Self-adaptive brushless motor control method and system

Abstract

The invention discloses a self-adaptive brushless motor control method and system, and relates to the field of intelligent control, and the method comprises the steps: collecting the original data of the operation state of a motor through a sensor group, and carrying out the preprocessing; time-varying parameter identification is completed through combination of an extended Kalman filtering algorithm and a radial basis function neural network, an evaluation index system is constructed based on an analytic hierarchy process to obtain a comprehensive evaluation value, and a related trend is predicted through a long and short-term memory neural network; constructing a multi-modal control strategy library, determining an adaptive strategy, optimizing core parameters by using an improved particle swarm optimization algorithm, generating a control instruction, and outputting a corresponding current through a power driving module; and monitoring motor parameters in real time, comparing with a control target value, calculating deviation, correcting an identification result, adjusting a strategy threshold value, and updating and optimizing an objective function. The method has the advantages that by accurately sensing the state of the motor, dynamically adapting the control strategy and optimizing parameters in real time, it is ensured that the motor stably and efficiently operates under the complex working condition, and the characteristics of energy conservation and long service life are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent control, in particular to a self-adaptive brushless motor control method and system. BACKGROUND

[0002] Brushless motors have been widely used in new energy vehicles, industrial automation, drones, smart homes and other fields due to their high efficiency, long service life, low noise and other advantages. However, the demand for dynamic performance and robustness under complex working conditions is becoming increasingly stringent. With the development of device miniaturization and integration, higher requirements are placed on the real-time performance, self-adaptive ability and compatibility of the control system, and it is urgent to break through the limitations of traditional control frameworks.

[0003] Current brushless motor control methods on the market cannot adapt to complex and variable working condition requirements. Most methods rely on a single control strategy, such as traditional vector control or PID control, and lack multi-modal adaptation capability. When switching between different modes such as light load, heavy load and braking, response lag or precision decline problems may occur. In terms of parameter identification, single Kalman filtering or fixed models are commonly used, which cannot effectively compensate for parameter nonlinear changes caused by temperature drift and magnetic saturation, resulting in insufficient identification accuracy and affecting control effectiveness. State assessment focuses on a single operating indicator and lacks a comprehensive assessment system. It also lacks forward-looking prediction of load changes and parameter drift, making it difficult to avoid fault risks in advance. Control parameters are mostly offline set or simply iteratively optimized, without introducing adaptive mechanisms and global optimization algorithms, which can easily fall into local optima and lack stability in the case of load mutation or interference. In addition, some methods ignore the braking energy recovery function, resulting in low energy utilization efficiency. At the same time, they lack a closed-loop iterative optimization mechanism, which cannot dynamically correct the model and control threshold based on real-time errors. The overall adaptive ability, robustness and intelligence of the control are lower than those of the proposed method. SUMMARY

[0004] To improve existing methods and systems, a self-adaptive brushless motor control method and system are provided. The method achieves full-link adaptive control through multi-technology fusion, accurately senses motor state, dynamically adapts control strategy and optimizes parameters in real time, ensuring stable and efficient operation under complex working conditions, and has energy-saving and long-life characteristics.

[0005] To achieve the above objectives, the technical solution adopted by the present application is as follows:

[0006] A self-adaptive brushless motor control method, comprising:

[0007] Real-time acquisition of motor operating state parameters is performed by a sensor group, and the acquired raw data is preprocessed to obtain standardized state monitoring data;

[0008] Based on the standardized monitoring data, the extended Kalman filter algorithm is used to establish the motor state equation and the observation equation, the output results are input into the radial basis function neural network for error correction, and the real-time motor time-varying parameter identification result is obtained;

[0009] Based on the standardized monitoring data and the parameter identification result, a motor operating state evaluation index system is constructed, the comprehensive state evaluation value is calculated by the analytic hierarchy process, the long short-term memory neural network is used to predict the load change trend, the parameter drift rule and the state evaluation value evolution direction in the future time period, and the trend prediction result is generated;

[0010] Based on the comprehensive state evaluation value and the trend prediction result, a multi-modal control strategy library is constructed, and according to the mode and the comprehensive state evaluation value of the motor, the control strategy is determined to dynamically adjust the motor parameters through different algorithms;

[0011] Based on the trend prediction result and the selected control strategy, the improved particle swarm optimization algorithm is used to introduce adaptive inertia weight and chaotic mutation operator to online optimize the core parameters of the control strategy, and the optimal control parameter combination is obtained;

[0012] Based on the optimal control parameters and the selected control strategy, the corresponding control instructions are generated, the driving signals generated based on the control instructions are transmitted to the power driving module, and the three-phase sinusoidal current or braking feedback current is output;

[0013] Real-time monitoring of motor parameter data and comparison with motor control target value, calculation of speed error, current error and state evaluation deviation, based on the deviation correction identification result, adjustment of control strategy switching threshold and dynamic update of optimization objective function.

