Method and system for predicting motor efficiency

Abstract

The invention relates to a method and system for predicting motor efficiency, and the method comprises the steps: obtaining first data related to the testing of a motor, and the first data comprises the result of a routine test of the motor and the result of a type test of the motor; dividing the first data into a first subset and a second subset according to a preset proportion; inputting the first subset into the initialized machine learning model, executing an iterative operation of the initialized machine learning model, and determining an optimal hyper-parameter for the machine learning model using a parameter optimization algorithm; applying the optimal hyper-parameter to a machine learning model to obtain an optimized machine learning model; inputting the first subset into the optimized machine learning model, and training the optimized machine learning model to obtain a trained machine learning model; and inputting the second subset into the trained machine learning model and outputting the first predicted motor efficiency from the trained machine learning model.

Description

Technical Field

[0001] This invention relates to a method and system for predicting motor efficiency. Background Technology

[0002] Efficiency is a crucial performance indicator for motors and a key objective in motor optimization design. Currently, only a very small percentage of motors designed and manufactured in factories have clearly defined efficiency results. Efficiency is only determined through comprehensive testing for prototypes, quality inspections, and after-sales maintenance. For mass-produced motors, only a few prescribed tests are available, lacking efficiency information. This makes it impossible to directly determine whether a motor's efficiency is up to standard, or whether the efficiency fluctuation range of mass-produced motors is reasonable. Under these circumstances, there is not only the risk of substandard motor efficiency performance but also the risk of inappropriate material usage due to deviations between mass production and theoretical design, which contradicts lean design principles. Summary of the Invention

[0003] In view of this, embodiments of the present invention propose a novel method and system for predicting motor efficiency. This design can obtain highly accurate predicted motor efficiency values, providing a reference for performance evaluation and design optimization of motors before delivery. This method can fully utilize existing data resources, use machine learning methods to obtain a motor efficiency prediction model, realizing the prediction of motor efficiency from scratch, and can efficiently achieve large-scale motor efficiency prediction. Based on the gap between the predicted motor efficiency fluctuation range and the efficiency standard requirements, the subsequent optimization design space is determined, thereby saving motor costs.

[0004] According to a preferred embodiment of the present invention, a method for predicting motor efficiency is provided. The method includes: acquiring first data related to motor testing, wherein the first data includes results of routine motor tests and results of type tests of the motor; dividing the first data into a first subset and a second subset according to a predetermined ratio; inputting the first subset into an initialized machine learning model and performing iterative operations on the initialized machine learning model, wherein a parameter optimization algorithm is used to determine optimal hyperparameters for the machine learning model; applying the optimal hyperparameters to the machine learning model to obtain an optimized machine learning model; inputting the first subset into the optimized machine learning model to train the optimized machine learning model to obtain a trained machine learning model; and inputting the second subset into the trained machine learning model, and outputting a first predicted motor efficiency from the trained machine learning model. The method according to the present invention can achieve accurate prediction of batch motor efficiency in an automated manner, providing a basis for motor design optimization and performance evaluation.

[0005] According to an exemplary embodiment of the method of the present invention, the method further includes: acquiring second data related to the testing of the motor, the second data being more up-to-date than the first data in time; inputting the second data into a trained machine learning model, and having the trained machine learning model output a second predicted motor efficiency; and evaluating the performance of the trained machine learning model by comparing the second predicted motor efficiency with the first predicted motor efficiency. The method of the present invention enables the model to predict motor efficiency more accurately, thereby improving the overall prediction effect and enhancing the prediction accuracy and generalization ability of the evaluation model on novel samples.

[0006] According to an exemplary embodiment of the method of the present invention, the method further includes: performing data standardization on the first data and the second data. For example, data standardization is performed using Z-Score or OneHotEncoder methods to ensure that the model processes the data consistently during training and testing, thereby improving the model's generalization ability.

