Test adjustment method and system of intelligent brushless motor

Abstract

The invention relates to the technical field of brushless motor testing, and discloses a test adjusting method and system for an intelligent brushless motor, and the method comprises the steps: carrying out the real-time monitoring of the current, rotating speed and load characteristics of the brushless motor, obtaining the operation data of the motor, and obtaining a motor working mode recognition result; constructing a motor parameter hierarchical optimization scheme; performing dynamic optimization to obtain a motor adaptive control parameter set; performing rolling optimization solution to obtain a motor prediction control instruction sequence; performing multi-objective optimization calculation to obtain a motor multi-performance balance adjustment scheme; a motor multi-performance balance adjustment scheme is integrated into a digital twin test platform, simulation verification is carried out on the performance of the brushless motor under different working conditions, a motor test adjustment strategy is obtained, the working mode of the motor can be recognized in real time, parameter optimization and performance prediction can be carried out based on historical data and the current state, and the stability of the motor test adjustment strategy is improved. Therefore, the optimal operation state of the motor under different working conditions is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of brushless motor testing, and in particular to a testing and adjusting method and system for an intelligent brushless motor. Background Art

[0002] With the rapid development of industrial automation and intelligent manufacturing, brushless motors are widely used in various application scenarios due to their advantages such as high efficiency, low noise and long life. However, traditional brushless motor testing and adjustment methods are often difficult to adapt to complex and changing working environments and load conditions, resulting in the motor performance not being fully utilized. This not only affects production efficiency, but may also cause energy waste and shorten equipment life.

[0003] At present, the control strategy of brushless motors usually adopts fixed parameter settings, which makes it difficult to optimize in real time for different working conditions. In addition, many performance indicators of motors, such as efficiency, torque pulsation and noise, often have a mutually restrictive relationship, and traditional methods are difficult to achieve a good balance between these indicators. At the same time, data collection, analysis and optimization and adjustment in the motor testing process are often separated, lacking a unified intelligent platform to achieve automated management of the entire process. Summary of the invention

[0004] The present invention provides a testing and adjustment method and system for an intelligent brushless motor, which are used to identify the working mode of the motor in real time, perform parameter optimization and performance prediction based on historical data and current status, so as to achieve the optimal operating state of the motor under different working conditions.

[0005] In a first aspect, the present invention provides a test and adjustment method for an intelligent brushless motor, the test and adjustment method for an intelligent brushless motor comprising: The current, speed and load characteristics of the brushless motor are monitored in real time to obtain motor operation data, and the motor operation data is processed to obtain a motor operation mode recognition result; Based on the motor working mode recognition result, a multi-time scale optimization structure including day-of-day, hour-level and second-level is constructed to obtain a hierarchical optimization scheme for motor parameters; According to the motor parameter hierarchical optimization scheme, the control parameters of the brushless motor are dynamically optimized to obtain a motor adaptive control parameter set; Based on the motor adaptive control parameter set, a control sequence of the brushless motor is subjected to rolling optimization and solution to obtain a motor predictive control instruction sequence; Based on the motor predictive control instruction sequence, multi-objective optimization calculation is performed on the efficiency, torque ripple and noise of the brushless motor to obtain a motor multi-performance balance adjustment scheme; The motor multi-performance balance adjustment scheme is integrated into the digital twin test platform, the performance of the brushless motor under different working conditions is simulated and verified, and the motor test adjustment strategy is obtained.

[0006] In a second aspect, the present invention provides a test and adjustment system for an intelligent brushless motor, the test and adjustment system for the intelligent brushless motor comprising: A monitoring module is used to monitor the current, speed and load characteristics of the brushless motor in real time, obtain motor operation data, and process the motor operation data to obtain a motor operation mode recognition result; A construction module is used to construct a multi-time scale optimization structure including day-ahead level, hour-level and second-level based on the motor working mode recognition result, and obtain a hierarchical optimization scheme for motor parameters; An optimization module, configured to dynamically optimize the control parameters of the brushless motor according to the motor parameter hierarchical optimization scheme to obtain a motor adaptive control parameter set; A solution module, used for performing rolling optimization solution on the control sequence of the brushless motor based on the motor adaptive control parameter set to obtain a motor predictive control instruction sequence; A calculation module, used for performing multi-objective optimization calculation on the efficiency, torque ripple and noise of the brushless motor based on the motor prediction control instruction sequence, and obtaining a motor multi-performance balance adjustment scheme; The verification module is used to integrate the motor multi-performance balance adjustment scheme into the digital twin test platform, simulate and verify the performance of the brushless motor under different working conditions, and obtain the motor test adjustment strategy.

[0007] The third aspect of the present invention provides a test and adjustment device for an intelligent brushless motor, comprising: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the test and adjustment device for the intelligent brushless motor executes the above-mentioned test and adjustment method for the intelligent brushless motor.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned test and adjustment method for the intelligent brushless motor.

[0009] In the technical solution provided by the present invention, the working mode of the motor can be accurately identified by real-time monitoring of the motor current, speed and load characteristics, laying the foundation for subsequent parameter optimization. A multi-time scale optimization structure including day-ahead level, hour level and second level is constructed to improve the accuracy and adaptability of motor parameter adjustment. By using reinforcement learning and deep Q network, dynamic optimization and adaptive adjustment of motor control parameters are realized, and the adaptability and operating efficiency of the motor under different working conditions are improved. By rolling optimization and solving the motor control sequence, combined with the online correction of the predicted state trajectory, accurate prediction and control of the future operating state of the motor are achieved, and the stability and reliability of the system are improved. By adopting a multi-objective evaluation system and a Pareto optimal solution strategy, the comprehensive optimization of multiple performance indicators such as motor efficiency, torque pulsation and noise is achieved, and the goal of multi-performance balance is achieved. The optimization scheme is integrated into the digital twin test platform, and the performance verification and optimization of the motor under different working conditions are realized through parallel simulation calculation and multi-dimensional performance evaluation, which improves the test efficiency and reliability. From data acquisition, pattern recognition to parameter optimization and test verification, a complete closed-loop management system is formed, which realizes the full process intelligence and automation of the brushless motor test and adjustment process.

[0010] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0011] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of an embodiment of a test and adjustment method for an intelligent brushless motor in an embodiment of the present invention; Figure 2 A schematic diagram of an embodiment of a test and adjustment system for an intelligent brushless motor in an embodiment of the present invention; Figure 3 It is a schematic diagram of an embodiment of a testing and adjusting device for an intelligent brushless motor in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0015] To facilitate understanding of this embodiment, a test and adjustment method for an intelligent brushless motor disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps: 101. Monitor the current, speed and load characteristics of the brushless motor in real time to obtain motor operation data, and process the motor operation data to obtain a motor operation mode recognition result; It is understandable that the execution subject of the present invention may be a test and adjustment system for an intelligent brushless motor, or may be a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0016] Specifically, the current, speed and load characteristic data of the brushless motor are monitored in real time. During the data acquisition process, these data are divided into sliding time windows to generate an initial feature sequence. Fast Fourier transform is applied to the initial feature sequence to convert the time domain data into a frequency domain feature matrix. Based on the frequency domain feature matrix, the mutual information coefficient between each feature is calculated to generate a feature correlation matrix. In order to optimize data processing and reduce complexity, the feature correlation matrix is ​​sorted by feature importance to obtain a reduced feature set. Fuzzy reasoning is used to analyze the reduced feature set to obtain the membership value of each working mode. The membership value is used to quantify the fuzziness of the current working mode of the motor, and the fuzzy entropy is calculated based on the membership value, which reflects the effectiveness of the current mode clustering. By evaluating the effectiveness index of fuzzy clustering, the reduced feature set is adaptively threshold segmented to generate a set of candidate working modes. For each candidate working mode, its time domain statistical features and wavelet packet energy features are extracted to construct a multidimensional feature vector. The multidimensional feature vector is classified and predicted by a classification algorithm to obtain the initial working mode recognition result. In order to evaluate the reliability of the initial recognition result, the corresponding confidence is calculated and a confidence sequence is generated. In order to ensure the stability of the working mode recognition, the confidence sequence is processed by sliding window averaging and a threshold is set for judgment to obtain a stable working mode recognition result. The recognition result is divided into the constant torque mode and constant power mode of the motor to obtain the motor working mode recognition result.

[0017] 102. Based on the results of motor working mode recognition, a multi-time scale optimization structure including day-ahead level, hour-level and second-level is constructed to obtain a hierarchical optimization scheme for motor parameters; Specifically, the motor working mode recognition result is decomposed into trend items, period items and random items. Based on the trend items and period items, the motor working mode in the next 24 hours is predicted to generate the day-ahead working mode prediction sequence. Based on the day-ahead working mode prediction sequence, the historical operation data of the motor is clustered and the typical working condition set of the motor is identified. For each typical working condition identified, a vector that can characterize its characteristics is extracted to form a working condition feature matrix. After obtaining the working condition feature matrix, the day-ahead optimization objective function is constructed through probability modeling. The function aims to form a day-ahead parameter optimization scheme by optimizing the basic operating parameters of the motor to adapt to future operating requirements. The generated day-ahead parameter optimization scheme is subdivided into time scales, and the optimization scheme is gradually refined from the day-ahead level to the hourly level. Based on the day-ahead optimization scheme, the hourly parameter optimization initial value is constructed. On this basis, an hourly dynamic optimization model is established, and the optimal control sequence at the hourly level is obtained by constructing a model predictive control objective function and solving it. Based on the hourly optimal control sequence, the second-level optimization target is generated. The second-level optimization goal is to achieve fine adjustment of the motor control parameters. Based on this goal, a second-level parameter fine-tuning strategy is generated to ensure that the motor can accurately adjust its operating parameters in a short time. The day-ahead parameter optimization scheme, hour-level optimal control sequence, and second-level parameter fine-tuning strategy are integrated to form a complete motor parameter hierarchical optimization scheme. This scheme can optimize the operation of the motor on different time scales, ensure that it can maintain an efficient and stable working state in both the short and long term, and achieve comprehensive optimization and adjustment of the brushless motor performance.