[0014] Preferably, the real-time acquisition of motor operating state parameters by the sensor group and the preprocessing of the collected raw data to obtain standardized state monitoring data specifically includes:

[0015] Based on the calibrated sensor group, the rotor real-time speed, three-phase stator instantaneous current, stator winding temperature, motor shell vibration amplitude and frequency, and DC bus voltage motor operating key parameters are collected in real time;

[0016] According to the interference characteristics of different types of data, the corresponding preprocessing method is adopted, the current and speed signals susceptible to electromagnetic interference are processed by adaptive Kalman filter to eliminate power grid noise and electromagnetic interference, and the temperature and vibration signals with random fluctuations are processed by sliding average filter to weaken random errors;

[0017] The filtered raw data are converted into standardized state monitoring data meeting the requirements of subsequent processing.

[0018] Preferably, the standardized monitoring data is used to establish motor state equations and observation equations by using an extended Kalman filter algorithm, the output results are input into a radial basis function neural network for error correction, and real-time motor time-varying parameter identification results are obtained, which specifically include:

[0019] Based on the standardized monitoring data, the motor state equations and observation equations are constructed by using the extended Kalman filter algorithm, the speed and three-phase current are used as observation inputs, and the stator resistance, d / q axis inductance and back electromotive force coefficient are preliminarily estimated.

[0020] The preliminary estimation results of the EKF and the winding temperature data are input into the radial basis function neural network, the EKF estimation error is corrected through online learning, and the stator resistance drift caused by temperature and the inductance nonlinear change caused by magnetic saturation are compensated.

[0021] The control identification period is consistent with the sampling period, and the real-time parameter identification results after error correction and disturbance compensation are finally output.

[0022] Preferably, the standardized monitoring data and parameter identification results are used to construct a motor operating state evaluation index system, the comprehensive state evaluation value is calculated by using the analytic hierarchy process, the long short-term memory neural network is used to predict the load change trend, parameter drift rule and state evaluation value evolution direction in the future time period, and the trend prediction result is generated, which specifically includes:

[0023] Based on the standardized monitoring data and parameter identification results, a motor operating state evaluation index system is constructed, and the index system includes the speed fluctuation coefficient, current harmonic distortion rate, winding temperature deviation value, vibration amplitude exceeding times and parameter drift rate.

[0024] The weights of each index are determined by using the analytic hierarchy process, the comprehensive state evaluation value is calculated in combination with the actual values of the indexes, and according to the preset threshold, the current operating mode of the motor is distinguished, including light load, rated load, heavy load and braking, and the health state is determined, including normal, sub-health and pre-fault.

[0025] The long short-term memory neural network is used to take the historical monitoring data and identification parameters as training samples to predict the load change trend, parameter drift rule and state evaluation value evolution direction in the future time period.

[0026] The comprehensive state evaluation result, operating mode and health state determination conclusion are integrated, and the trend prediction data is superimposed to form a complete operating state evaluation and trend prediction result.

[0027] Preferably, a multi-modal control strategy library is constructed based on the comprehensive state evaluation value and the trend prediction result, and the control strategy is determined according to the mode of the motor and the comprehensive state evaluation value to dynamically adjust the motor parameters by using different algorithms, which specifically includes:

[0028] Based on the running state evaluation and trend prediction results, a strategy library including an optimized vector control algorithm, a model predictive control algorithm, a PID-sliding mode composite control algorithm and a braking energy recovery control algorithm is built;

[0029] Based on the current operating mode and health state of the motor, the control strategy is matched, the MPC algorithm is started under light load and normal conditions to optimize the voltage vector with the goal of reducing energy consumption and current harmonics, the FOC algorithm is used under rated load to ensure speed stability and control accuracy, the PID-sliding mode composite control algorithm is used under heavy load or sub-health to ensure steady state through PID and anti-disturbance through sliding mode, and the energy recovery control algorithm is used in braking mode to coordinate motor power generation and inverter feedback.

[0030] Preferably, based on the trend prediction results and the selected control strategy, the core parameters of the control strategy are optimized online by introducing adaptive inertia weight and chaotic mutation operator into the improved particle swarm optimization algorithm to obtain the optimal control parameter combination, which specifically includes:

[0031] Based on the load and state prediction results and the selected control strategy, the core parameters of the control strategy are optimized using the improved particle swarm optimization algorithm;

[0032] For the MPC algorithm, the prediction time domain length, control time domain length and weight coefficient are optimized, for the vector control algorithm, the d / q axis current loop PI regulator parameters are optimized, and for the PID-sliding mode composite control algorithm, the PID gain, sliding mode surface coefficient and exponential reaching law parameters are optimized;

[0033] The improved PSO algorithm introduces adaptive inertia weight and chaotic mutation operator, uses large inertia weight for global search in the early iteration to obtain optimal solution in a large range, reduces inertia weight in the later iteration to focus on local search, breaks out of local optimal trap through chaotic mutation to continuously approach the target value, and obtains the optimal control parameter combination.