[0007] According to an exemplary embodiment of the method of the present invention, the method further includes: randomly selecting multiple different combinations of hyperparameters in the hyperparameter space of the machine learning model as initial sampling points for a parameter optimization algorithm. For motor efficiency prediction, randomly selecting combinations of hyperparameters as initial sampling points allows the optimal model configuration to be found through an intelligent optimization algorithm. This not only improves prediction accuracy but also enhances the robustness and generalization ability of the model, thereby improving motor performance evaluation, design optimization, and lean manufacturing.

[0008] According to an exemplary embodiment of the method of the present invention, determining the optimal hyperparameters for a machine learning model using a parameter optimization algorithm includes: evaluating a machine learning model with initial sampling points and a combination of hyperparameters used for iteration using five-fold cross-validation; evaluating a probability distribution model of the objective function used to build the machine learning model in the hyperparameter space; calculating a sampling function based on the probability distribution model, and determining the combination of hyperparameters for the next iteration in the hyperparameter space using the sampling function; and when a preset maximum number of iterations is reached and the iteration terminates, selecting the hyperparameter combination with the optimal objective function value from all hyperparameter combinations applied to the iteration as the optimal hyperparameters. This method, by combining five-fold cross-validation and Bayesian optimization algorithms to iteratively select the optimal hyperparameter combination, can significantly improve the performance of motor efficiency prediction models, making them more accurate and efficient.

[0009] According to an exemplary embodiment of the method of the present invention, obtaining first data related to the testing of the motor includes: establishing a relationship between the results of routine tests and the results of type tests for the same motor. In this invention, the method of establishing a relationship between the results of routine tests and the results of type tests for the same motor can fully utilize more test data about the motor, transforming it into effective information for motor efficiency prediction, increasing the amount of data available for prediction, thereby improving the prediction efficiency and accuracy of motor efficiency.

[0010] According to an exemplary embodiment of the method of the present invention, the machine learning model is multiple linear regression, random forest regression, GBDT, Adaboost, neural network, or LightGBM. By selecting a suitable machine learning model, the accuracy and speed of predicting motor efficiency can be significantly improved, enhancing the model's generalization ability and its ability to process complex data.

[0011] According to an exemplary embodiment of the method of the present invention, the parameter optimization algorithm is a Bayesian optimization algorithm, a grid search algorithm, or a random search algorithm. Optimizing hyperparameters using these algorithms can improve the performance of prediction models for motor efficiency, reduce computational resource consumption, enhance the model's generalization ability, and improve the accuracy and reliability of predictions.

[0012] According to an exemplary embodiment of the method of the present invention, routine tests include at least one of resistance testing, insulation testing, withstand voltage testing, no-load testing, and locked-rotor testing, and type testing includes motor efficiency testing. By modeling the correlation between efficiency data from routine tests and type tests, the accuracy and efficiency of motor efficiency prediction can be significantly improved, costs can be reduced, and data-driven real-time decision-making and optimization can be facilitated.

[0013] According to another preferred embodiment of the present invention, a system for predicting motor efficiency is provided. The system includes: one or more processors; and a non-transient memory for storing computer code instructions, which, when executed, cause the one or more processors to: acquire first data related to motor testing, wherein the first data includes results of routine tests and type tests of the motor; divide the first data into a first subset and a second subset according to a predetermined ratio; input the first subset into an initialized machine learning model and perform iterative operations on the initialized machine learning model, wherein a parameter optimization algorithm is used to determine optimal hyperparameters for the machine learning model; apply the optimal hyperparameters to the machine learning model to obtain an optimized machine learning model; input the first subset into the optimized machine learning model and train the optimized machine learning model to obtain a trained machine learning model; and input the second subset into the trained machine learning model, and output a first predicted motor efficiency from the trained machine learning model.

[0014] According to an exemplary embodiment of the system of the present invention, one or more processors are further configured to: acquire second data related to the testing of the motor, the second data being more up-to-date than the first data in time; input the second data into a trained machine learning model, and output a second predicted motor efficiency by the trained machine learning model; and evaluate the performance of the trained machine learning model by comparing the second predicted motor efficiency with the first predicted motor efficiency.

[0015] According to an exemplary embodiment of the system of the present invention, one or more processors are further configured to perform data standardization of the first data and the second data.