[0018] The day-ahead parameter optimization scheme is subdivided into time scales, and 24 hour-level time nodes are obtained by sampling at equal intervals. Based on the 24 hour-level time nodes, the motor operation data is segmented to generate an hour-level parameter optimization data set. This data set covers the motor's operating status and control parameters in each hour, ensuring that the optimization scheme can be effectively adjusted within a shorter time scale. An hour-level dynamic model of the motor is created based on the hour-level parameter optimization data set to reflect the dynamic behavior of the motor on the hour-level time scale and capture the state changes and response characteristics of the motor. By analyzing the model, the state variables and control variables of the motor are extracted. These variables are the core parameters in the optimization process and are used to describe the motor's operating status and control strategy. Based on the extracted state variables and control variables, the state trajectory equation and output equation in the prediction time domain are constructed to form a prediction model. The purpose of the prediction model is to predict the future state of the motor in a given time domain and guide the motor's operation and control strategy based on these predictions. In order to optimize the performance of the motor, a quadratic performance indicator function containing state deviation, control deviation and control increment is constructed based on the prediction model. This function aims to optimize the motor's control strategy by minimizing the state deviation and control deviation, thereby improving the overall performance of the motor. In order to consider various constraints of the motor in actual operation, soft constraints are introduced into the quadratic performance index function. Soft constraints allow the constraint conditions to be violated within a certain range to avoid the control problem being unable to be solved due to overly strict constraints. By introducing soft constraints, the objective function is expanded to consider not only the deviation of state and control, but also the state constraints and control constraints in actual operation. The expanded objective function is converted into an unconstrained optimization problem to simplify the solution process and obtain the model predictive control objective function. Based on the model predictive control objective function, an hourly dynamic optimization model including a state predictor, a controller and a feedback corrector is constructed. The state predictor is responsible for predicting the future state changes of the motor, the controller generates the optimal control sequence based on the prediction results, and the feedback corrector is used to adjust the control strategy in real time to cope with the uncertainty and disturbance in the operation of the motor. Through the dynamic optimization model, the operating parameters of the motor can be accurately optimized on an hourly time scale, thereby ensuring that the motor is always in the best working state within a short time range, and finally realizing the efficient regulation and control of the intelligent brushless motor.

[0019] 103. According to the motor parameter hierarchical optimization scheme, the control parameters of the brushless motor are dynamically optimized to obtain a motor adaptive control parameter set; Specifically, the parameters of the hierarchical optimization scheme of motor parameters are decoupled. Through the decoupling operation, the complex motor optimization problem is decomposed into three optimization parameter subsets of different time scales, namely, day-ahead, hour-level and second-level optimization parameter subsets. These subsets represent the optimization requirements of the motor at different time scales, ensuring that the operating state of the motor can be effectively adjusted in each time scale. Based on the optimization parameter subsets, a motor dynamic mathematical model including motor current, speed and torque is constructed, which is used to describe the dynamic behavior of the motor under different control conditions. The motor dynamic mathematical model is trained by reinforcement learning to obtain the initial Q value function. The Q value function is the core of reinforcement learning and is used to evaluate the benefits of taking different actions under different states. Based on the initial Q value function, a deep Q network (DQN) including a target network and an evaluation network is constructed. In DQN, the target network is used to generate a stable Q value, while the evaluation network is used to update the Q value in real time, so that the model can gradually approach the optimal strategy. The deep Q network is trained by experience replay. Experience replay breaks the correlation between data by repeatedly using past experience data in training to obtain an optimized Q value function. Based on the optimized Q-value function, an adaptive action selection strategy is generated, which can adaptively select the optimal control action according to the real-time motor state. Based on the adaptive action selection strategy, the motor control parameters are adjusted online to generate a candidate control parameter set. In order to ensure that these parameters can effectively optimize the motor performance, the square error of the candidate control parameter set is calculated to obtain the parameter evaluation index. The square error reflects the deviation between the current parameter and the target value, which is an important basis for evaluating the quality of the parameter. Based on the parameter evaluation index, a multi-criteria decision model including parameter importance and parameter sensitivity is constructed. The model is used to screen the candidate control parameter set and select the optimized control parameter set that can perform well under multiple standards. For the selected optimized control parameter set, adaptive weight allocation is performed, and different weights are assigned according to the importance of each parameter to ensure that the final control parameter can achieve the best effect under different operating conditions. The weighted control parameter set is smoothed to eliminate the mutation and discontinuity between parameters and ensure the stability of motor control. After smoothing, the final motor adaptive control parameter set can flexibly adjust the motor's operating state in different time scales, realize precise control and dynamic optimization of the brushless motor, and thus improve the overall performance and operating efficiency of the motor.

[0020] Based on the adaptive action selection strategy, the control parameter space of the motor is discretized and sampled to obtain the initial parameter sampling point set. By discretizing the parameter space, the entire parameter space can be represented by a limited number of sampling points, which is convenient for subsequent optimization calculations. The radial basis function is used to interpolate and fit the initial parameter sampling point set to generate a continuous parameter space model. The continuous parameter space model can more accurately describe each position in the parameter space, thereby providing a smooth and consistent estimate over the entire parameter range. The continuous parameter space model is gradient-calculated to obtain a parameter sensitivity matrix. The parameter sensitivity matrix reflects the degree of influence of the control parameters on the motor performance. By analyzing the matrix, the most critical parameters in the adjustment process can be identified. Based on the parameter sensitivity matrix, the control parameters are iteratively optimized using the gradient descent method with an adaptive step size. By adjusting the adaptive step size, the optimization process can be ensured to converge quickly and avoid falling into the local optimum, thereby obtaining the local optimal parameter set. A global search is performed on the local optimal parameter set to obtain the global control parameter set, thereby avoiding missing a better solution due to the limitation of the local optimal solution. On this basis, a multidimensional search space containing upper and lower limit constraints of parameters is constructed to ensure that the parameter combinations considered in the search process meet the physical constraints and design requirements of the actual operation of the motor. Based on the multidimensional search space, an unconstrained optimization problem is generated. By eliminating the upper and lower limit constraints of the parameters, the optimization problem is simplified to a standard mathematical optimization problem, which is convenient for solving using conventional optimization algorithms. After solving the unconstrained optimization problem, a candidate control parameter set is obtained. The candidate control parameter set contains the possible optimal parameter combination under the current optimization conditions. The candidate control parameter set is compared with the target parameter value and its square error is calculated. The square error reflects the gap between the candidate parameter and the ideal target value, and is an important indicator for evaluating the quality of the candidate parameter. The effectiveness of the candidate parameter set is quantified by calculating the square error.

[0021] 104. Based on the motor adaptive control parameter set, the control sequence of the brushless motor is subjected to rolling optimization and solution to obtain the motor predictive control instruction sequence; Specifically, according to the motor adaptive control parameter set, the parameter changes in the future control time domain are predicted to generate a parameter prediction sequence. Through the prediction sequence, the operating state of the motor in the future period can be mastered in advance. Based on the parameter prediction sequence, a motor prediction model containing state equations and output equations is constructed. The model is used to describe the dynamic behavior of the motor under different control inputs and can effectively simulate the response characteristics of the motor in the future time domain. The motor prediction model is linearized to simplify the complex nonlinear system to obtain the state matrix and input matrix at each prediction time. These matrices are the key to control sequence optimization and can link the future state of the system with the control input. Based on the state matrix and the input matrix, a comprehensive optimization objective function in the prediction time domain is constructed. The objective function usually includes multiple optimization objectives for motor performance, such as minimizing control deviation, reducing energy consumption, and improving motor efficiency, etc., to ensure that the system can operate in the best way under different trade-off conditions. The constructed comprehensive optimization objective function is solved to obtain the optimal control increment at the current moment. The optimal control increment is a fine-tuning of the current control input, which optimizes the running state of the motor at the current moment to be closest to the target state. Based on the optimal control increment, the control input of the motor is updated to generate a rolling optimization control sequence. The rolling optimization process means that the system will continuously repeat the above process at a new time point to ensure that the motor is always in the optimal control state during the entire operation. The rolling optimization control sequence is forward simulated to predict the future state trajectory of the motor. The predicted state trajectory shows the various states that the motor may reach in the future time domain, which can provide a more intuitive reference for control decisions. In order to improve the control accuracy, the predicted state trajectory is corrected online, and the control input is adjusted in real time to cope with various disturbances and uncertainties that occur in actual operation. After online correction, the motor's predicted control instruction sequence is obtained, which contains the control instructions for the motor in the future time domain, which can guide the motor to maintain a stable and efficient operating state in a changing environment.