[0034] Preferably, based on the optimal control parameters and the selected control strategy, corresponding control instructions are generated, drive signals generated based on the control instructions are transmitted to the power drive module, and three-phase sinusoidal current or braking feedback current is outputted, which specifically includes:

[0035] For the vector control, MPC and PID-sliding mode composite control strategies, adaptive SVPWM algorithm is used, combined with motor rotor position information and optimal control parameters, three-phase voltage values are calculated, voltage vector sectors are divided, and high-frequency PWM drive signals are generated;

[0036] For the braking energy recovery control strategy, energy feedback control signals are directly generated to adjust the inverter to energy feedback working mode;

[0037] The generated PWM driving signal or energy feedback control signal is transmitted to the power driving module, which operates at a preset switching frequency and duty cycle to output three-phase sinusoidal current or braking feedback current.

[0038] Preferably, the motor parameters are monitored in real time and compared with the motor control target value to calculate the speed error, current error and state evaluation deviation, and the identification result is corrected, the control strategy switching threshold is adjusted and the optimization target function is dynamically updated based on the deviation.

[0039] The motor monitoring data are collected in real time and compared with the preset control target value to calculate the speed error, current error and state evaluation deviation.

[0040] The calculated speed error, current error and state evaluation deviation are used as feedback signals for data feedback.

[0041] According to the feedback deviation, the identification result is corrected, the strategy switching threshold is adjusted and the optimization target function is dynamically updated, and the above deviation calculation, feedback transmission and adjustment process are repeated to realize continuous iterative optimization of control performance.

[0042] Further, an adaptive brushless motor control system is proposed, which comprises:

[0043] Parameter identification module: based on standardized monitoring data, the extended Kalman filter algorithm is used to establish equations to preliminarily estimate parameters, and the radial basis function neural network error correction is used to output real-time motor time-varying parameter identification results;

[0044] State evaluation and trend prediction module: an evaluation index system is constructed, the comprehensive state evaluation value is calculated by the analytic hierarchy process, and the future state trend is predicted by the long short-term memory neural network to form complete evaluation and prediction results;

[0045] Control strategy decision module: based on the comprehensive state evaluation value and the trend prediction result, a multi-modal control strategy library is constructed, and the control strategy is matched and determined according to the motor operation mode and health state;

[0046] Control parameter optimization module: combined with the trend prediction result and the selected control strategy, an improved particle swarm optimization algorithm with adaptive inertia weight and chaos mutation operator is used to optimize the core parameters to obtain the optimal combination;

[0047] Driving signal generation module: according to the optimal control parameters and the selected control strategy, the PWM driving signal or energy feedback control signal is generated and transmitted to the power driving module to output three-phase sinusoidal current or braking feedback current;

[0048] Feedback and iterative optimization module: real-time monitoring of motor parameters and comparison with target value to calculate deviation, based on deviation correction identification results, adjustment strategy switching threshold and optimization objective function;

[0049] Processor: the processor is used for processing the calculation process of each formula and the construction calculation process of each model.

[0050] Compared with the prior art, the advantages of the present application are that:

[0051] Through multi-sensor data acquisition and targeted filtering pretreatment, the reliability of the state data is ensured; combined with the parameter identification scheme of extended Kalman filter and radial basis function neural network, the temperature drift and magnetic saturation nonlinear influence are effectively compensated, and the time-varying parameter identification accuracy is improved; based on the evaluation and prediction mechanism of analytic hierarchy process and long short-term memory neural network, the motor operation mode and health state can be accurately determined, and the load change and parameter evolution trend can be predicted in advance, providing forward-looking support for control decision. The multi-modal control strategy library can flexibly switch the adaptive inertia weight and chaotic mutation operator to realize online optimization of core control parameters, taking into account control accuracy and dynamic response speed; the adaptive SVPWM technology and braking energy recovery function improve the energy utilization efficiency, and the closed-loop feedback iteration mechanism continuously optimizes the control performance by real-time correction of the model, adjustment of the threshold and update of the objective function, so that the motor can still maintain stable operation, high efficiency and long service life characteristics under complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 The method proposed by the present application is shown in the schematic diagram;

[0053] Figure 2 The schematic diagram for obtaining standardized state monitoring data proposed by the present application is shown in the schematic diagram;

[0054] Figure 3 The schematic diagram for obtaining real-time motor time-varying parameter identification results proposed by the present application is shown in the schematic diagram;

[0055] Figure 4 The schematic diagram for generating trend prediction results proposed by the present application is shown in the schematic diagram;

[0056] Figure 5 The schematic diagram for dynamic adjustment of motor parameters proposed by the present application is shown in the schematic diagram;

[0057] Figure 6 The schematic diagram for obtaining optimal control parameter combination proposed by the present application is shown in the schematic diagram;

[0058] Figure 7 The schematic diagram for generating control instructions proposed by the present application is shown in the schematic diagram;

[0059] Figure 8An error calculation schematic diagram is provided for the present application. DETAILED DESCRIPTION

[0060] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only used as examples, and other obvious modifications can be thought of by those skilled in the art.