[0016] According to an exemplary embodiment of the system of the present invention, one or more processors are further configured to: randomly select multiple different combinations of hyperparameters in the hyperparameter space of the machine learning model as initial sampling points for parameter optimization algorithms.

[0017] According to an exemplary embodiment of the system of the present invention, one or more processors are further configured to: evaluate a machine learning model applied with initial sampling points and a combination of hyperparameters for iteration using five-fold cross-validation; evaluate a probability distribution model based on the objective function of the machine learning model built on the hyperparameter space; calculate a sampling function based on the probability distribution model, and determine a combination of hyperparameters for the next iteration in the hyperparameter space by the sampling function; and when a preset maximum number of iterations is reached and the iteration terminates, select the hyperparameter combination with the optimal objective function value from all hyperparameter combinations applied to the iteration as the optimal hyperparameters.

[0018] According to an exemplary embodiment of the system of the present invention, one or more processors are further configured to: establish a relationship between the results of routine tests and the results of type tests for the same motor.

[0019] According to another preferred embodiment of the present invention, a computer program product is provided, which includes a computer program that, when executed by a processor, performs the methods described above. Attached Figure Description

[0020] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which will make the above and other features and advantages of the present invention more apparent to those skilled in the art. In the drawings: Figure 1 A flowchart of a method for predicting motor efficiency according to a first embodiment of the present invention is shown.

[0021] Figure 2 A flowchart of a method for predicting motor efficiency according to a second embodiment of the present invention is shown.

[0022] Figure 3A flowchart of a method for predicting motor efficiency according to a third embodiment of the present invention is shown.

[0023] Figure 4 A schematic diagram of a system for predicting motor efficiency according to an embodiment of the present invention is shown. Detailed Implementation

[0024] To enable those skilled in the art to better understand the solutions of the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other solutions obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a product or device that comprises a series of units is not necessarily limited to those units explicitly listed, but may include other units not listed or inherent to such product or device.

[0026] Figure 1 A flowchart of a method for predicting motor efficiency according to a first embodiment of the present invention is shown. In this embodiment, the method includes: step S1, acquiring first data related to motor testing, wherein the first data includes the results of routine tests and type tests of the motor; step S2, dividing the first data into a first subset and a second subset according to a predetermined ratio; step S3, inputting the first subset into an initialized machine learning model and performing iterative operations on the initialized machine learning model, wherein a parameter optimization algorithm is used to determine the optimal hyperparameters for the machine learning model; step S4, applying the optimal hyperparameters to the machine learning model to obtain an optimized machine learning model; step S5, inputting the first subset into the optimized machine learning model and training the optimized machine learning model to obtain a trained machine learning model; and step S6, inputting the second subset into the trained machine learning model, and outputting a first predicted motor efficiency from the trained machine learning model.

[0027] In a preferred embodiment, step S1 first collects comprehensive data related to motor testing. This data includes two parts: the results of routine motor tests and the results of type tests. Routine test data includes, but is not limited to, the results of resistance tests, insulation tests, withstand voltage tests, no-load tests, and locked-rotor tests. Type test data includes the results of motor efficiency tests, which provide information on the motor's efficiency performance under specific operating conditions. The collected test data are integrated to ensure data completeness and consistency. For motors with both factory routine test and type test results, the factory test results correspond to their efficiency results in the type test, thus establishing an indirect correspondence between routine tests and motor efficiency. Based on these data conditions and assumptions, and considering that the final application is the efficiency prediction of factory-tested motors, the type test efficiency of motors with both type test and routine test results is used as the target, and the corresponding routine test items are used as model inputs. To improve model performance, additional basic information about the motor, including the number of poles, power, rated voltage, and wiring method, is also collected as input parameters for the model.

[0028] In a preferred embodiment, in step S2, the data is cleaned to remove possible outliers or erroneous data, providing high-quality input for model training. The integrated and cleaned first data is divided into a first subset (training set) and a second subset (test set) according to a preset ratio, such as 80% and 20%. The first subset is used to train the machine learning model, while the second subset is used to evaluate the model's predictive performance.

