formed a complete path, assign values ​​to the parameters of the convolutional neural network based on the movement paths of the ants. Step 34: Train the convolutional neural network using the training set; the convolutional neural network structure contains 6 layers connected end to end, namely the input layer, the first convolutional layer, the pooling layer, the second convolutional layer, the fully connected layer, and the regression layer; Step 35: The regression layer is responsible for outputting the prediction results of the convolutional neural network.

[0032] In step 35, the input layer is responsible for inputting the input parameters into the model. The input parameters are: ; In the formula, Representing the One input parameter; The first and second convolutional layers have the same structure and obtain data features through convolution. Their kernel size is 3*1 and their stride is 1. The output after this convolution is represented as follows: ; In the formula, Indicates the first One output unit For this convolutional layer, the input unit is... This represents the convolution kernel of the convolutional layer. For activation functions; The pooling layer is responsible for downsampling the output features of the convolutional layer to avoid the curse of dimensionality and overfitting. The downsampled output is: ; In the formula, For the pooling layer One output unit for The downsampling coefficients of the function, in addition The function uses average pooling. The fully connected layer is connected to all nodes of the second convolutional layer, and the output expression is: ; In the formula, For the fully connected layer One output unit For the first The convolutional layer neural unit and the first The connection weights between fully connected layer neural units, The first convolutional layer One output unit Indicates the first The threshold of a neural unit; Random dropout should also be performed after the fully connected layer, and its function expression is: ; In the formula, The probability that each neuron is preserved.

[0033] This also includes step 36, which calculates the performance function of each path in the current iteration and compares it with the best performance function of the previous iteration to obtain the path with the largest performance function; Step 37, update the pheromones along the path. The formula for updating pheromone intensity is shown below: For ants at nodes Any reachable node It is defined as follows: ; In the formula, The dissipation coefficient of the pheromone. For the first Only one ant is on the path in this loop. The pheromone material left on the surface per unit length t Represents the number of iterations. t+k Representative at the t After the first iteration, and then after... k After the nth iteration, the total number of iterations is now; The formula is defined as follows: ; In the formula, The calculation formula is: ; in, F It is a constant; Q i For ants l In the path ij The performance function for moving upwards; Step 38: Verify whether the maximum number of iterations has been reached. If the optimal number of iterations has been reached, output the parameter combination represented by the optimal path. Step 39: Calculate the model's coefficient of determination R2, root mean square error (RMSE), and mean absolute percentage error (MAPE) using the true values ​​of the training set and the predicted values ​​output by the prediction model based on the training set. These are used as evaluation indicators of the model's predictive performance. The smaller the RMSE and mean absolute percentage error, and the closer the coefficient of determination is to 1, the better the model's predictive performance.

[0034] In step 39, the coefficient of determination R 2 The formulas for defining the root mean square error (RMSE) and the mean absolute percentage error (MAPE) are as follows: ; ; ; In the formula, For predicted values, For the true value, This is the average value. This indicates the number of samples in the training set.

[0035] Example 3 This embodiment provides a method for predicting the corona initiation voltage of a motor stator winding, including the following steps: Step 1: Obtain multi-dimensional data of the motor stator winding environment. The multi-dimensional data includes temperature, humidity and air pressure, as well as the corona voltage at the stator slot, stator outlet and "R" corner corresponding to the environmental factors.

[0036] Optional, depending on the needs of the prediction target, the above multidimensional data can be the environmental factors and corona initiation voltage at different time points within a preset time period for the same motor electronic winding, in order to predict the corona initiation voltage of the stator winding of the motor that may cause corona discharge in time sequence; or it can be the environmental factors and corona initiation voltage of the stator windings of motors located in different environments at the same time, in order to predict the corona initiation voltage of the stator windings of multiple motors.

[0037] Step 2: Based on the preset standardization function and partitioning ratio, preprocess the multidimensional data and divide it into training set and prediction set; Temperature, humidity, and air pressure, as well as the corona initiation voltage at the stator slot, stator outlet, and "R" corner corresponding to these environmental factors, are preprocessed. After processing, the size of each element in the resulting dataset is linearly transformed to between 0 and 1. The advantages are: it can remove the dimensions of the data and avoid some sample data being too large or too small, which would cause the data distribution to be flat, improve the feature recognition of the data, facilitate the data feature extraction of convolutional neural networks, and make the prediction model more reliable.