[0061] An adaptive brushless motor control system comprises:

[0062] A parameter identification module: based on standardized monitoring data, a preliminary parameter estimation equation is established using an extended Kalman filter algorithm, and a radial basis function neural network error correction is performed to output real-time motor time-varying parameter identification results;

[0063] A state evaluation and trend prediction module: an evaluation index system is constructed, an analytic hierarchy process is used to calculate a comprehensive state evaluation value, a long short-term memory neural network is used to predict future state trends, and complete evaluation and prediction results are formed;

[0064] A control strategy decision module: a multi-modal control strategy library is constructed based on the comprehensive state evaluation value and the trend prediction result, and a control strategy is matched and determined according to the motor operating mode and the health state;

[0065] A control parameter optimization module: combined with the trend prediction result and the selected control strategy, an improved particle swarm optimization algorithm with adaptive inertia weight and chaos mutation operator is used to optimize the core parameters to obtain the optimal combination;

[0066] A drive signal generation module: according to the optimal control parameters and the selected control strategy, PWM drive signals or energy feedback control signals are generated and transmitted to the power drive module to output three-phase sinusoidal currents or braking feedback currents;

[0067] A feedback and iterative optimization module: real-time motor parameters are monitored and compared with target values to calculate deviations, and the identification results are corrected based on the deviations, the strategy switching threshold is adjusted, and the objective function is optimized;

[0068] A processor: the processor is used to process the calculation process of each formula and the construction and calculation process of each model.

[0069] Referring to Figure 1 An adaptive brushless motor control method is shown, comprising:

[0070] Step one: real-time acquisition of motor operating state parameters is performed through a sensor group, and the acquired raw data is preprocessed to obtain standardized state monitoring data;

[0071] Step two: Based on standardized monitoring data, the extended Kalman filter algorithm is used to establish the motor state equation and observation equation, and the output results are input into the radial basis function neural network for error correction to obtain real-time motor time-varying parameter identification results;

[0072] Step three: Based on the standardized monitoring data and parameter identification results, a motor operating state evaluation index system is constructed, the comprehensive state evaluation value is calculated by the analytic hierarchy process, the long short-term memory neural network is used to predict the load change trend, parameter drift rule and state evaluation value evolution direction in the future time period, and the trend prediction result is generated;

[0073] Step four: Based on the comprehensive state evaluation value and the trend prediction result, a multi-modal control strategy library is constructed, and according to the mode and comprehensive state evaluation value of the motor, the control strategy is determined to dynamically adjust the motor parameters through different algorithms;

[0074] Step five: Based on the trend prediction result and the selected control strategy, the improved particle swarm optimization algorithm is used to introduce adaptive inertia weight and chaos mutation operator to optimize the core parameters of the control strategy online to obtain the optimal control parameter combination;

[0075] Step six: Based on the optimal control parameters and the selected control strategy, the corresponding control instructions are generated, the drive signals generated based on the control instructions are transmitted to the power drive module, and the three-phase sinusoidal current or braking feedback current is output;

[0076] Step seven: Real-time monitoring of motor parameter data and comparison with motor control target value, calculation of speed error, current error and state evaluation deviation, based on the deviation correction identification result, adjustment of control strategy switching threshold and dynamic update of optimization objective function.

[0077] Referring to Figure 2 As shown, the motor operating state parameters are collected in real time by the sensor group, and the collected raw data is preprocessed to obtain standardized state monitoring data, which specifically includes:

[0078] Based on the calibrated sensor group, the rotor real-time speed, three-phase stator instantaneous current, stator winding temperature, motor housing vibration amplitude and frequency, and DC bus voltage motor operating key parameters are collected in real time;

[0079] According to the interference characteristics of different types of data, the corresponding preprocessing method is adopted, the current and speed signals susceptible to electromagnetic interference are processed by adaptive Kalman filter to eliminate power grid noise and electromagnetic interference, and the temperature and vibration signals with random fluctuations are processed by sliding average filter to weaken random errors;

[0080] The filtered raw data is converted into standardized state monitoring data that meets the requirements of subsequent processing.