[0022] 105. Based on the motor predictive control instruction sequence, multi-objective optimization calculations are performed on the efficiency, torque ripple, and noise of the brushless motor to obtain a motor multi-performance balance adjustment scheme; Specifically, feature extraction is performed on the motor predictive control instruction sequence to generate a control parameter vector. The control parameter vector is an important data set that describes the motor's operating state. Based on the control parameter vector, a multi-objective evaluation system including an efficiency model, a torque ripple model, and a noise model is constructed. By integrating multiple performance indicators, the system can comprehensively evaluate the motor's operating performance under different control conditions. The constructed multi-objective evaluation system is subjected to initial population generation to obtain a candidate solution set. The initial population represents different combinations of motor control parameters, which show different advantages and disadvantages in terms of performance indicators. In order to effectively select the optimal solution set, the candidate solution set is subjected to non-dominated sorting and crowding calculation to obtain the initial Pareto frontier. The Pareto frontier contains a solution set that is not dominated by other solutions in all objectives, showing the possibility space of multi-objective optimization. Based on the initial Pareto frontier, the candidate solution set is iteratively optimized. Through repeated iterations, the optimization algorithm gradually approaches a better solution set to form an optimized Pareto solution set. This solution set not only considers the trade-offs between various objectives, but also provides solutions that perform well in all objectives. Based on the optimized Pareto solution set, the membership function of each objective is calculated. The membership function reflects the fitness of each solution set on different objectives and can quantify the performance of each solution in the multi-objective evaluation system. In order to find the best compromise solution among multiple objectives, the optimal compromise solution is constructed based on the membership function. The optimal compromise solution balances the relationship between efficiency, torque ripple and noise to ensure that the motor can achieve a better state in different performance indicators. Based on the optimal compromise solution, the brushless motor is simulated and calculated to obtain the predicted values ​​of each performance indicator. The simulation calculation can simulate the performance level that the motor may achieve in actual operation and provide reliable prediction data. Based on the predicted values ​​of each performance indicator, multiple groups of optimization schemes are generated. Each optimization scheme represents a possible control strategy under different weight combinations. Based on the hierarchical analysis method, multiple groups of optimization schemes are comprehensively evaluated and ranked. The hierarchical analysis method comprehensively evaluates each scheme by establishing a multi-level goal and criterion system, and finally obtains the best motor multi-performance balance adjustment scheme. This scheme can achieve the best balance between efficiency, torque ripple and noise, ensuring that the brushless motor can maintain an efficient and stable operating state under different working conditions.

[0023] 106. Integrate the motor's multi-performance balance adjustment scheme into the digital twin test platform, simulate and verify the performance of the brushless motor under different working conditions, and obtain the motor test adjustment strategy.

[0024] Specifically, the parameters of the motor multi-performance balance adjustment scheme are mapped, and the parameters in the adjustment scheme are converted into the initial configuration parameters of the digital twin model to ensure that the digital twin model can accurately reflect the actual physical characteristics and operating status of the motor. Based on the initial configuration parameters, a complete motor digital twin system including the motor body, drive system and load model is constructed. The system realizes the virtual reproduction of the entire life cycle of the motor by accurately modeling each component of the motor. Based on the motor digital twin system, multiple sets of operating condition parameter sets are generated. The generation of operating condition parameter sets is aimed at simulating the performance of the motor under different operating conditions, covering a wide range of scenarios from steady-state conditions to dynamic conditions. A test scheme matrix including steady-state conditions and dynamic conditions is designed. Through systematic test design, the test scheme matrix ensures comprehensive verification of motor performance under different conditions and can reveal the operating characteristics of the motor in a changing environment. Based on the test scheme matrix, parallel simulation calculations are performed on the motor digital twin system. Through parallel simulation, multiple sets of simulation result data are obtained in a relatively short time, which reflect the performance of the motor under different conditions. The simulation result data is statistically analyzed and feature extracted to obtain the distribution characteristics of performance indicators and reveal the changing rules of key performance indicators such as efficiency, torque pulsation and noise of the motor. Based on the obtained performance indicator distribution characteristics, a multidimensional performance evaluation space including efficiency, torque pulsation and noise is constructed. This evaluation space forms a comprehensive performance evaluation model by integrating multiple performance indicators. The simulation results are comprehensively scored based on the performance evaluation model to obtain the advantages and disadvantages of different adjustment schemes, which helps to identify the optimal adjustment scheme and ensure that the motor can achieve the best balance in various performance indicators. A sensitivity analysis is performed on the advantages and disadvantages of the schemes to identify the key factors affecting the motor performance. By analyzing the target influencing factors, the contribution of each factor to the performance is determined, and the corresponding optimization direction and adjustment interval are generated. Based on the determined optimization direction and adjustment interval, a comparative analysis is performed to refine and optimize the motor adjustment strategy. Through comparative analysis, the influence of various factors can be comprehensively considered to obtain the optimal motor test adjustment strategy. This strategy can ensure that the brushless motor always maintains an efficient and stable operating state under different working conditions and achieves the best balance between key performance indicators such as efficiency, torque pulsation and noise.

[0025] In the embodiment of the present invention, by real-time monitoring of the motor current, speed and load characteristics, the working mode of the motor can be accurately identified, laying the foundation for subsequent parameter optimization. A multi-time scale optimization structure including day-ahead level, hour level and second level is constructed to improve the accuracy and adaptability of motor parameter adjustment. By using reinforcement learning and deep Q network, dynamic optimization and adaptive adjustment of motor control parameters are realized, and the adaptability and operating efficiency of the motor under different working conditions are improved. By rolling optimization and solving the motor control sequence, combined with online correction of the predicted state trajectory, accurate prediction and control of the future operating state of the motor are achieved, and the stability and reliability of the system are improved. By adopting a multi-objective evaluation system and Pareto optimal solution strategy, the comprehensive optimization of multiple performance indicators such as motor efficiency, torque pulsation and noise is achieved, and the goal of multi-performance balance is achieved. The optimization scheme is integrated into the digital twin test platform, and the performance verification and optimization of the motor under different working conditions are realized through parallel simulation calculation and multi-dimensional performance evaluation, which improves the test efficiency and reliability. From data acquisition, pattern recognition to parameter optimization and test verification, a complete closed-loop management system is formed, realizing the full process intelligence and automation of the brushless motor test and adjustment process.

[0026] In a specific embodiment, the process of executing step 101 may specifically include the following steps: The current, speed and load characteristic data of the brushless motor are segmented by a sliding time window to obtain an initial feature sequence, and the initial feature sequence is subjected to a fast Fourier transform to obtain a frequency domain feature matrix; Based on the frequency domain feature matrix, the mutual information coefficient between each feature is calculated to obtain the feature correlation matrix, and the feature correlation matrix is ​​sorted by feature importance to obtain a reduced-dimensional feature set; Perform fuzzy reasoning on the reduced dimension feature set to obtain the membership value of each working mode, and calculate the fuzzy entropy based on the membership value to obtain the fuzzy clustering effectiveness index; According to the fuzzy clustering effectiveness index, the dimension reduction feature set is segmented by adaptive threshold to obtain a set of candidate working modes, and the time domain statistical features and wavelet packet energy features are extracted from each mode in the candidate working mode set to obtain a multi-dimensional feature vector; Classify and predict the multi-dimensional feature vector to obtain the initial working mode recognition result, and calculate the confidence of the initial working mode recognition result to obtain a confidence sequence; The confidence sequence is averaged by sliding windows and the threshold is judged to obtain a stable working mode recognition result. Based on the stable working mode recognition result, the motor is divided into a constant torque mode and a constant power mode to obtain a motor working mode recognition result.

[0027] Specifically, the current, speed and load characteristic data of the brushless motor are segmented by sliding time windows to extract useful feature information. Sliding time window segmentation captures the dynamic changes of the motor in different time segments by setting a fixed-length window on the time axis and gradually moving the window. For example, setting a length of time window, and with a step size Slide, and each slide will generate a new data segment to obtain the initial feature sequence. Perform fast Fourier transform on the initial feature sequence. Fast Fourier transform is an effective tool to convert time domain signals into frequency domain signals, which can reveal the frequency components and their intensity distribution in the signal. By performing fast Fourier transform on the initial feature sequence, the frequency domain feature matrix is ​​obtained. Assume that the original time series is , after fast Fourier transform, the obtained frequency domain signal is expressed as , its mathematical expression is: ; in, is the frequency, is the length of the sequence, For time point, is an imaginary unit. The frequency domain feature matrix contains the signal amplitude and phase information under multiple frequency components, which can reflect the dynamic characteristics of the motor at different frequencies. Based on the frequency domain feature matrix, the mutual information coefficient between each feature is calculated. The mutual information coefficient is an indicator used to measure the dependency between two random variables, which can quantify the degree of correlation between frequency domain features. Assume that the two features and The mutual information is , and its calculation formula is: ; in, It is a feature and The joint probability distribution of and Characteristics and The marginal probability distribution of . By calculating the mutual information coefficient between the frequency domain features, the feature correlation matrix is ​​obtained, and each element in the matrix represents the strength of association between two features. The feature correlation matrix is ​​sorted by feature importance to identify the most representative features in describing the motor state. By sorting, the features with higher importance are retained to form a reduced dimension feature set. Fuzzy reasoning is performed on the reduced dimension feature set to identify different working modes of the motor. Fuzzy reasoning handles uncertainty and ambiguity through fuzzy logic, and is suitable for handling fuzzy states caused by changes in operating conditions during the operation of the motor. Each working mode can be characterized by a membership value, which indicates the degree to which a feature belongs to a certain mode. Based on the membership value calculated by fuzzy reasoning, the fuzzy entropy is calculated. The fuzzy entropy is used to quantify the ambiguity between different working modes and the effectiveness of clustering. The formula for fuzzy entropy can be expressed as: ; in, is the fuzzy entropy, is the number of working modes. The discrimination of different working modes is evaluated by calculating fuzzy entropy, and the effectiveness index of fuzzy clustering is obtained. According to the effectiveness index of fuzzy clustering, the reduced dimension feature set is adaptively threshold segmented to generate a set of candidate working modes. Each mode in the candidate working mode set represents the working state of the motor under specific conditions. Time domain statistical features and wavelet packet energy features are extracted from the candidate working modes. The extraction of these features can reflect the dynamic behavior of the motor from different angles, such as time domain features such as mean, variance, skewness, and energy distribution features extracted by wavelet packet transform. The multidimensional feature vector is classified and predicted to identify the current working mode of the motor. Classification prediction can use machine learning algorithms such as support vector machine, decision tree or neural network. The predicted result is the possible working mode of the motor, and the confidence of the initial working mode recognition result is calculated. The confidence reflects the credibility of the classification result, usually expressed in the form of probability. By calculating the confidence for each initial working mode recognition result, a confidence sequence is obtained. In order to enhance the stability of the recognition result, the confidence sequence is processed by sliding window averaging, and a threshold is set for judgment. The sliding window average can smooth the fluctuation of confidence and eliminate short-term abnormal changes. By thresholding the smoothed confidence sequence, a stable working mode recognition result is obtained. Based on this result, the motor's operating state is divided into constant torque mode and constant power mode, and finally a complete motor working mode recognition result is obtained.