[0038] It is important to note that, based on the prediction requirements, the acquired multidimensional data from the motor stator winding environment should be divided into two parts: the first part consists of environmental factor data from the target motor, excluding corona initiation voltage data, with each sample being a 3x1 data matrix; the second part consists of data from the sample database and simulation experiments, including both corona initiation voltage and environmental factor data, with each sample being a 6x1 data matrix. When dividing the preprocessed multidimensional data, the first part is used as the prediction set, and the second part is used as the training set. For example, if the first part contains 30 sets of 3x1 multidimensional data and the second part contains 200 sets of 6x1 multidimensional data, the resulting prediction set will contain 30 sets of 3x1 multidimensional data.

[0039] Step 3: Input the formed training set into the convolutional neural network, use the ant colony optimization algorithm to select the convolutional neural network parameters, complete the training of the convolutional neural network, and form the final prediction model. Based on the possible ranges of learning rate and random discard rate, list all possible parameter combinations that may optimize model performance and construct the corresponding convolutional neural network model.

[0040] It's important to note that convolutional neural networks (CNNs) are black-box models; they automatically perform feature finding after receiving sample data. Ant colony optimization (ACO) can be viewed as a feedback mechanism within the CNN to train the model. Its principle lies in mimicking the behavior of ants searching for the shortest path while foraging.

[0041] In some feasible embodiments, to further ensure the reliability of the prediction results, the input parameters of the convolutional neural network are not limited to the environmental factors mentioned above, including temperature, humidity, and air pressure. More multidimensional data can also be considered to extract data features and complete the prediction of the motor's corona initiation voltage. For example, the input time series 09-30-23 represents the month, date, and hour, respectively. Different months, dates, and hours result in different weather conditions and environmental factors affecting motor insulation in the same region, which can serve as feature samples for the prediction model. Similarly, the input spatial features of latitude and longitude, which are highly correlated with specific environmental factors, can also be used as feature samples for the prediction model.

[0042] In this implementation, the convolutional neural network structure consists of six interconnected layers: an input layer, a first convolutional layer, a pooling layer, a second convolutional layer, a fully connected layer, and a regression layer. The ant colony optimization algorithm is used as the feedback mechanism of the convolutional neural network to select the optimal combination of parameters.

[0043] Furthermore, each ant is considered as a convolutional neural network parameter configuration scheme, and all possible parameters of the convolutional neural network are considered as nodes of the ant colony algorithm; the ant colony is initialized, and the ants are randomly distributed on the nodes; the pheromone matrix is ​​initialized, so that the pheromone between each node is equal; the maximum number of iterations is set, and the iteration begins. For example, based on experience, the optimal learning rate of the model appears in the range of 0.0001-0.0010, containing 10 points, and the optimal random discard rate appears in the range of 0.1-0.5, containing 5 points. Therefore, the ant colony algorithm has 15 nodes, and the number of ants is initialized to 25. The pheromone intensity matrix is ​​initialized, and the pheromone concentration of all nodes is set to an equal value, such as an initial value of 1.

[0044] Furthermore, ants choose paths based on probability, which is determined by the pheromone concentration and the attractiveness of each path. The function for calculating the ant's transition probability at a node is shown below: Ants at the node Upward node The transition probability is defined as: ; In the formula: This indicates that the ant is located at the node. The set of path points that can be traversed in the next step; The relative importance of the trajectory reflects the role of the information accumulated by the ant during its movement. The relative importance of visibility reflects the degree to which heuristic factors are valued in ants' path selection during movement. For the ant located at the node Node The intensity of pheromones along the path, The attractiveness of each path to the ant can be defined as a performance function of the ant. t This indicates the current iteration number of the algorithm; The transition probability function For performance functions, the ant's position on the node Any reachable node It is defined as follows: ; In the formula, , and These are the true values ​​of the validation set; , and These are the predicted values ​​output by the prediction model; Specifically, during the path selection phase, each ant travels from one node to another based on the probabilities calculated above, ultimately forming a complete path. This probability is determined by pheromone concentration. And the attractiveness of each path to the ants It is defined by [the relevant authority / method]. This indicates that the ant is at the node. The set of nodes that can be selected at any time.