[0081] Referring toFigure 3 As shown, based on the standardized monitoring data, the extended Kalman filter algorithm is used to establish the motor state equation and observation equation, the output results are input into the radial basis function neural network for error correction, and the real-time motor time-varying parameter identification result is obtained, which specifically includes:

[0082] Based on the standardized monitoring data, the extended Kalman filter algorithm is used to construct the motor state equation and observation equation, the speed, three-phase current is used as the observation input, and the stator resistance, d / q axis inductance and back electromotive force coefficient are preliminarily estimated;

[0083] The preliminary estimation result of EKF and the winding temperature data are input into the radial basis function neural network, the EKF estimation error is corrected through online learning, and the stator resistance drift caused by temperature and the inductance nonlinear change caused by magnetic saturation are compensated;

[0084] The control identification period is consistent with the sampling period, and the real-time parameter identification result after error correction and disturbance compensation is finally output.

[0085] Specifically, based on the motor mathematical model, the state equation and observation equation required by EKF are established, wherein the state equation describes the change law of motor parameters with running state, and the observation equation relates the actual collected speed, current data and parameter estimation value; The screened speed, three-phase current data are input into the EKF algorithm, through the two core links of prediction and update, the stator resistance, d / q axis inductance and back electromotive force coefficient are preliminarily estimated, a group of parameter estimation values to be corrected are obtained, and the error fluctuation range in the EKF estimation process is output;

[0086] The preliminary parameter estimation value output by EKF and the stator winding temperature data are used as the input layer variables of RBF-NN, wherein the temperature data are used to quantify the influence degree of temperature on the stator resistance; The difference between the ideal parameter value and the EKF estimation value is set as the output layer target of RBF-NN; The correlation strength between the input variables and the output correction amount is analyzed through the mutual information criterion, and the number of RBF-NN hidden layer nodes is adjusted adaptively;

[0087] Based on the EKF estimation value-actual parameter value-temperature sample data accumulated in the historical operation of the motor, the RBF-NN is initialized and trained to ensure that the network has basic correction ability; In the real-time identification process, the new estimation value of EKF and the current temperature data are continuously input into RBF-NN, and the parameter correction amount output by RBF-NN is used to dynamically adjust the preliminary estimation value of EKF;

[0088] The hidden layer output equation formula of the radial basis function neural network is:

[0089]

[0090] The output layer output equation formula is:

[0091]

[0092] The weight update equation is:

[0093]

[0094] wherein, is the output of the jth hidden layer node, is the input vector, is the center vector of the jth hidden layer node, is the width parameter of the jth hidden layer node, is the number of hidden layer nodes, is the parameter correction vector of the RBF-NN output, is the connection weight from the hidden layer to the output layer, is the output layer bias, is the learning rate, is the error;

[0095] The EKF estimation error range of the parameter value corrected by the RBF-NN is compared, if the corrected parameter value is in the reasonable error interval, the parameter value is directly output as the final identification result; if it is out of the error interval, the RBF-NN learning process is triggered again, and the network weight is adjusted for correction again until the parameter value meets the accuracy requirement.

[0096] Referring to Figure 4 , a motor operation state evaluation index system is constructed based on the standardized monitoring data and the parameter identification result, a comprehensive state evaluation value is calculated by an analytic hierarchy process, a long short-term memory neural network is used to predict the load change trend, parameter drift rule and state evaluation value evolution direction in a future time period, and a trend prediction result is generated, specifically including:

[0097] A motor operation state evaluation index system is constructed based on the standardized monitoring data and the parameter identification result, and the index system includes a speed fluctuation coefficient, a current harmonic distortion rate, a winding temperature deviation value, a vibration amplitude exceeding times and a parameter drift rate;

[0098] The weights of each index are determined by an analytic hierarchy process, a comprehensive state evaluation value is calculated in combination with the actual values of the indexes, and according to a preset threshold, the current operation mode of the motor is distinguished, including light load, rated load, heavy load and braking, and the health state is determined, including normal, sub-health and pre-fault;

[0099] A long short-term memory neural network is used to predict the load change trend, parameter drift rule and state evaluation value evolution direction in a future time period, using historical monitoring data and identified parameters as training samples;

[0100] The integrated comprehensive state evaluation result, operation mode and health state determination conclusion are superimposed with trend prediction data to form complete operation state evaluation and trend prediction result.

[0101] Specifically, according to the standardized monitoring data, the actual value of each index is calculated, and then combined with the corresponding weight, the comprehensive state evaluation value is obtained by weighted summation; according to the comprehensive evaluation value interval, the current operation mode of the motor is judged: when the comprehensive evaluation value is low and the load data shows that the load rate is < 30% of the rated load, it is determined as light load mode; when the load rate is 30%-80% and the indexes are all in the normal range, it is determined as rated load mode; when the load rate is > 80% or part of the indexes are in the sub-health range, it is determined as heavy load mode; when the speed shows a downward trend and the current direction changes, it is determined as braking mode; at the same time, the health state is judged, all indexes are normal, 1-2 indexes are in sub-health state, and there are indexes reaching the threshold of failure precursor or more than 3 indexes in sub-health state, it is determined as failure precursor state.