[0028] In a specific embodiment, the process of executing step 102 may specifically include the following steps: The motor working mode recognition results are decomposed into time series to obtain trend items, period items and random items. Based on the trend items and period items, the motor working mode in the next 24 hours is predicted to obtain the day-ahead working mode prediction sequence. Based on the previous working mode prediction sequence, cluster analysis is performed on the motor's historical operating data to obtain a typical working condition set, and feature vectors are extracted from each working condition in the typical working condition set to obtain a working condition feature matrix; Probabilistic modeling is performed on the operating condition characteristic matrix to obtain the day-ahead optimization objective function. Based on the day-ahead optimization objective function, the basic operating parameters of the motor are preliminarily optimized to obtain the day-ahead parameter optimization solution. The day-ahead parameter optimization scheme is subdivided into time scales to obtain the initial values ​​of hourly parameter optimization. Based on the initial values ​​of hourly parameter optimization, the model predictive control objective function is constructed to obtain the hourly dynamic optimization model. Solve the hourly dynamic optimization model to obtain the hourly optimal control sequence, and generate the second-level optimization target based on the hourly optimal control sequence; A second-level parameter fine-tuning strategy is generated according to the second-level optimization target, and the day-ahead parameter optimization scheme, hour-level optimal control sequence and second-level parameter fine-tuning strategy are integrated to obtain a hierarchical optimization scheme for motor parameters.

[0029] Specifically, the motor working mode recognition results are decomposed into three basic components: trend term, periodic term and random term. The trend term reflects the long-term change trend of the motor working mode, which is usually caused by the gradual increase or decrease of the motor load. The periodic term represents the periodic changes in the motor operating state, such as mode fluctuations caused by daily usage patterns or seasonal changes. The random term captures short-term fluctuations or unpredictable emergencies in the motor operation. Time series decomposition can be achieved through an additive model or a multiplicative model, and the specific choice depends on the data characteristics. For the additive model, the time series It can be expressed as: ; in, is the trend item, is the periodic term, is a random term. By decomposing, the stable trend and periodic changes in the motor working mode can be clearly distinguished, so as to better understand and predict future patterns. Based on the trend term and the periodic term, a time series prediction model such as ARIMA (autoregressive integrated moving average model) or LSTM (long short-term memory neural network) is used to predict the motor working mode in the next 24 hours, and the day-ahead working mode prediction sequence is obtained. The ARIMA model captures the pattern in the time series through autoregression and moving average, and its model formula is: ; in, is the time series value, is the autoregression coefficient, is the moving average coefficient, is a random error. Based on the day-ahead working mode prediction sequence, the historical operating data of the motor is clustered and analyzed to identify a set of typical operating conditions for the motor. Typical operating conditions represent the common operating states of the motor under different conditions. Using clustering algorithms such as K-means or DBSCAN, the historical data are divided into multiple clusters, each cluster representing a typical operating condition. The characteristic vector of each operating condition is obtained by analyzing its performance under different conditions, and these characteristic vectors constitute the operating condition feature matrix. After obtaining the operating condition feature matrix, the probability modeling method is used to establish the day-ahead optimization objective function. The objective function is used to describe the optimization objectives under different operating conditions, such as minimizing energy consumption or improving efficiency. The objective function can be expressed as: ; in, For the The probability of a working condition, is the optimization objective function value under this condition, is the total number of working conditions. Based on this objective function, the basic operating parameters of the motor are preliminarily optimized to generate a day-ahead parameter optimization solution. The day-ahead parameter optimization solution is subdivided into time scales, and the long-time optimization solution is converted into initial values ​​of shorter time scales to form hourly parameter optimization initial values. Based on the hourly parameter optimization initial values, the model predictive control objective function is constructed to form an hourly dynamic optimization model. The objective function of model predictive control usually includes state deviation and control cost, and its general form is: ; in, is the system state vector, is the control input vector, and is the weight matrix, is the length of the prediction time domain. By solving the objective function, the hourly optimal control sequence is obtained. Based on the hourly optimal control sequence, the second-level optimization target is generated. The second-level optimization aims to make subtle adjustments to the motor control to cope with fluctuations and changes in a short period of time. According to the second-level optimization target, a second-level parameter fine-tuning strategy is generated to ensure that the motor can respond quickly to changes during operation and maintain the best state. The day-ahead parameter optimization scheme, the hourly optimal control sequence and the second-level parameter fine-tuning strategy are integrated to form a complete motor parameter hierarchical optimization scheme. Through optimization at different time scales, comprehensive control and adjustment of the motor operating state are achieved. For example, when the brushless motor of a factory is running at peak hours, through the above method, the day-ahead optimization scheme ensures that the motor runs at the optimal efficiency as a whole, while the hourly optimization refines the output power of the motor in each hour, and the second-level optimization is responsible for real-time fine-tuning to avoid current fluctuations caused by sudden changes in load, thereby ensuring the stability and efficiency of the motor at different time scales.

[0030] In a specific embodiment, the execution step subdivides the day-ahead parameter optimization scheme into time scales to obtain the hourly parameter optimization initial values, and constructs the model predictive control objective function based on the hourly parameter optimization initial values. The process of obtaining the hourly dynamic optimization model may specifically include the following steps: The day-ahead parameter optimization scheme is sampled at equal intervals to obtain 24 hourly time nodes. Based on the 24 hourly time nodes, the motor operation data is processed in segments to obtain an hourly parameter optimization data set. Create an hourly-level dynamic model of the motor based on the hourly-level parameter optimization data set, and extract state variables and control variables based on the hourly-level dynamic model of the motor; According to the state variables and control variables, the state trajectory equation and output equation in the prediction time domain are constructed to obtain the prediction model, and based on the prediction model, a quadratic performance indicator function including state deviation, control deviation and control increment is constructed; The soft constraint term is introduced into the quadratic performance index function to obtain the extended objective function considering the state constraint and control constraint, and the extended objective function is transformed into an unconstrained optimization problem to obtain the model predictive control objective function. Based on the model predictive control objective function, an hourly dynamic optimization model including a state predictor, a controller and a feedback corrector is constructed.

[0031] Specifically, the time series of the day-ahead parameter optimization scheme is evenly divided, and the entire 24-hour time range is divided into 24 equal time periods, each of which corresponds to an hourly time node. By sampling at equal intervals, it is ensured that the operating characteristics of each hour are fully captured, forming a more comprehensive hourly parameter optimization data set. The motor operation data is segmented based on the time node. The operating characteristics of the motor are extracted in each hour, including parameters such as current, speed, and load. By independently analyzing the data of each hour, the corresponding hourly parameter optimization data set is generated. The hourly dynamic model of the motor is created based on the hourly parameter optimization data set. The hourly dynamic model aims to describe the dynamic behavior of the motor in each hour, reflects the response of the motor under different control inputs, and can capture the evolution of the internal state of the system over time. In the modeling process, the state variables and control variables of the system are identified and extracted. State variables are variables that describe the internal state of the system, such as the speed, current, and temperature of the motor, while control variables are adjustable parameters of the system input, such as voltage and current set values. Based on the extracted state variables and control variables, the state trajectory equation and output equation in the prediction time domain are constructed to form a prediction model of the motor. The prediction model is used to predict the behavior of the system in the future under given control inputs. Usually, the prediction model can be expressed in the form of the following state space equation: ; ; in, is the state variable vector, describing the system at time Status; is the control input vector, representing the control input of the system; and They are the state matrix and input matrix of the system, respectively, describing the relationship between state variables and control variables; is the output vector, representing the output of the system; and is the output matrix, which is used to describe the influence of state and control on the output. In order to optimize the control strategy of the system, a quadratic performance index function including state deviation, control deviation and control increment is constructed based on the prediction model. The goal of this performance index function is to minimize the deviation and control cost of the system in the prediction time domain, which is usually expressed as: ; in, is the performance indicator function; is the predicted state vector, is the reference trajectory; is the control input, To control the increment; , and is a weight matrix, which represents the weights of state deviation, control input and control increment respectively. By minimizing , ensuring that the system has the smallest deviation when tracking the reference trajectory while maintaining the rationality of the control input and control increment. Introducing soft constraints in the performance indicator function enhances the flexibility and practicality of the model. Soft constraints allow the system to violate certain constraints within a certain range. For example, when the temperature or current of the motor exceeds a certain safety range, the system can make adaptive adjustments instead of strictly following hard constraints. The extended objective function after introducing soft constraints can be expressed as: ; in, is the weight of the soft constraint, For the Constraints, is the number of constraints. In this way, the extended objective function can be transformed into an unconstrained optimization problem, making the optimization process more flexible and efficient. Based on the above model predictive control objective function, an hourly dynamic optimization model including a state predictor, a controller and a feedback corrector is finally constructed. The state predictor is responsible for predicting the future state of the system under a given control input; the controller generates the optimal control sequence based on the prediction results; and the feedback corrector is used to correct the deviation caused by uncertainty or external disturbances during the actual operation. For example, in industrial applications, if the load of a brushless motor suddenly increases, the state predictor will predict the motor state changes in the next few seconds, and the controller will adjust the voltage or current to ensure that the motor remains in the best operating state, while the feedback corrector will adjust the control strategy in real time when the actual motor response deviates from the expected to maintain the stability of the system. Through this hourly dynamic optimization model, the operating parameters of the motor are accurately adjusted on an hourly time scale to ensure that the motor always maintains the best performance and stability during operation.