[0045] For example, in a certain iteration, if an ant is located on the node representing a learning rate of 0.0002, then all nodes representing learning rates will be excluded. In the dataset, the only nodes available for the ant to choose from are those representing random discard rates. In this example, the nodes represent only five random discard rates: 0.1, 0.2, 0.3, 0.4, and 0.5. This embodiment optimizes the learning rate and random discard rate of the convolutional neural network. Therefore, when the ant chooses another node, a complete path is formed. In this embodiment, this complete path has five possible combinations. Before the ant selects a node representing the random discard rate, the convolutional neural network is trained and predicted based on the parameters of the above five combinations and multidimensional data, and the performance function is calculated. , , , and Based on the pheromone intensities of the five possible paths obtained in the previous iteration, five transition probabilities can be calculated. , , , and The ants will choose the next node according to this probability and leave pheromones as they pass through, forming a complete path.

[0046] Furthermore, after all the ants have formed a complete path, the parameters of the convolutional neural network are assigned values ​​based on the ants' movement paths; the input layer of the aforementioned convolutional neural network is responsible for inputting the input parameters into the model, and the input parameters are: ; In the formula, Representing the One input parameter; This prediction model comprehensively considers the environmental factors of the motor. In a representative example, the training set formed by obtaining multidimensional data contains 180 sets of 6*1 data, so the input parameters are a data matrix of 180*6.

[0047] The first and second convolutional layers have the same structure and obtain data features through convolution. Their kernel size is 3*1 and their stride is 1. The output after this convolution is represented as follows: ; In the formula, Indicates the first One output unit For this convolutional layer, the input unit is... This represents the convolution kernel of the convolutional layer. For activation functions; The pooling layer is responsible for downsampling the output features of the convolutional layer to avoid the curse of dimensionality and overfitting. The downsampled output is: ; In the formula, For the pooling layer One output unit for The downsampling coefficients of the function, in addition The function uses average pooling. The fully connected layer is connected to all nodes of the second convolutional layer, and the output expression is: ; In the formula, For the fully connected layer One output unit For the first The convolutional layer neural unit and the first The connection weights between fully connected layer neural units, The first convolutional layer One output unit Indicates the first The threshold of a neural unit; Random dropout should also be performed after the fully connected layer, and its function expression is: ; In the formula, The probability that each neuron is preserved.

[0048] Furthermore, the regression layer is responsible for outputting the prediction results of the convolutional neural network. In a representative example, if the prediction set formed by obtaining multidimensional data contains 30 sets of 3*1 data matrices, then the final regression layer outputs a 30*3 data matrix.

[0049] Calculate the performance function of each path in the current iteration and compare it with the best performance function from the previous iteration. Find the path with the highest performance function. Update this path as the best path for the current iteration, and update its best performance function to the best performance function for the current iteration.

[0050] Furthermore, the pheromones along the path are updated, and the formula for updating the pheromone intensity is as follows: For ants at nodes Any reachable node It is defined as follows: ; In the formula, The dissipation coefficient of the pheromone. t Represents the number of iterations. t+k Representative at the t After the first iteration, and then after... k After the nth iteration, the total number of iterations is now; For the first Only one ant is on the path in this loop. The pheromone material left on the surface per unit length The formula is defined as follows: ; In the formula, The calculation formula is: ; F It is a constant; Q i For ants l In the path ij The performance function for moving upwards; Furthermore, verify whether the maximum number of iterations has been reached. If the optimal number of iterations has been reached, output the parameter combination represented by the optimal path. Furthermore, the coefficient of determination R of the model is calculated using the true values ​​of the test set and the predicted values ​​output by the prediction model based on the test set. 2 The root mean square error (RMSE) and mean absolute percentage error (MAPE) are used as indicators to evaluate the predictive performance of the model. The smaller the RMSE and MAPE, and the closer the coefficient of determination is to 1, the better the predictive performance of the model.

[0051] Among them, the coefficient of determination R 2 The formulas for defining the root mean square error (RMSE) and the mean absolute percentage error (MAPE) are as follows: ; ; ; In the formula, For predicted values, For the true value, This is the average value. Indicates the number of samples in the test set. Step 4: Input the prediction set into the prediction model to obtain the predicted corona initiation voltage of the target motor stator winding.