[0029] In a preferred embodiment, the machine learning model is multiple linear regression, random forest regression, GBDT, Adaboost, neural network, or LightGBM. For example, a benchmark model is selected from the aforementioned machine learning models, where the model evaluation metric is MAE (mean absolute error). A smaller MAE value indicates better model prediction performance, and its calculation formula is: , n is the corresponding sample size. predict For the model prediction results, actualThe actual efficiency of the motor refers to the efficiency value obtained during type testing. In a preferred embodiment, LightGBM is selected as the machine learning model. This is a highly efficient gradient boosting algorithm based on decision trees, characterized by fast training and high prediction accuracy. When initializing the model, initial hyperparameters are set, such as tree depth and minimum number of samples per leaf node. In step S3, the first subset of data is input into the initialized LightGBM model, and iterative training of the model begins. The goal of each iteration is to optimize the model's predictive ability by learning patterns from the data. The model's hyperparameters (such as learning rate, number of trees, maximum depth, minimum number of samples per leaf node, etc.—may not be optimal) are pre-set, and the initialized machine learning model serves as the basis or starting point for subsequent hyperparameter optimization processes.

[0030] In a preferred embodiment, during the iterative training process in step S3, a parameter optimization algorithm, such as Bayesian optimization, grid search, or random search, is used to determine the optimal hyperparameters. For LightGBM as a benchmark machine learning model, the hyperparameters that need to be tuned are shown in Table 1:

[0031] Table 1

[0032] In a preferred embodiment, a Bayesian optimization algorithm is used. This algorithm is a global optimization technique that estimates the performance distribution in the hyperparameter space by constructing a surrogate model (usually a Gaussian process) and a sampling function, thereby intelligently selecting the next evaluation point. The goal is to find the globally optimal hyperparameters or design parameters with as few evaluations as possible.

[0033] In another preferred embodiment, a grid search algorithm is used as the parameter optimization algorithm. For this, discrete parameter values ​​are predefined (e.g., hyperparameter num_leaves: [2, 5, 10]), instead of the continuous space of Bayesian optimization (e.g., hyperparameter learning_rate: (0.01, 0.1)). The continuous parameters are discretized, and the range is narrowed based on the Bayesian optimization results to avoid combinatorial explosion. Furthermore, the parameter set is optimized in stages (first adjusting the tree structure parameter max_depth / num_leaves, then adjusting the regularization parameter lambda_l1) to reduce the word search dimension, rather than relying on Bayesian multi-parameter combination optimization.

[0034] In another preferred embodiment, a random search algorithm is used as the parameter optimization algorithm. For this purpose, a probability distribution for the parameters is defined (e.g., hyperparameter max_depth: randint(3, 15)), rather than a fixed list. A log-uniform distribution (loguniform(0.001, 0.1)) is used for exponential parameters such as the learning rate.

[0035] In a preferred embodiment, after finding the optimal hyperparameter combination in step S4, it is applied to the LightGBM model to adjust the model configuration, resulting in an optimized machine learning model. The selection of optimal hyperparameters can significantly improve the model's generalization ability, reduce the risk of overfitting, and improve prediction accuracy.

[0036] In a preferred embodiment, in step S5, the first subset of data is again input into the optimized machine learning model for deep training until the model converges or reaches a preset number of iterations, thereby obtaining the trained machine learning model. This model can predict efficiency based on routine motor test data without the need for time-consuming and expensive type testing.

[0037] In a preferred embodiment, in step S6, the trained machine learning model is tested using a second subset of data. By inputting this data into the model, the model outputs a series of predicted efficiency values. The pre-divided test set data is then substituted into the final model trained in the previous step, and the model performance for this batch of samples, i.e., the mean absolute error (MAE), is calculated. Ideally, the MAE on the test set and the training set should be similar. These predicted values ​​are compared with the efficiency results of actual type trials to evaluate the model's predictive accuracy. By inputting the second subset of data into the trained machine learning model, the predicted efficiency value corresponding to the test set motor is output, i.e., the first predicted motor efficiency. These predicted values ​​reflect the model's generalization ability on unseen data, verifying the effectiveness of its performance in practical applications.