[0032] In a specific embodiment, the process of executing step 103 may specifically include the following steps: Decouple the parameters of the hierarchical optimization scheme of motor parameters to obtain the optimization parameter subsets at the day-ahead level, hour-level and second-level. Based on the optimization parameter subsets, construct the motor dynamic mathematical model including motor current, speed and torque. Perform reinforcement learning training on the motor dynamic mathematical model to obtain the initial Q-value function, and based on the initial Q-value function, build a deep Q network including the target network and the evaluation network; Perform experience replay training on the deep Q network to obtain an optimized Q-value function, and generate an adaptive action selection strategy based on the optimized Q-value function; Based on the adaptive action selection strategy, the motor control parameters are adjusted online to obtain a candidate control parameter set, and the square error of the candidate control parameter set is calculated to obtain a parameter evaluation index; Based on the parameter evaluation index, a multi-criteria decision-making model including parameter importance and parameter sensitivity is constructed, and the candidate control parameter set is screened based on the multi-criteria decision-making model to obtain the optimized control parameter set; Adaptive weight allocation is performed on the optimized control parameter set to obtain a weighted control parameter set, and the weighted control parameter set is smoothed to obtain a motor adaptive control parameter set.

[0033] Specifically, the overall optimization scheme is decomposed into subsets of different time scales, namely, the day-ahead, hour-level, and second-level optimization parameter subsets. Parameter decoupling optimizes the motor parameters on each time scale by decomposing the complex multi-scale problem into multiple simple problems. The day-ahead parameter subset focuses on the long-term trend and large-scale adjustment of the motor, such as the overall efficiency optimization and long-term operation stability of the motor; the hour-level parameter subset focuses on the adjustment in the medium and short term, such as the load fluctuation response of the motor in different hours; and the second-level parameter subset is responsible for handling the control with high real-time requirements, such as the response speed and torque pulsation control of the motor during instantaneous load changes. Based on the optimized parameter subset, a dynamic mathematical model of the motor including motor current, speed, and torque is constructed. The goal of the model is to describe the dynamic behavior of the motor under different control parameters and capture the key physical characteristics of the motor in operation. Mathematical models can usually be represented by differential equations or state space. For example, the state space model of the motor can be expressed as: ; ; in, Represents the state vector of the motor, including current, speed, torque, etc. is the control input vector, and are system noise and measurement noise, respectively, is the system matrix, describing the relationship between state and input. The constructed motor dynamic mathematical model is trained by reinforcement learning. Reinforcement learning aims to enable the model to learn how to select the optimal control strategy under different states through continuous interaction with the environment. The Q-value function is learned by the Q-learning algorithm, which is used to evaluate the long-term reward of taking a certain action in a certain state. The initial Q-value function is usually randomly initialized and gradually approaches the optimal Q value through continuous learning and updating. In order to enhance the learning effect, a deep Q network (DQN) including a target network and an evaluation network is constructed based on the initial Q-value function. The role of the target network is to generate a stable target Q value to avoid oscillation during the learning process, while the evaluation network updates the Q-value function in real time to gradually approach the optimal value. DQN approximates the Q value through a deep neural network, enabling it to handle high-dimensional state spaces and complex control tasks. Experience replay training is performed on the deep Q network. Experience replay training breaks the time correlation between data by reusing past experience data during the training process, thereby improving the generalization ability and convergence speed of the model. After this training process, an optimized Q-value function is obtained. Based on the optimized Q-value function, an adaptive action selection strategy is generated. The adaptive action selection strategy selects the optimal control action based on the current state of the motor to maximize future rewards. Through this strategy, the control parameters of the motor are adjusted online to form a candidate control parameter set. The candidate parameter set includes multiple optimal control schemes that are possible under the current conditions. In order to evaluate the pros and cons of these candidate control parameter sets, square error calculations are performed on them. The square error calculation is used to quantify the deviation between the candidate parameters and the ideal target, and the formula is as follows: ; in, is the square error, is the actual value, is the predicted value, is the number of samples. By calculating the square error, an evaluation index is assigned to each candidate control parameter set. Based on the parameter evaluation index, a multi-criteria decision model including parameter importance and parameter sensitivity is constructed. The multi-criteria decision model can identify the most critical control parameters by comprehensively considering the importance and sensitivity of each parameter, and screen the candidate control parameter set accordingly, and finally obtain the optimized control parameter set. Adaptive weight allocation is performed on the optimized control parameter set. Adaptive weight allocation assigns different weights according to the degree of influence of each parameter on system performance, making the final control strategy more accurate and effective. The weighted control parameter set is smoothed. Smoothing can eliminate mutations and discontinuities between parameters, ensure the smooth transition of control input and the stability of motor operation. Through these steps, the adaptive control parameter set of the motor is obtained.

[0034] In a specific embodiment, the execution step is based on the adaptive action selection strategy, the motor control parameters are adjusted online, a candidate control parameter set is obtained, and the square error calculation is performed on the candidate control parameter set to obtain the parameter evaluation index. The process may specifically include the following steps: Based on the adaptive action selection strategy, the motor control parameter space is discretized and sampled to obtain the initial parameter sampling point set, and the initial parameter sampling point set is interpolated and fitted based on the radial basis function to obtain a continuous parameter space model; The continuous parameter space model is gradient-calculated to obtain the parameter sensitivity matrix, and based on the parameter sensitivity matrix, the control parameters are iteratively optimized using the gradient descent method with an adaptive step size to obtain the local optimal parameter set; Perform a global search on the local optimal parameter set to obtain a global control parameter set, and based on the global control parameter set, construct a multidimensional search space containing upper and lower limit constraints on the parameters; Generate an unconstrained optimization problem according to the multidimensional search space, and solve the unconstrained optimization problem to obtain a candidate control parameter set; The square error between the candidate control parameter set and the target parameter value is calculated to obtain the parameter evaluation index.

[0035] Specifically, the range and resolution of the motor control parameters are defined. Based on the adaptive action selection strategy, the motor control parameter space is discretized and sampled, and these points cover the key areas of the parameter space. A series of sampling points are uniformly selected within the parameter range by setting the step size, and these sampling points constitute the initial parameter sampling point set. For brushless motors, typical control parameters may include current set values, speed set values, and torque set values, which often need to be adjusted within a certain range in practical applications. For example, the current set value may vary between 0 and 10 amperes, the speed set value may vary between 1000 and 5000 revolutions per minute, and the torque set value may vary between 0 and 100 Newton meters. The initial parameter sampling point set is interpolated and fitted based on the radial basis function to construct a continuous parameter space model. The radial basis function is a commonly used interpolation method that generates a smooth, continuous function model by fitting discrete data points. This model can not only accurately restore data at existing sampling points, but also perform reasonable interpolation between sampling points to form a complete continuous parameter space model. Assume that there is sampling points, the coordinates of each sampling point are ( ), the corresponding target value is The function form of radial basis function interpolation fitting can be expressed as: ; in, is the continuous function after fitting, is the coefficient to be determined, is a radial basis function, usually in the form of a Gaussian function, namely: ; in, represents the Euclidean distance between the input variable and the sampling point, is a parameter used to control the width of the function. In this way, a continuous model describing the motor control parameter space is obtained. The gradient calculation of the continuous parameter space model is performed to obtain the parameter sensitivity matrix. The purpose of the gradient calculation is to determine the influence of each control parameter on the objective function (such as energy consumption, efficiency or torque ripple, etc.). Each element in the sensitivity matrix represents the influence of a small change in a parameter at that point on the objective function. Assume that the objective function is , then the elements of the parameter sensitivity matrix can be expressed as: ; in, The control parameter variables, / is the partial derivative of the objective function with respect to the parameter. Based on the parameter sensitivity matrix, the control parameters are iteratively optimized using the gradient descent method with an adaptive step size, gradually approaching the local optimal solution. The gradient descent method with an adaptive step size is different from the traditional fixed step size method. It can dynamically adjust the step size according to the change of the gradient, thereby achieving a good balance between the convergence speed and the convergence accuracy. The update formula of the gradient descent is: ; in, For the The parameter vector at the iteration, is the adaptive step size, is the gradient of the objective function at the current parameter value. Through multiple iterations, the local optimal parameter set is gradually approached. A global search is performed on the local optimal parameter set to explore the entire parameter space and find a better control parameter combination. The purpose of global search is to avoid falling into the local optimal solution and find a better solution overall. Global search can be achieved through global optimization algorithms such as particle swarm optimization, genetic algorithm or simulated annealing. These algorithms can find global optimal solutions that cannot be found by local search by exploring the parameter space in a larger range. Based on the results of global search, a multidimensional search space containing upper and lower limit constraints of parameters is constructed. This search space limits the value range of each control parameter to ensure that the optimization process is physically reasonable and in line with engineering practice. For example, the values ​​of current and speed should be limited by the rated parameters of the motor, and the torque should meet the mechanical design specifications of the motor. Based on the multidimensional search space, an unconstrained optimization problem is generated and solved. Unconstrained optimization problems usually transform constrained optimization problems into unconstrained problems through the Lagrange multiplier method or the penalty function method, thereby simplifying the solution process. The candidate control parameter set obtained after the solution represents the parameter combination that is most likely to achieve the best performance under the current conditions. The square error between the candidate control parameter set and the target parameter value is calculated to quantify their advantages and disadvantages. The formula for calculating the square error is: ; in, is the square error, is the target parameter value, is the candidate control parameter value, is the total number of parameters. The smaller the square error, the closer the candidate control parameter is to the target value, and the better the optimization effect.