[0052] Example 4 like Figure 2 As shown, in a second aspect, this embodiment provides a prediction system for the corona initiation voltage of a motor stator winding, including: The detection module is used to acquire environmental data of the stator winding of the target motor; The data processing module preprocesses the multidimensional data according to a preset standardization function and divides it into a training set and a prediction set. The model training module is used to input the formed training set into the convolutional neural network, train the convolutional neural network through the ant colony optimization algorithm, obtain the optimal combination of learning rate and random drop-off rate of the convolutional neural network, and form the final prediction model. The performance evaluation module is used to calculate the model's coefficient of determination, root mean square error, and mean absolute percentage error by using the true values ​​of the training set and the predicted values ​​output by the prediction model based on the training set, thereby evaluating the model's predictive performance. The voltage prediction module is used to calculate the target motor stator winding environment dataset based on the optimized and trained prediction model, and output the prediction data.

[0053] The data processing module includes: The first data processing submodule is used to perform standardization processing on feature sample data and prediction target data according to the standardization equation of the prediction method described in Embodiment 2 above, and to remove dimensions. The second data processing submodule is used to sort the standardized data by category to obtain a dataset that conforms to the model input format corresponding to the detection data and sample data. The third data processing submodule is used to divide the sample feature dataset into a training set and a prediction set according to a preset ratio.

[0054] The ant colony optimization module includes: The first ant colony optimization submodule is used to initialize the data and assign values ​​to the number of optimization nodes and the number of ants according to the number of training set groups and the number of model parameter groups. The second ant colony optimization submodule is used to calculate the probability based on the probability transfer formula of ants on nodes in the above prediction method, and guide the ants to transfer. The third ant colony optimization submodule is used to update the pheromone intensity on the path according to the pheromone intensity adjustment formula of the ants in the above prediction method. The fourth ant colony optimization submodule is used to determine whether the maximum number of iterations has been reached and to output the optimal parameters.

[0055] Example 5 like Figure 3 As shown, in a third aspect, this embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the prediction method described in Embodiment 2 above.

[0056] The processor can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) or a program loaded from storage into random access memory (RAM). The processor may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may include a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of the present invention.

[0057] The RAM stores various programs and data required for the operation of the electronic device. The processor, ROM, and RAM are interconnected via a bus. The processor executes various operations of the method flow according to embodiments of the present invention by executing programs in the ROM and / or RAM. It should be noted that the programs may also be stored in one or more memories other than ROM and RAM. The processor may also execute various operations of the method flow according to embodiments of the present invention by executing programs stored in said one or more memories.

[0058] According to embodiments of the present invention, the electronic device may further include an input / output (I / O) interface, which is also connected to a bus. The electronic device may also include one or more of the following components connected to the input / output (I / O) interface: an input section including a keyboard, mouse, etc.; an output section including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a LAN card, modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the input / output (I / O) interface as needed. Removable media, such as magnetic disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage section as needed.

[0059] The program can rely on tangible storage media such as optical storage devices or magnetic storage devices. In another embodiment, the computer program can also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via a communication component, and / or installed from a removable medium. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0060] In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a processor, it performs the functions defined in the system of this embodiment of the invention. According to embodiments of the invention, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0061] According to embodiments of the present invention, program code for executing the computer programs provided in the embodiments of the present invention can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0062] Example 6 Fourthly, this embodiment provides a computer-readable storage medium that stores computer instructions that cause a computer to execute the prediction method described in Embodiment 2 above.

[0063] The computer-readable storage medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of the present invention.

[0064] According to embodiments of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of the present invention, a computer-readable storage medium may include ROM and / or RAM and / or one or more memories other than ROM and RAM as described above.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for predicting the corona initiation voltage of a motor stator winding, characterized in that, Includes the following steps: Step 1: Obtain multidimensional data from the motor stator winding environment; Step 2: Based on the preset standardization function and partitioning ratio, preprocess the multidimensional data and divide it into training set and prediction set; Step 3: Input the resulting training set into the convolutional neural network and train the convolutional neural network. Step 4: Input the prediction set into the prediction model to obtain the predicted corona initiation voltage of the target motor stator winding.

2. The method for predicting the corona initiation voltage of a motor stator winding according to claim 1, characterized in that: In step one, the multidimensional data includes temperature, humidity, air pressure, and the stator slot, stator outlet, and corona voltage at the "R" corner corresponding to the environmental factor.