[0038] Figure 2 A flowchart of a method for predicting motor efficiency according to a second embodiment of the present invention is shown. Figure 2 In the embodiments, except Figure 1 The method steps further include: step S7, acquiring second data related to the test of the motor, the second data being more up-to-date than the first data in time; step S8, inputting the second data into a trained machine learning model, and having the trained machine learning model output a second predicted motor efficiency; and step S9, evaluating the performance of the trained machine learning model by comparing the second predicted motor efficiency with the first predicted motor efficiency.

[0039] In a preferred embodiment, although the test set and the training set are different samples, they overlap in the time dimension. That is, they may be samples with similar characteristics such as production or testing within a certain time period. The second data used as the validation set in step S7 is also a completely new sample compared to the training set. Not only are the samples not learned by the model, but the samples in the validation set are also samples from a completely new time interval. That is, the samples in this time period do not appear in the training set or the test set, so as to better evaluate the model's prediction accuracy and generalization ability on the new samples.

[0040] Figure 3 A flowchart of a method for predicting motor efficiency according to a third embodiment of the present invention is shown. Figure 3 In the embodiments, except Figure 1 The method and steps also include the following steps.

[0041] In a further embodiment, the method further includes: step S10, performing data standardization of the first data and the second data.

[0042] In a preferred embodiment, for example, GBDT is selected as the benchmark model for machine learning. For this, appropriate data standardization needs to be performed during the data preprocessing stage. For example, missing values ​​need to be handled. Since GBDT lacks a built-in missing value handling mechanism, if this model is used as a benchmark model for machine learning, missing value handling needs to be performed before model training. Common methods include explicit imputation of missing values, such as imputation with the mean / median (for continuous features) or adding a missing value indicator column. Furthermore, the target variable (e.g., the motor efficiency involved in this invention) needs to be transformed. GBDT is sensitive to the sample distribution in regression tasks; common methods, such as logarithmic transformation or quantile transformation, are used to make it closer to a normal distribution, thus making the model suitable.

[0043] In a preferred embodiment, for example, AdaBoost is chosen as the benchmark model for machine learning. For this, missing values ​​need to be handled. Similarly, since AdaBoost lacks a built-in missing value handling mechanism, if this model is used as a benchmark machine learning model, missing value handling needs to be performed before model training. Common methods include explicit imputation of missing values, such as using the mean / median (for continuous features) or adding a missing value indicator column. Furthermore, because AdaBoost fits difficult samples by adjusting sample weights, it is highly sensitive to noise and outliers. Therefore, noisy data needs to be removed to avoid overfitting to noisy samples and increase the model's robustness.

[0044] In a further embodiment, the method further includes: step S11, randomly selecting multiple different combinations of hyperparameters in the hyperparameter space of the machine learning model as initial sampling points for parameter optimization algorithms.

[0045] In a preferred embodiment, the possible value ranges of all hyperparameters of the model are determined. For example, when using the LightGBM model, it may be necessary to set the ranges for num_leaves (number of leaves in the tree), max_depth (maximum depth of the tree), min_data_in_leaf (minimum number of samples in a leaf node), learning_rate (learning rate), num_iterations (number of iterations), bagging_fraction (data sampling ratio), feature_fraction (feature sampling ratio), lambda_l1, and lambda_l2 (L1 and L2 regularization parameters), etc. The hyperparameter space is explored within the set hyperparameter ranges. Considering that the hyperparameter space may be very large and that there may be complex interdependencies between hyperparameters, random sampling can perform a preliminary scan of the hyperparameter space from different perspectives, helping the parameter optimization algorithm converge to the potential optimal region more quickly. Multiple hyperparameter combinations are generated using a random strategy, each combination being randomly selected within the defined value range. For example, `num_leaves` might randomly select an integer value from 1 to n, `max_depth` might select a value from -1 to infinity, `min_data_in_leaf` might select an integer from 1 to n, and so on. These randomly generated combinations of hyperparameters will serve as the initial sampling points for the Bayesian optimization algorithm.