[0102] Extract 50 groups of historical standardized monitoring data and parameter identification results as training samples of long short-term memory neural network, deploy the trained LSTM model to the control core module, input the latest 10 groups of monitoring data and parameter data, predict the load change trend, key parameter drift rule and comprehensive state evaluation value evolution direction in the future 5ms-10ms, and finally generate trend prediction results including prediction results, confidence and risk prompt.

[0103] Referring to Figure 5 As shown, a multi-modal control strategy library is constructed based on the comprehensive state evaluation value and the trend prediction result, and the control strategy is determined according to the mode and the comprehensive state evaluation value of the motor, and the motor parameters are dynamically adjusted by different algorithms, which specifically includes:

[0104] Based on the operation state evaluation and trend prediction result, a strategy library including optimized vector control algorithm, model predictive control algorithm, PID-sliding mode composite control algorithm and braking energy recovery control algorithm is built;

[0105] Based on the current operation mode and health state of the motor, the control strategy is matched, the MPC algorithm is started when the motor is in light load and normal state, the voltage vector is optimized to reduce energy consumption and current harmonic; the FOC algorithm is used when the motor is in rated load to ensure speed stability and control accuracy, the PID-sliding mode composite control algorithm is used when the motor is in heavy load or sub-health state, the PID is used to keep steady state and the sliding mode is used to resist disturbance, and the energy recovery control algorithm is used in braking mode to coordinate motor power generation and inverter feedback.

[0106] Specifically, the smooth transition switching process is performed, and the parameter gradient adjustment is started 10 sampling periods before the switching starts: on the one hand, the parameter weight of the original strategy is gradually reduced, for example, the voltage vector selection weight of the original MPC algorithm is reduced every period; on the other hand, the parameter weight of the target strategy is increased synchronously, for example, the current loop adjustment weight of the target vector control algorithm is increased by 10% every period, to realize the gradient superposition of the two types of strategies; at the same time, the motor torque change is monitored in real time, and if the torque fluctuation exceeds 5% of the rated torque, the weight adjustment rate is immediately slowed down until the fluctuation returns to the threshold; after 10 sampling periods, the original strategy weight is reduced to 0% and the target strategy weight is increased to 100%, the impact-free switching is completed, the motor speed and current are ensured to be free of sudden changes, and the operation stability is maintained.

[0107] Referring to Figure 6 Based on the trend prediction result and the selected control strategy, the adaptive inertia weight and the chaos mutation operator are introduced into the improved particle swarm optimization algorithm to optimize the core parameters of the control strategy, and the optimal control parameter combination is obtained, which specifically includes:

[0108] Based on the load and state prediction results and the selected control strategy, the improved particle swarm optimization algorithm is used to optimize the core parameters of the control strategy;

[0109] For the MPC algorithm, the prediction time domain length, the control time domain length and the weight coefficient are optimized, for the vector control algorithm, the d / q axis current loop PI regulator parameters are optimized, and for the PID-sliding mode compound control algorithm, the PID gain, the sliding mode surface coefficient and the exponential reaching law parameters are optimized;

[0110] The adaptive inertia weight and the chaos mutation operator are introduced into the improved PSO algorithm, a large inertia weight is used for global search in the early iteration to obtain a large range of optimal solution, and the inertia weight is reduced in the later iteration to focus on local search, the chaos mutation is used to break the local optimal trap and continuously approach the target value, and the optimal control parameter combination is obtained.

[0111] Specifically, for different control strategies, the core parameters are optimized one by one by the improved PSO algorithm: if the current is the MPC algorithm, the optimal prediction time domain, control time domain and weight coefficient are searched by taking energy consumption + current THD + response time as the fitness function, for example, when the load increases, the prediction time domain length is appropriately increased to cope with the load change in advance; if it is a vector control algorithm, the d / q axis current loop PI parameters are optimized by taking speed fluctuation + current THD as the fitness function, when the speed fluctuation exceeds the limit, the proportional coefficient is adjusted to suppress the fluctuation; if it is a PID-sliding mode compound control algorithm, the PID gain and the sliding mode surface coefficient are optimized by taking the disturbance rejection ability + response time as the fitness function, when the parameter drift is obvious, the sliding mode surface coefficient is increased to improve the disturbance rejection ability; during the optimization process, every 10 iterations, the current optimal parameter combination is extracted and compared with the optimization target, and if the target is met, the iteration can be terminated in advance.