3. The method for predicting the corona initiation voltage of a motor stator winding according to claim 2, characterized in that: In step two, the temperature, humidity, air pressure, and the corona voltage at the stator slot, stator outlet, and "R" corner corresponding to the environmental factors are preprocessed. After processing, the size of each element in the resulting dataset is linearly transformed to between 0 and 1.

4. The method for predicting the corona initiation voltage of a motor stator winding according to claim 1, characterized in that: In step one, the multidimensional data includes environmental factors, corona initiation voltage, and corona initiation voltage of the same motor electronic winding at different time points within a preset time period to predict the possible corona discharge of the motor stator winding in time sequence; or environmental factors, corona initiation voltage, and corona initiation voltage of the same motor electronic winding located in different environments at the same time to predict the corona initiation voltage of the stator windings of multiple motors.

5. The method for predicting the corona initiation voltage of a motor stator winding according to claim 4, characterized in that: In step two, according to the prediction requirements, the multidimensional data obtained in the motor stator winding environment should be divided into two parts: the first part is the environmental factor data from the target motor being predicted, which does not include corona initiation voltage data, and its single sample is a 3*1 data matrix; the second part is the data from the sample database and simulation experiment data, which includes both corona initiation voltage data and environmental factor data, and its single sample is a 6*1 data matrix; when dividing the preprocessed multidimensional data, the first part is used as the prediction set and the second part is used as the training set.

6. The method for predicting the corona initiation voltage of a motor stator winding according to claim 1, characterized in that: In step three, based on the possible ranges of learning rate and random discard rate, all possible parameter combinations that may optimize model performance are listed and the corresponding convolutional neural network model is constructed.

7. The method for predicting the corona initiation voltage of a motor stator winding according to claim 6, characterized in that: Step three includes the following steps: Step 31: Treat each ant as a convolutional neural network parameter configuration scheme, and treat all possible parameters of the convolutional neural network as nodes of the ant colony algorithm; initialize the ant colony, so that the ants are randomly distributed on the nodes, initialize the pheromone matrix so that the pheromone between each node is equal, set the maximum number of iterations, and start the iteration. Step 32: Ants choose paths based on probability, determined by pheromone concentration and the attractiveness of each path. The transfer probability calculation function for ants at nodes is shown below: Ants at the node Upward node The transition probability is defined as: ； In the formula: This indicates that the ant is located at the node. The set of path points that can be traversed in the next step; The relative importance of the trajectory reflects the role of the information accumulated by the ant during its movement. The relative importance of visibility reflects the degree to which heuristic factors are valued in ants' path selection during movement. For the ant to be located at the node Node The intensity of pheromones along the path, The attractiveness of each path to the ant can be defined as a performance function of the ant. t This indicates the current iteration number of the algorithm; The transition probability function For performance functions, the ant's position on the node Any reachable node It is defined as follows: ； In the formula, , and These are the true values ​​of the validation set; , and These are the predicted values ​​output by the prediction model; Step 33: After all the ants have formed a complete path, assign values ​​to the parameters of the convolutional neural network based on the movement paths of the ants. Step 34: Train the convolutional neural network using the training set; the convolutional neural network structure consists of 6 interconnected layers, namely the input layer, the first convolutional layer, the pooling layer, the second convolutional layer, the fully connected layer, and the regression layer. Step 35: The regression layer is responsible for outputting the prediction results of the convolutional neural network.

8. The method for predicting the corona initiation voltage of a motor stator winding according to claim 7, characterized in that: In step 35, the input layer is responsible for inputting the input parameters into the model. The input parameters are: ； In the formula, Representing the One input parameter; The first and second convolutional layers have the same structure and obtain data features through convolution. Their kernel size is 3*1 and their stride is 1. The output after this convolution is represented as follows: ； In the formula, Indicates the first One output unit For this convolutional layer, the input unit is... This represents the convolution kernel of the convolutional layer. For activation functions; The pooling layer is responsible for downsampling the output features of the convolutional layer to avoid the curse of dimensionality and overfitting. The downsampled output is: ； In the formula, For the pooling layer One output unit for The downsampling coefficients of the function, in addition The function uses average pooling. The fully connected layer is connected to all nodes of the second convolutional layer, and the output expression is: ； In the formula, For the fully connected layer One output unit For the first The convolutional layer neural unit and the first The connection weights between fully connected layer neural units, The first convolutional layer One output unit Indicates the first The threshold of a neural unit; Random dropout should also be performed after the fully connected layer, and its function expression is: ； In the formula, The probability that each neuron is preserved.