[0036] In a specific embodiment, the process of executing step 104 may specifically include the following steps: Predict the parameter changes in the future control time domain according to the motor adaptive control parameter set to obtain a parameter prediction sequence, and build a motor prediction model including state equations and output equations based on the parameter prediction sequence; The motor prediction model is linearized to obtain the state matrix and input matrix at each prediction moment, and based on the state matrix and input matrix, a comprehensive optimization objective function in the prediction time domain is constructed; Solve the comprehensive optimization objective function to obtain the optimal control increment at the current moment, and update the control input based on the optimal control increment to obtain the rolling optimization control sequence; The control sequence of the rolling optimization is forward simulated to obtain the predicted state trajectory, and the predicted state trajectory is corrected online to obtain the motor predictive control instruction sequence.

[0037] Specifically, the current adaptive control parameter set is used to predict and analyze the future motor operation state. The adaptive control parameter set includes key control variables of the motor such as current, speed, and torque. By performing time series analysis on these variables, their changing trends in the future period of time, namely, the parameter prediction sequence, are obtained. The prediction method can use methods such as ARIMA model, Kalman filter, or long short-term memory (LSTM) network, etc., through which the state changes of the motor in the future time domain are accurately predicted. Based on the obtained parameter prediction sequence, a motor prediction model including state equations and output equations is constructed. The state equation describes how the internal state of the motor changes with time and input signals, while the output equation reflects how the state of the motor is converted into observable output. For example, the state equation of the motor can be expressed as: ; The output equation is: ; in, For the motor at time The state vector usually includes variables such as speed, current, and torque; is the control input vector, representing the externally applied control signal; and are the state matrix and input matrix, describing the dynamic characteristics of the system; and are system noise and measurement noise, respectively, and is the output matrix, which defines the relationship between the state and the output. In order to simplify the calculation, the motor prediction model is linearized. Linearization refers to approximating the nonlinear system to a linear system near a certain operating point, so that the model can be solved using linear algebra. Through processing, the linearized state matrix and input matrix at each prediction moment are obtained. Based on the state matrix and the input matrix, a comprehensive optimization objective function in the prediction time domain is constructed. The objective function usually includes multiple optimization objectives, such as minimizing the control error, minimizing the energy consumption of the control input, and smoothing the control output. The comprehensive optimization objective function can be expressed as: ; in, is the value of the objective function, which represents the cumulative deviation and control cost in the entire prediction time domain; is the predicted state vector, is the desired reference trajectory; is the control input, is the increment of the control input; , and is a weight matrix, which is used to adjust the influence of state deviation, control input and control increment on the objective function. , optimize the control strategy of the motor so that the system can maintain the lowest energy consumption and smoothness of control while tracking the reference trajectory. Solve the comprehensive optimization objective function to obtain the optimal control increment at the current moment. This increment represents how to adjust the control input to achieve the optimal performance under the current state. The solution process usually uses optimization algorithms such as quadratic programming algorithm or gradient descent method. After obtaining the optimal control increment, the control sequence of rolling optimization is generated by updating the control input. The characteristic of rolling optimization is that the prediction model and optimization target are continuously updated over time to adjust the control strategy in real time. Forward simulation is performed on the control sequence of rolling optimization to verify the accuracy of the prediction model and analyze the response of the system in the future time domain. Forward simulation simulates the operation of the motor in the future period of time to obtain the predicted state trajectory. This trajectory reflects the state changes of the motor under different control inputs. However, due to the complexity of the actual operating environment and unpredictable disturbances, the actual state trajectory may deviate from the predicted trajectory. The predicted state trajectory is corrected online. By collecting the feedback data of the motor in real time and comparing it with the predicted results, the control strategy is dynamically adjusted to ensure that the motor is always in the optimal operating state. Through continuous correction and adjustment, the motor predictive control instruction sequence is obtained.

[0038] In a specific embodiment, the process of executing step 105 may specifically include the following steps: Feature extraction is performed on the motor predictive control instruction sequence to obtain the control parameter vector, and based on the control parameter vector, a multi-objective evaluation system including efficiency model, torque ripple model and noise model is constructed; Generate the initial population of the multi-objective evaluation system to obtain the candidate solution set, and perform non-dominated sorting and crowding calculation on the candidate solution set to obtain the initial Pareto frontier; Based on the initial Pareto frontier, the candidate solution set is iteratively optimized to obtain the optimized Pareto solution set; The membership function of each objective is calculated according to the optimized Pareto solution set, and the optimal compromise solution is constructed based on the membership function; Based on the optimal compromise solution, the brushless motor is simulated and calculated to obtain the predicted value of each performance index, and multiple groups of optimization schemes are generated based on the predicted value of each performance index; Based on the hierarchical analysis method, multiple groups of optimization schemes are comprehensively evaluated and ranked to obtain the motor multi-performance balance adjustment scheme.

[0039] Specifically, the motor predictive control instruction sequence is feature extracted to obtain a control parameter vector. The control parameter vector includes the main variables that affect the motor performance, such as current, speed, torque, etc. These parameters directly affect the key performance indicators of the motor, such as efficiency, torque ripple, and noise. For example, if the predictive control instruction sequence contains the current set value and speed set value of the motor in the future, these values ​​are extracted and formed into a control parameter vector, recorded as , where each Represents a specific control parameter. Based on the extracted control parameter vector, a multi-objective evaluation system including efficiency model, torque pulsation model and noise model is constructed. The multi-objective evaluation system evaluates the overall performance of the motor under different control conditions by integrating multiple performance indicators. The efficiency model is used to evaluate the energy utilization of the motor, the torque pulsation model measures the torque stability of the motor under load fluctuations, and the noise model is used to evaluate the noise level of the motor during operation. Each model can be described by the following formula: Efficiency model: ; in, Indicates the efficiency of the motor, is the motor output power, is the input power, are the control parameter vectors function.

[0040] Torque ripple model: ; in, represents the torque ripple, is the motor torque, is the average torque.

[0041] Noise model: ; in, is the noise level, and SPL represents the sound pressure level as a function of frequency and torque. By constructing the above multi-objective evaluation system, a comprehensive evaluation of multiple performances of the motor is performed. Based on the multi-objective evaluation system, the initial population is generated. The initial population refers to a set of solutions randomly generated in the initial stage of the optimization algorithm, which represent different combinations of control parameters. Each solution in the population can be evaluated by the aforementioned model to form a candidate solution set. The candidate solution set is non-dominated sorted and the crowding degree is calculated to determine the superiority and inferiority relationship of these solutions. Non-dominated sorting classifies candidate solutions according to their performance on each objective, while crowding degree calculation is used to measure the diversity between solutions. Through these steps, the initial Pareto frontier is obtained. Pareto frontier is a concept in multi-objective optimization, which refers to the set of solutions in a solution set that cannot be further optimized by improving a certain objective without deteriorating other objectives. Based on the initial Pareto frontier, the candidate solution set is iteratively optimized to obtain the optimized Pareto solution set. Through multiple cycles of iterative optimization, the solution set continuously approaches the true Pareto optimal frontier. This process is usually carried out using methods such as genetic algorithms or multi-objective particle swarm optimization. After obtaining the optimized Pareto solution set, the membership function of each objective is calculated. The membership function is used to quantify the pros and cons of each solution on each objective, and to describe the trade-off relationship between objectives through fuzzy logic. For example, for the efficiency objective, the membership function can be defined as: ; in, represents the membership of the solution on the efficiency target, and are the maximum and minimum values ​​of efficiency, respectively. Based on the membership function, the optimal compromise solution is constructed, that is, the solution that performs well on all objectives. Based on the optimal compromise solution, the brushless motor is simulated and calculated to predict the values ​​of each performance index. These predicted values ​​reflect the efficiency, torque pulsation and noise level of the motor under the optimal control parameters. Based on the predicted performance indicators, multiple groups of optimization schemes are generated, each group of schemes corresponds to a different combination of control parameters. In order to select the best scheme, the hierarchical analysis method is used to comprehensively evaluate and sort the multiple groups of optimization schemes. The hierarchical analysis method decomposes the complex multi-criteria decision-making problem into multiple simple judgments by constructing a hierarchical structure model, and determines the relative importance of each factor by pairwise comparison. The hierarchical analysis method generates a comprehensive score for each scheme, sorts the schemes according to these scores, and obtains the multi-performance balance adjustment scheme of the motor.

[0042] In a specific embodiment, the process of executing step 106 may specifically include the following steps: Parameter mapping is performed on the motor multi-performance balance adjustment scheme to obtain the initial configuration parameters of the digital twin model, and based on the initial configuration parameters, a motor digital twin system including the motor body, drive system and load model is constructed; Generate multiple sets of operating condition parameter sets according to the motor digital twin system, and design a test scheme matrix including steady-state and dynamic conditions based on the multiple sets of operating condition parameter sets; According to the test scheme matrix, the motor digital twin system is simulated and calculated in parallel to obtain multiple sets of simulation result data. The multiple sets of simulation result data are statistically analyzed and feature extracted to obtain the distribution characteristics of performance indicators. Based on the distribution characteristics of performance indicators, a multidimensional performance evaluation space including efficiency, torque pulsation and noise is constructed to obtain a performance evaluation model. Based on the performance evaluation model, the simulation results are comprehensively scored to obtain a ranking of the solutions. Conduct sensitivity analysis on the pros and cons of the schemes, obtain the target influencing factors, and generate optimization directions and adjustment intervals based on the target influencing factors; Based on the comparative analysis of optimization direction and adjustment range, the motor test adjustment strategy is obtained.