[0046] In a further embodiment, using a parameter optimization algorithm to determine the optimal hyperparameters for the machine learning model includes: step 12, using five-fold cross-validation to evaluate the machine learning model with initial sampling points and a combination of hyperparameters used for iteration; step 13, evaluating a probability distribution model of the objective function used to build the machine learning model in the hyperparameter space; step 14, calculating a sampling function based on the probability distribution model, and determining the combination of hyperparameters used for the next iteration in the hyperparameter space using the sampling function; and step 15, when a preset maximum number of iterations is reached and the iteration terminates, selecting the hyperparameter combination with the optimal objective function value from all hyperparameter combinations applied to the iteration as the optimal hyperparameters.

[0047] In a preferred embodiment, in step 12, the first subset input to the initialized machine learning model is randomly divided into five equal-sized subsets, or the sample class distributions of each subset are ensured to be approximately the same. This is then iterated five times. In each iteration, one subset from the five subsets is selected as the validation (test) set, and the remaining four subsets form the training set for training the model. After training the model using the training set data in each iteration, the model is applied to the validation set to evaluate its performance. Performance evaluation metrics typically include accuracy, precision, recall, F1 score, mean squared error (MSE), and mean absolute error (MAE), with specific metrics depending on the type of learning task. Finally, the performance evaluation results from the five iterations are collected, and the average or relevant statistics are calculated as an overall evaluation of the model's performance.

[0048] In a preferred embodiment, in step 13, using the results evaluated based on the initial random points set in step S11 and the subsequent iteration points in step S14—that is, the loss or prediction accuracy of the model corresponding to each set of hyperparameters on a certain validation set—a probability distribution model of the objective function (prediction performance) is established in the hyperparameter space. Simply put, the Gaussian Process (GP) of the Bayesian optimization algorithm, based on existing data points, "learns" and "guesses" which hyperparameter combinations might yield better model performance (motor efficiency prediction accuracy), and the degree of uncertainty in this guess.

[0049] In a preferred embodiment, in step 14, the GP Upper Confidence Bound (UCB) is used as the acquisition function. This acquisition function is derived from the formula (which includes the mean and coefficients of the objective function estimate). The standard deviation / uncertainty of the objective function estimate is used to calculate a value. The set of unevaluated hyperparameter combinations with the highest UCB value is selected as the "next evaluation point" (the next set of hyperparameters to be tried). This set of hyperparameters is considered most likely to improve model performance (based on a comprehensive consideration of the prediction and uncertainty of the GP model). Based on the current GP model, the GP-UCB function is used to search the entire hyperparameter space to find the next hyperparameter combination with the greatest potential to improve performance.

[0050] In a preferred embodiment, in step 15, during the hyperparameter optimization process, a maximum number of iterations is set (e.g., 50, 100, etc.). The algorithm checks whether the current number of evaluations has reached or exceeded the preset maximum number of iterations, serving as a stopping condition for the Bayesian optimization process. When the preset maximum number of iterations is reached (the judgment result is "yes"), the iteration terminates. From all evaluated hyperparameter combinations, the set of hyperparameters with the optimal objective function value (minimum loss or highest prediction accuracy on the validation set) is selected as the final optimal (or best) LightGBM model parameter combination. This set of parameters is considered to be the best-performing hyperparameters found through the Bayesian optimization process within this budget (number of iterations). Furthermore, using the optimal parameter combination output from the previous step, and the original data that underwent data standardization at the beginning of the process, the final LightGBM prediction model, i.e., the trained machine learning model, is trained (constructed). This trained machine learning model is fully trained on the training set using the optimal hyperparameters, serving as the best-performing (expected) model for subsequent actual predictions.

[0051] In a further embodiment, acquiring the first data related to motor testing includes: establishing a relationship between the results of routine tests and the results of type tests for the same motor. All test results for the same motor in routine tests are collected, including but not limited to resistance tests, insulation tests, withstand voltage tests, no-load tests, and stall tests. Simultaneously, type test data for the motor, particularly the results of motor efficiency tests, are also acquired. These two types of data form the basis for establishing the relationship. Typically, each motor undergoes routine factory tests before leaving the production line. Therefore, in production, there exists a batch of motors with both type test results and routine factory test results. For the same motor, assuming all designs remain unchanged, its efficiency value should be a fixed level. Therefore, for a motor that has both routine factory test results and type test results, its routine factory test results correspond to its efficiency results in the type test, thus establishing an indirect correspondence between routine factory test results and motor efficiency.