9. The method for predicting the corona initiation voltage of a motor stator winding according to claim 8, characterized in that: It also includes step 36, which calculates the performance function of each path in the current iteration and compares it with the best performance function of the previous iteration to obtain the path with the largest performance function; Step 37, update the pheromones along the path. The formula for updating pheromone intensity is shown below: For ants at nodes Any reachable node It is defined as follows: ； In the formula, The dissipation coefficient of the pheromone. For the first Only one ant is on the path in this loop. The pheromone material left on the surface per unit length t Represents the number of iterations. t+k Representative at the t After the first iteration, and then after... k After the nth iteration, the total number of iterations is now; The formula is defined as follows: ； In the formula, The calculation formula is: ； in, F It is a constant; Q i For ants l In the path ij The performance function for moving upwards; Step 38: Verify whether the maximum number of iterations has been reached. If the optimal number of iterations has been reached, output the parameter combination represented by the optimal path. Step 39: Calculate the model's coefficient of determination R2, root mean square error (RMSE), and mean absolute percentage error (MAPE) using the true values ​​of the training set and the predicted values ​​output by the prediction model based on the training set. These are used as evaluation indicators of the model's predictive performance. The smaller the RMSE and mean absolute percentage error, and the closer the coefficient of determination is to 1, the better the model's predictive performance.

10. The method for predicting the corona initiation voltage of a motor stator winding according to claim 9, characterized in that: In step 39, the coefficient of determination R 2 The formulas for defining the root mean square error (RMSE) and the mean absolute percentage error (MAPE) are as follows: ； ； ； In the formula, For predicted values, For the true value, This is the average value. This indicates the number of samples in the training set.

11. The prediction system of the method for predicting the corona initiation voltage of a motor stator winding according to any one of claims 1-10, characterized in that, include: The detection module is used to acquire environmental data of the stator winding of the target motor; The data processing module preprocesses the multidimensional data according to a preset standardization function and divides it into a training set and a prediction set. The model training module is used to input the formed training set into the convolutional neural network, train the convolutional neural network through the ant colony optimization algorithm, obtain the optimal combination of learning rate and random drop-off rate of the convolutional neural network, and form the final prediction model. The performance evaluation module is used to calculate the model's coefficient of determination, root mean square error, and mean absolute percentage error based on the true values ​​of the training set and the predicted values ​​output by the prediction model from the training set, thereby evaluating the model's predictive performance. The voltage prediction module is used to calculate the target motor stator winding environment dataset based on the optimized and trained prediction model, and output the prediction data.

12. The prediction system for corona initiation voltage of motor stator winding according to claim 11, characterized in that, The data processing module includes: The first data processing submodule is used to standardize the feature sample data and the prediction target data according to the standardization equation in the above prediction method, and remove the dimensions. The second data processing submodule is used to sort the standardized data by category to obtain a dataset that conforms to the model input format corresponding to the detection data and sample data. The third data processing submodule is used to divide the sample feature dataset into a training set and a prediction set according to a preset ratio.

13. The prediction system for corona initiation voltage of motor stator winding according to claim 12, characterized in that, The ant colony optimization module includes: The first ant colony optimization submodule is used to initialize the data and assign values ​​to the number of optimization nodes and the number of ants according to the number of training set groups and the number of model parameter groups. The second ant colony optimization submodule is used to calculate the probability based on the probability transfer formula of ants on nodes in the above prediction method, and guide the ants to transfer. The third ant colony optimization submodule is used to update the pheromone intensity on the path according to the pheromone intensity adjustment formula of the ants in the above prediction method. The fourth ant colony optimization submodule is used to determine whether the maximum number of iterations has been reached and to output the optimal parameters.

14. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the prediction method described in any one of claims 1-10.

15. A computer-readable storage medium, characterized in that: A computer-readable storage medium stores computer instructions that cause a computer to perform the prediction method according to any one of claims 1-10.

Images