[0043] Specifically, the key parameters in the adjustment scheme are converted into the initial configuration parameters of the digital twin model. The multi-performance balance adjustment scheme usually contains a set of optimized motor control parameters, such as current setpoints, speed setpoints, and torque setpoints, which can ensure the best operation of the motor under different working conditions. The purpose of parameter mapping is to accurately map the optimized parameters to the various modules of the digital twin system so that the simulation system can truly reflect the physical characteristics of the motor. For example, the current setpoint can be mapped to the electromagnetic characteristic parameters in the motor body model, the speed setpoint can be mapped to the control logic in the drive system, and the torque setpoint can be mapped to the load model to simulate different mechanical load conditions. Based on the initial configuration parameters, a motor digital twin system including the motor body, drive system, and load model is constructed. The motor body model usually includes electromagnetic, mechanical, and thermal characteristics, the drive system model includes inverters, controllers, and related power electronic equipment, and the load model simulates the load conditions of the motor under actual working conditions, including factors such as inertia, resistance, and external disturbances. Through a comprehensive digital twin system, the real operating state of the motor is accurately simulated in a virtual environment. Multiple sets of working condition parameter sets are generated based on the constructed digital twin system. The operating parameter set reflects the state of the motor under different operating conditions, including different loads, speeds, ambient temperatures and other factors. These operating condition sets must cover a wide range of operating conditions to ensure a comprehensive evaluation of the motor. In order to fully test the performance of the motor, a test scheme matrix containing steady-state conditions and dynamic conditions is designed. Steady-state conditions refer to the operating state of the motor under stable load and speed, while dynamic conditions simulate non-steady-state conditions such as load mutations, acceleration and deceleration. Through this matrix, the performance of the motor under various conditions is systematically tested. According to the test scheme matrix, the motor digital twin system is simulated and calculated in parallel. Parallel simulation allows multiple operating conditions to be simulated at the same time, greatly improving the simulation efficiency. During the simulation process, the system records the output data of the motor under different operating conditions, such as current, speed, torque, efficiency and noise. Statistical analysis and feature extraction are performed on the simulation result data to identify the distribution characteristics of the motor performance indicators. These characteristics include the distribution range of efficiency, the amplitude of torque pulsation, and the spectral characteristics of noise. Through these statistical analyses, the performance of the motor under different operating conditions is understood. For example, the distribution characteristics of efficiency may reveal that the motor has low operating efficiency under certain loads, while the characteristics of torque pulsation may indicate that the output of the motor is unstable under certain working conditions. Based on the distribution characteristics of performance indicators, a multidimensional performance evaluation space including efficiency, torque pulsation and noise is constructed. The evaluation space integrates different performance indicators to form a comprehensive performance evaluation model. The performance evaluation model is used to quantitatively score different simulation results and rank the simulation schemes according to these scores. Assume that a multi-objective optimization function is defined , which can be expressed as: ; in, is the control parameter vector, is the motor efficiency, is the torque ripple, is the noise level, , and are weight coefficients, representing the importance weights of each performance indicator. By adjusting the weight coefficients, the comprehensive performance of the motor is optimized according to the actual application requirements. By comprehensively scoring the simulation results, different adjustment schemes are ranked. This ranking helps to determine which control parameter combinations can achieve the best balance in multiple performance indicators and provide a clear direction for subsequent optimization. A sensitivity analysis is performed on the ranking of the schemes to identify which control parameters have the most significant impact on the motor performance. Through sensitivity analysis, it is determined which parameters are the key factors affecting the motor performance. Based on the analysis results, the optimization direction and adjustment range are generated to determine how each parameter should be adjusted during the optimization process. Based on the determined optimization direction and adjustment range, a comparative analysis is performed. By comparing different adjustment schemes, the best performing scheme under specific working conditions is found. A test adjustment strategy for the motor is formed. This strategy guides how the motor adjusts the control parameters in actual operation to achieve the best balance of multiple performance indicators.

[0044] The above describes the test and adjustment method of the intelligent brushless motor in the embodiment of the present invention. The following describes the test and adjustment system of the intelligent brushless motor in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a test and adjustment system for an intelligent brushless motor includes: The monitoring module 201 is used to monitor the current, speed and load characteristics of the brushless motor in real time, obtain the motor operation data, and process the motor operation data to obtain the motor operation mode recognition result; A construction module 202 is used to construct a multi-time scale optimization structure including day-ahead level, hour-level and second-level based on the motor working mode recognition result, and obtain a hierarchical optimization scheme for motor parameters; An optimization module 203 is used to dynamically optimize the control parameters of the brushless motor according to the motor parameter hierarchical optimization scheme to obtain a motor adaptive control parameter set; A solution module 204 is used to perform rolling optimization solution on the control sequence of the brushless motor based on the motor adaptive control parameter set to obtain a motor predictive control instruction sequence; The calculation module 205 is used to perform multi-objective optimization calculations on the efficiency, torque ripple, and noise of the brushless motor based on the motor predictive control instruction sequence to obtain a motor multi-performance balance adjustment scheme; The verification module 206 is used to integrate the motor multi-performance balance adjustment scheme into the digital twin test platform, simulate and verify the performance of the brushless motor under different working conditions, and obtain the motor test adjustment strategy.

[0045] Through the collaboration of the above components, the real-time monitoring of the motor current, speed and load characteristics can accurately identify the working mode of the motor, laying the foundation for subsequent parameter optimization. A multi-time scale optimization structure including day-ahead, hourly and second-level is constructed to improve the accuracy and adaptability of motor parameter adjustment. Reinforcement learning and deep Q network are used to realize dynamic optimization and adaptive adjustment of motor control parameters, and improve the adaptability and operating efficiency of the motor under different working conditions. By rolling optimization and solving the motor control sequence, combined with online correction of the predicted state trajectory, accurate prediction and control of the future operating state of the motor are achieved, and the stability and reliability of the system are improved. The multi-objective evaluation system and Pareto optimal solution strategy are adopted to achieve comprehensive optimization of multiple performance indicators such as motor efficiency, torque pulsation and noise, and achieve the goal of multi-performance balance. The optimization scheme is integrated into the digital twin test platform, and the performance verification and optimization of the motor under different working conditions are realized through parallel simulation calculation and multi-dimensional performance evaluation, which improves the test efficiency and reliability. From data acquisition, pattern recognition to parameter optimization and test verification, a complete closed-loop management system is formed, realizing the full process intelligence and automation of the brushless motor test and adjustment process.

[0046] above Figure 2 The test and adjustment system of the intelligent brushless motor in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The test and adjustment device of the intelligent brushless motor in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0047] Figure 3It is a structural schematic diagram of a test and adjustment device for an intelligent brushless motor provided by an embodiment of the present invention. The test and adjustment device 300 for the intelligent brushless motor may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the test and adjustment device 300 for the intelligent brushless motor. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the test and adjustment device 300 for the intelligent brushless motor to implement the steps of the test and adjustment method for the intelligent brushless motor described above.

[0048] The intelligent brushless motor test and adjustment device 300 may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3 The structure of the test and adjustment device for the intelligent brushless motor shown does not constitute a limitation on the test and adjustment device for the intelligent brushless motor provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0049] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the test and adjustment method of the intelligent brushless motor.

[0050] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0051] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0052] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A test and adjustment method for an intelligent brushless motor, characterized in that: The method comprises: Monitor the current, speed and load characteristics of the brushless motor in real time to obtain motor operation data, and process the motor operation data to obtain a motor operation mode recognition result; Based on the motor working mode recognition results, a multi-time scale optimization structure including day-of-day, hour-level and second-level is constructed to obtain a hierarchical optimization scheme for motor parameters; According to the motor parameter hierarchical optimization scheme, the control parameters of the brushless motor are dynamically optimized to obtain a motor adaptive control parameter set; Based on the motor adaptive control parameter set, a rolling optimization solution is performed on the control sequence of the brushless motor to obtain a motor predictive control instruction sequence; Based on the motor predictive control instruction sequence, a multi-objective optimization calculation is performed on the efficiency, torque ripple, and noise of the brushless motor to obtain a motor multi-performance balance adjustment scheme; The motor multi-performance balance adjustment scheme is integrated into the digital twin test platform, and the performance of the brushless motor under different working conditions is simulated and verified to obtain the motor test adjustment strategy.

2. The test and adjustment method of the intelligent brushless motor according to claim 1, characterized in that: The real-time monitoring of the current, speed and load characteristics of the brushless motor to obtain motor operation data, and processing the motor operation data to obtain a motor operation mode recognition result, includes: Performing sliding time window segmentation on the current, speed, and load characteristic data of the brushless motor to obtain an initial feature sequence, and performing fast Fourier transform on the initial feature sequence to obtain a frequency domain feature matrix; Based on the frequency domain feature matrix, the mutual information coefficient between each feature is calculated to obtain a feature correlation matrix, and the feature correlation matrix is ​​sorted by feature importance to obtain a reduced-dimensional feature set; Performing fuzzy reasoning on the dimension-reduced feature set to obtain a membership value of each working mode, and calculating fuzzy entropy based on the membership value to obtain a fuzzy clustering effectiveness index; According to the fuzzy clustering validity index, the dimension reduction feature set is adaptively threshold segmented to obtain a set of candidate working modes, and time domain statistical features and wavelet packet energy features are extracted for each mode in the set of candidate working modes to obtain a multidimensional feature vector; Performing classification prediction on the multidimensional feature vector to obtain an initial working mode recognition result, and calculating the confidence of the initial working mode recognition result to obtain a confidence sequence; The confidence sequence is subjected to sliding window averaging and threshold judgment to obtain a stable working mode recognition result, and based on the stable working mode recognition result, the motor is divided into a constant torque mode and a constant power mode to obtain a motor working mode recognition result.