[0052] Figure 4 A schematic diagram of a system for predicting motor efficiency according to an embodiment of the present invention is shown. Figure 3In one embodiment, the system 100 for predicting motor efficiency includes: one or more processors 10; and a non-transient memory 20 for storing computer code instructions 30, which, when executed, cause the one or more processors to: acquire first data related to motor testing, wherein the first data includes the results of routine tests and type tests of the motor; divide the first data into a first subset and a second subset according to a predetermined ratio; input the first subset into an initialized machine learning model and perform iterative operations on the initialized machine learning model, wherein a parameter optimization algorithm is used to determine the optimal hyperparameters for the machine learning model; apply the optimal hyperparameters to the machine learning model to obtain an optimized machine learning model; input the first subset into the optimized machine learning model and train the optimized machine learning model to obtain a trained machine learning model; and input the second subset into the trained machine learning model, and output a first predicted motor efficiency from the trained machine learning model.

[0053] In a preferred embodiment, multiple modules executed by the processor are stored in the aforementioned memory. These modules include a data acquisition module for acquiring first data related to motor testing; a data processing module for dividing the first data into a first subset and a second subset according to a predetermined ratio; a machine learning module for performing iterative operations on an initialized machine learning model based on the input first subset, determining optimal hyperparameters, and training the model based on the first subset after applying the optimal hyperparameters to obtain a trained machine learning model; and a prediction model for calculating a first predicted motor efficiency using the machine learning module based on the input second subset.

[0054] The technical advantages of this invention are as follows: By utilizing the motor efficiency prediction method and system of this application, and employing a Bayesian optimization algorithm to find the optimal hyperparameter combination of the LightGBM model, the accuracy and precision of the model's motor efficiency prediction can be significantly improved, thereby reducing reliance on actual type testing and increasing the efficiency of efficiency evaluation. This makes it possible to predict the efficiency of a large number of motors that have not undergone type testing, which helps to monitor and evaluate efficiency in real time on the production line and promptly identify potential problems. This method provides a reference for the performance evaluation and design optimization of motors before they leave the factory. Accurate efficiency prediction helps to identify motors with substandard efficiency in the early stages of the production line, thereby allowing for timely intervention, avoiding the production of defective products, and improving the overall quality of motors. Moreover, through prediction by machine learning models, reliance on expert experience is reduced, providing objective efficiency evaluation standards, making motor efficiency evaluation more systematic and standardized. Motor efficiency prediction can provide a basis for motor maintenance and fault diagnosis. By monitoring efficiency fluctuations, potential motor problems can be detected in advance, and preventive maintenance measures can be taken.

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting motor efficiency, characterized in that, The method includes: Acquire first data related to the testing of the motor, wherein the first data includes the results of routine tests of the motor and the results of type tests of the motor; The first data is divided into a first subset and a second subset according to a predetermined ratio; The first subset is input into an initialized machine learning model, and an iterative operation of the initialized machine learning model is performed, wherein a parameter optimization algorithm is used to determine the optimal hyperparameters for the machine learning model. The optimal hyperparameters are applied to the machine learning model to obtain an optimized machine learning model; The first subset is input into the optimized machine learning model to train the optimized machine learning model, resulting in a trained machine learning model; and The second subset is input into the trained machine learning model, and the trained machine learning model outputs a first predicted motor efficiency.

2. The method according to claim 1, characterized in that, The method further includes: Acquire second data related to the testing of the motor, the second data being more up-to-date than the first data in time; The second data is input into the trained machine learning model, and the trained machine learning model outputs a second predicted motor efficiency; and The performance of the trained machine learning model is evaluated by comparing the second predicted motor efficiency with the first predicted motor efficiency.