[0112] Referring to Figure 7 As shown, the corresponding control instructions are generated based on the optimal control parameters and the selected control strategy, and the drive signals generated based on the control instructions are transmitted to the power drive module to output three-phase sinusoidal current or braking feedback current, which specifically includes:

[0113] For vector control, MPC, and PID-sliding mode compound control strategy, an adaptive SVPWM algorithm is adopted, combined with motor rotor position information and optimal control parameters, to calculate three-phase voltage values, divide voltage vector sectors, and generate high-frequency PWM drive signals;

[0114] For the braking energy recovery control strategy, an energy feedback control signal is directly generated to adjust the inverter to the energy feedback working mode;

[0115] The generated PWM drive signals or energy feedback control signals are transmitted to the power drive module, which operates at a preset switching frequency and duty cycle to output three-phase sinusoidal current or braking feedback current.

[0116] Specifically, if the current is vector control MPC or PID-sliding mode compound control, the adaptive space vector pulse width modulation algorithm is started: according to the real-time position data of the rotor, the basic sector where the current voltage vector is located is determined; then combined with the motor speed and load change rate, the sector division precision is dynamically adjusted, when the speed ≥3000rpm or the load change rate >10% / ms, the basic 6 sectors are subdivided into 12 sectors; when the speed <3000rpm and the load is stable, maintain 6 sector division to reduce the calculation amount; according to the voltage reference value in the optimal control parameters, the voltage vector action time of each sector is calculated, and the dead zone time is dynamically adjusted combined with the bus voltage to avoid IGBT upper and lower arm shoot-through; according to the optimized switching time, high-frequency PWM drive signals are generated;

[0117] If the current is the braking energy recovery control strategy, the control core module first judges the motor reverse drag state, and determines the safety threshold of energy feedback combined with the bus voltage data; according to the optimized energy recovery parameters, the inverter working mode switching instruction is generated to switch the inverter from the drive mode to the rectifier mode; at the same time, the limit value of the feedback current is calculated, the PWM modulation instruction is generated, and the IGBT module is controlled to act according to the specific duty cycle, so that the motor stator winding generates a reverse electromotive force, and the mechanical energy is converted into electrical energy.

[0118] Referring to Figure 8 As shown, real-time monitoring of motor parameter data is performed, and compared with the motor control target value to calculate speed error, current error and state evaluation deviation, based on which the identification result is corrected, the control strategy switching threshold is adjusted, and the optimal objective function is dynamically updated

[0119] Real-time acquisition of motor monitoring data, and compared with the preset control target value, respectively calculate the speed error, current error, combined with the state evaluation index system, calculate the comprehensive state evaluation deviation;

[0120] The calculated speed error, current error and state evaluation deviation are used as feedback signals for data feedback;

[0121] According to the feedback deviation, the identification result is corrected, the strategy switching threshold is adjusted, and the optimization objective function is dynamically updated, and the above deviation calculation, feedback transmission and adjustment process are repeated to realize continuous iterative optimization of control performance.

[0122] Specifically, after the parameter identification module receives the deviation signal, if the current deviation continuously exceeds 5% of the rated current, it is judged that the estimation of the inductance parameter is deviated, and the observation equation weight of the EKF algorithm is immediately corrected to enhance the influence of the current signal in parameter estimation and improve the inductance identification accuracy; If the state evaluation deviation received by the control strategy switching module is ≤10 points, the threshold of the original load change rate >5% / ms switching strategy is down-regulated to 3% / ms, the strategy switching is triggered in advance to avoid further deterioration of the state; According to the speed deviation, if the speed fluctuation exceeds the target range, the optimization weight of the original speed fluctuation minimization is increased from 30% to 45%, the optimal control parameter combination is recalculated, and the control of speed stability is strengthened.

[0123] It should be noted that the above-mentioned embodiments of the present application are only for description, not representing the advantages and disadvantages of the embodiments. And the above describes the specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or may be advantageous.

[0124] Each embodiment in the present application is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment focuses on the difference from other embodiments.