3. The test and adjustment method of the intelligent brushless motor according to claim 1, characterized in that: Based on the motor working mode recognition result, a multi-time scale optimization structure including day-of-day, hour-level and second-level is constructed to obtain a hierarchical optimization scheme for motor parameters, including: Decomposing the motor working mode recognition result into a time series to obtain a trend item, a period item, and a random item, and predicting the motor working mode for the next 24 hours based on the trend item and the period item to obtain a day-ahead working mode prediction sequence; Based on the previous stage working mode prediction sequence, cluster analysis is performed on the historical operating data of the motor to obtain a typical working condition set, and a feature vector is extracted for each working condition in the typical working condition set to obtain a working condition feature matrix; Probabilistic modeling is performed on the operating condition characteristic matrix to obtain a day-ahead optimization objective function, and based on the day-ahead optimization objective function, preliminary optimization is performed on basic motor operating parameters to obtain a day-ahead parameter optimization solution; Subdividing the day-ahead parameter optimization scheme into time scales to obtain hourly parameter optimization initial values, and constructing a model predictive control objective function based on the hourly parameter optimization initial values ​​to obtain an hourly dynamic optimization model; Solving the hourly dynamic optimization model to obtain an hourly optimal control sequence, and generating a second-level optimization target based on the hourly optimal control sequence; A second-level parameter fine-tuning strategy is generated according to the second-level optimization target, and the day-ahead parameter optimization scheme, the hour-level optimal control sequence and the second-level parameter fine-tuning strategy are integrated to obtain a motor parameter hierarchical optimization scheme.

4. The test and adjustment method of the intelligent brushless motor according to claim 3, characterized in that: The day-ahead parameter optimization scheme is subdivided into time scales to obtain hourly parameter optimization initial values, and based on the hourly parameter optimization initial values, a model predictive control objective function is constructed to obtain an hourly dynamic optimization model, including: Performing equal-interval sampling on the day-ahead parameter optimization scheme to obtain 24 hourly time nodes, and segmenting the motor operation data based on the 24 hourly time nodes to obtain an hourly parameter optimization data set; Creating an hourly-level dynamic model of the motor according to the hourly-level parameter optimization data set, and extracting state variables and control variables based on the hourly-level dynamic model of the motor; Constructing a state trajectory equation and an output equation in a prediction time domain according to the state variables and the control variables to obtain a prediction model, and constructing a quadratic performance indicator function including a state deviation, a control deviation, and a control increment based on the prediction model; Introducing a soft constraint term into the quadratic performance index function to obtain an extended objective function that considers state constraints and control constraints, and converting the extended objective function into an unconstrained optimization problem to obtain a model predictive control objective function; Based on the model predictive control objective function, an hourly dynamic optimization model including a state predictor, a controller and a feedback corrector is constructed.

5. The test and adjustment method of the intelligent brushless motor according to claim 1, characterized in that: The method of dynamically optimizing the control parameters of the brushless motor according to the motor parameter hierarchical optimization scheme to obtain a motor adaptive control parameter set includes: Decoupling the motor parameter hierarchical optimization scheme to obtain day-level, hour-level, and second-level optimization parameter subsets, and constructing a motor dynamic mathematical model including motor current, speed, and torque based on the optimization parameter subsets; Performing reinforcement learning training on the motor dynamic mathematical model to obtain an initial Q-value function, and constructing a deep Q-network including a target network and an evaluation network based on the initial Q-value function; Performing experience replay training on the deep Q network to obtain an optimized Q-value function, and generating an adaptive action selection strategy based on the optimized Q-value function; Based on the adaptive action selection strategy, the motor control parameters are adjusted online to obtain a candidate control parameter set, and the square error of the candidate control parameter set is calculated to obtain a parameter evaluation index; Based on the parameter evaluation index, a multi-criteria decision model including parameter importance and parameter sensitivity is constructed, and the candidate control parameter set is screened based on the multi-criteria decision model to obtain an optimized control parameter set; Adaptive weight distribution is performed on the optimized control parameter set to obtain a weighted control parameter set, and smoothing is performed on the weighted control parameter set to obtain a motor adaptive control parameter set.

6. The test and adjustment method of the intelligent brushless motor according to claim 5, characterized in that: The method of adjusting the motor control parameters online based on the adaptive action selection strategy to obtain a candidate control parameter set, and performing square error calculation on the candidate control parameter set to obtain a parameter evaluation index includes: Based on the adaptive action selection strategy, the motor control parameter space is discretized and sampled to obtain an initial parameter sampling point set, and the initial parameter sampling point set is interpolated and fitted based on a radial basis function to obtain a continuous parameter space model; Performing gradient calculation on the continuous parameter space model to obtain a parameter sensitivity matrix, and iteratively optimizing the control parameters based on the parameter sensitivity matrix and a gradient descent method with an adaptive step size to obtain a local optimal parameter set; Performing a global search on the local optimal parameter set to obtain a global control parameter set, and constructing a multidimensional search space containing upper and lower parameter constraints based on the global control parameter set; generating an unconstrained optimization problem according to the multidimensional search space, and solving the unconstrained optimization problem to obtain a set of candidate control parameters; The square error between the candidate control parameter set and the target parameter value is calculated to obtain a parameter evaluation index.

7. The test and adjustment method of the intelligent brushless motor according to claim 1, characterized in that: The step of performing rolling optimization on the control sequence of the brushless motor based on the motor adaptive control parameter set to obtain a motor predictive control instruction sequence includes: Predicting parameter changes in a future control time domain according to the motor adaptive control parameter set to obtain a parameter prediction sequence, and constructing a motor prediction model including a state equation and an output equation based on the parameter prediction sequence; Linearizing the motor prediction model to obtain a state matrix and an input matrix at each prediction moment, and constructing a comprehensive optimization objective function in the prediction time domain based on the state matrix and the input matrix; Solving the comprehensive optimization objective function to obtain the optimal control increment at the current moment, and updating the control input based on the optimal control increment to obtain a rolling optimization control sequence; A forward simulation is performed on the rolling optimized control sequence to obtain a predicted state trajectory, and the predicted state trajectory is corrected online to obtain a motor predictive control instruction sequence.

8. The test and adjustment method of the intelligent brushless motor according to claim 1, characterized in that: The method of performing multi-objective optimization calculation on the efficiency, torque ripple, and noise of the brushless motor based on the motor predictive control instruction sequence to obtain a motor multi-performance balance adjustment scheme includes: Extracting features from the motor predictive control instruction sequence to obtain a control parameter vector, and constructing a multi-objective evaluation system including an efficiency model, a torque ripple model, and a noise model based on the control parameter vector; Generating an initial population for the multi-objective evaluation system to obtain a candidate solution set, and performing non-dominated sorting and congestion calculation on the candidate solution set to obtain an initial Pareto frontier; Based on the initial Pareto front, iteratively optimize the candidate solution set to obtain an optimized Pareto solution set; Calculating the membership function of each objective according to the optimized Pareto solution set, and constructing an optimal compromise solution based on the membership function; Performing simulation calculations on the brushless motor based on the optimal compromise solution to obtain predicted values ​​of various performance indicators, and generating multiple groups of optimization solutions based on the predicted values ​​of various performance indicators; Based on the hierarchical analysis method, the multiple groups of optimization schemes are comprehensively evaluated and ranked to obtain the motor multi-performance balance adjustment scheme.

9. The test and adjustment method of the intelligent brushless motor according to claim 1, characterized in that: The motor multi-performance balance adjustment scheme is integrated into the digital twin test platform, and the performance of the brushless motor under different working conditions is simulated and verified to obtain the motor test adjustment strategy, including: Parameter mapping is performed on the motor multi-performance balance adjustment scheme to obtain initial configuration parameters of the digital twin model, and based on the initial configuration parameters, a motor digital twin system including a motor body, a drive system, and a load model is constructed; Generating multiple sets of operating condition parameter sets according to the motor digital twin system, and designing a test scheme matrix including steady-state operating conditions and dynamic operating conditions based on the multiple sets of operating condition parameter sets; According to the test scheme matrix, parallel simulation calculations are performed on the motor digital twin system to obtain multiple groups of simulation result data, and statistical analysis and feature extraction are performed on the multiple groups of simulation result data to obtain performance indicator distribution characteristics; Based on the distribution characteristics of the performance indicators, a multidimensional performance evaluation space including efficiency, torque ripple, and noise is constructed to obtain a performance evaluation model. The simulation results are comprehensively scored based on the performance evaluation model to obtain a ranking of the solutions. Perform sensitivity analysis on the pros and cons of the schemes to obtain target influencing factors, and generate optimization directions and adjustment intervals based on the target influencing factors; A comparative analysis is performed based on the optimization direction and the adjustment interval to obtain a motor test adjustment strategy.

10. A test and adjustment system for an intelligent brushless motor, characterized in that: A system for performing a test and adjustment method for an intelligent brushless motor according to any one of claims 1 to 9, the system comprising: A monitoring module is used to monitor the current, speed and load characteristics of the brushless motor in real time to obtain motor operation data, and process the motor operation data to obtain a motor operation mode recognition result; A construction module is used to construct a multi-time scale optimization structure including day-of-day, hour-level and second-level based on the motor working mode recognition result to obtain a hierarchical optimization scheme for motor parameters; an optimization module, configured to dynamically optimize the control parameters of the brushless motor according to the motor parameter hierarchical optimization scheme to obtain a motor adaptive control parameter set; A solution module, configured to perform rolling optimization solution on the control sequence of the brushless motor based on the motor adaptive control parameter set to obtain a motor predictive control instruction sequence; a calculation module, configured to perform multi-objective optimization calculations on the efficiency, torque ripple, and noise of the brushless motor based on the motor predictive control instruction sequence, and obtain a motor multi-performance balance adjustment scheme; A verification module is used to integrate the motor multi-performance balance adjustment scheme into the digital twin test platform, simulate and verify the performance of the brushless motor under different working conditions, and obtain the motor test adjustment strategy.

Images