3. The method according to claim 2, characterized in that, The method further includes: Perform data standardization on the first data and the second data.

4. The method according to claim 1, characterized in that, The method further includes: In the hyperparameter space of the machine learning model, multiple different combinations of hyperparameters are randomly selected as initial sampling points for the parameter optimization algorithm.

5. The method according to claim 4, characterized in that, Determining the optimal hyperparameters for the machine learning model using parameter optimization algorithms includes: Five-fold cross-validation was used to evaluate the machine learning model that applied the initial sampling points and the hyperparameter combination used in the iteration; Based on the evaluation, a probability distribution model of the objective function of the machine learning model is established in the hyperparameter space; The acquisition function is calculated based on the probability distribution model, and the acquisition function determines the hyperparameter combination for the next iteration in the hyperparameter space; and When the preset maximum number of iterations is reached and the iteration terminates, the hyperparameter combination with the optimal objective function value is selected from all hyperparameter combinations applied to the iteration and is taken as the optimal hyperparameter.

6. The method according to claim 1, characterized in that, Obtaining first data related to the testing of the motor includes: For the same motor, establish the relationship between the results of the routine test and the results of the type test.

7. The method according to claim 1, characterized in that, The machine learning model is a multivariate linear regression, random forest regression, GBDT, Adaboost, neural network, or LightGBM.

8. The method according to claim 1, characterized in that, The parameter optimization algorithm is a Bayesian optimization algorithm, a grid search algorithm, or a random search algorithm.

9. The method according to claim 1, characterized in that, The routine tests include at least one of resistance testing, insulation testing, withstand voltage testing, no-load testing, and locked-rotor testing, and the type tests include motor efficiency testing.

10. A system for predicting motor efficiency, characterized in that, The system includes: One or more processors; and Non-transient memory for storing computer code instructions, which, when executed, cause the one or more processors to: Acquire first data related to the testing of the motor, wherein the first data includes the results of routine tests of the motor and the results of type tests of the motor; The first data is divided into a first subset and a second subset according to a predetermined ratio; The first subset is input into an initialized machine learning model, and an iterative operation of the initialized machine learning model is performed, wherein a parameter optimization algorithm is used to determine the optimal hyperparameters for the machine learning model. The optimal hyperparameters are applied to the machine learning model to obtain an optimized machine learning model; The first subset is input into the optimized machine learning model to train the optimized machine learning model, resulting in a trained machine learning model; and The second subset is input into the trained machine learning model, and the trained machine learning model outputs a first predicted motor efficiency.

11. The system according to claim 10, characterized in that, The one or more processors are further configured to: Acquire second data related to the testing of the motor, the second data being more up-to-date than the first data in time; The second data is input into the trained machine learning model, and the trained machine learning model outputs a second predicted motor efficiency. as well as The performance of the trained machine learning model is evaluated by comparing the second predicted motor efficiency with the first predicted motor efficiency.

12. The system according to claim 11, characterized in that, The one or more processors are further configured to: Perform data standardization on the first data and the second data.

13. The system according to claim 10, characterized in that, The one or more processors are further configured to: In the hyperparameter space of the machine learning model, multiple different combinations of hyperparameters are randomly selected as initial sampling points for the parameter optimization algorithm.

14. The system according to claim 13, characterized in that, The one or more processors are further configured to: Five-fold cross-validation was used to evaluate the machine learning model that applied the initial sampling points and the hyperparameter combination used in the iteration; Based on the evaluation, a probability distribution model of the objective function of the machine learning model is established in the hyperparameter space; The acquisition function is calculated based on the probability distribution model, and the acquisition function determines the hyperparameter combination for the next iteration in the hyperparameter space. as well as When the preset maximum number of iterations is reached and the iteration terminates, the hyperparameter combination with the optimal objective function value is selected from all hyperparameter combinations applied to the iteration and is taken as the optimal hyperparameter.

15. The system according to claim 10, characterized in that, The one or more processors are further configured to: For the same motor, establish the relationship between the results of the routine test and the results of the type test.

16. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, performs the method according to any one of claims 1-9.

Images