[0125] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of adaptive brushless motor control, characterized by, The application relates to a motor real-time state monitoring and control method based on a multi-mode control strategy. Real-time motor running state parameter data are collected through a sensor group, and original data collected are preprocessed to obtain standardized state monitoring data. Based on the standardized monitoring data, an extended Kalman filter algorithm is adopted to establish motor state equations and observation equations, output results are input into a radial basis function neural network for error correction, and real-time motor time-varying parameter identification results are obtained. Based on the standardized monitoring data and the parameter identification results, a motor running state evaluation index system is constructed, a comprehensive state evaluation value is calculated through an analytic hierarchy process, a long short-term memory neural network is adopted to predict load variation trends, parameter drift rules and state evaluation value evolution directions in a future time period, and trend prediction results are generated. Based on the comprehensive state evaluation value and the trend prediction results, a multi-mode control strategy library is constructed, a control strategy is determined according to a motor mode and the comprehensive state evaluation value, and motor parameters are dynamically adjusted through different algorithms. Based on the trend prediction results and the selected control strategy, an improved particle swarm optimization algorithm is adopted to introduce an adaptive inertia weight and a chaotic mutation operator to perform online optimization on core parameters of the control strategy, and optimal control parameter combinations are obtained. Based on the optimal control parameters and the selected control strategy, corresponding control instructions are generated, drive signals generated based on the control instructions are transmitted to a power drive module, and three-phase sinusoidal currents or braking feedback currents are output. Real-time motor parameter data are monitored and compared with motor control target values, speed error, current error and state evaluation deviation are calculated, the identification results are corrected, the control strategy switching threshold is adjusted, and an optimization objective function is dynamically updated.

2. The adaptive brushless motor control method of claim 1, wherein, The real-time motor running state parameter data are collected through the sensor group, and the original data collected are preprocessed to obtain the standardized state monitoring data, specifically including the following steps. Based on the calibrated sensor group, real-time rotor speed, three-phase stator instantaneous current, stator winding temperature, motor casing vibration amplitude and frequency, and DC bus voltage motor running key parameters are collected in real time. According to the interference characteristics of different types of data, corresponding preprocessing methods are adopted, adaptive Kalman filter is adopted for the current and speed signals susceptible to electromagnetic interference to eliminate power grid noise and electromagnetic interference, and sliding average filter is adopted for the temperature and vibration signals with random fluctuations to weaken random errors. The filtered original data are converted into standardized state monitoring data meeting the requirements of subsequent processing.

3. The adaptive brushless motor control method of claim 1, wherein, Based on the standardized monitoring data, the extended Kalman filter algorithm is adopted to establish the motor state equations and the observation equations, the output results are input into the radial basis function neural network for error correction, and the real-time motor time-varying parameter identification results are obtained, specifically including the following steps. Based on the standardized monitoring data, the extended Kalman filter algorithm is adopted to construct the motor state equations and the observation equations, the speed and three-phase current are taken as observation inputs, and the stator resistance, d / q axis inductance and back electromotive force coefficient are preliminarily estimated. The preliminary estimation results of the EKF and the winding temperature data are input into the radial basis function neural network, the EKF estimation error is corrected through online learning, and the stator resistance drift caused by the temperature and the inductance nonlinear change caused by the magnetic saturation are compensated. The control identification period is consistent with the sampling period, and finally the real-time parameter identification result after error correction and disturbance compensation is output.

4. The adaptive brushless motor control method of claim 1, wherein, The motor operation state evaluation index system is constructed based on the standardized monitoring data and the parameter identification result, the comprehensive state evaluation value is calculated by the analytic hierarchy process, the load change trend, the parameter drift rule and the state evaluation value evolution direction in the future time period are predicted by the long short-term memory neural network, and the trend prediction result specifically includes: A motor operation state evaluation index system is constructed based on standardized monitoring data and parameter identification results, and the index system includes a speed fluctuation coefficient, a current harmonic distortion rate, a winding temperature deviation value, a vibration amplitude exceeding times and a parameter drift rate. The weights of each index are determined by the analytic hierarchy process, the comprehensive state evaluation value is calculated in combination with the actual values of the indexes, and the current operation mode of the motor is distinguished according to the preset threshold, including light load, rated load, heavy load and braking, and the health state is determined, including normal, sub-health and pre-fault. The load change trend, the parameter drift rule and the state evaluation value evolution direction in the future time period are predicted by the long short-term memory neural network based on historical monitoring data and identification parameters. The comprehensive state evaluation result, the operation mode and the health state determination conclusion are integrated, and the trend prediction data is superimposed to form a complete operation state evaluation and trend prediction result.

5. The adaptive brushless motor control method of claim 1, wherein, The multi-modal control strategy library is constructed based on the comprehensive state evaluation value and the trend prediction result, and the control strategy is determined according to the mode and the comprehensive state evaluation value of the motor to dynamically adjust the motor parameters by different algorithms, specifically including: Based on the operation state evaluation and the trend prediction result, a strategy library containing an optimized vector control algorithm, a model predictive control algorithm, a PID-sliding mode composite control algorithm and a braking energy recovery control algorithm is built; Based on the current operation mode and the health state of the motor, the control strategy is matched, the MPC algorithm is started when the motor is in light load and normal state to optimize the voltage vector to reduce energy consumption and current harmonic; the FOC algorithm is used when the motor is in rated load to ensure speed stability and control accuracy; the PID-sliding mode composite control algorithm is used when the motor is in heavy load or sub-health state to stabilize the state by PID and resist disturbance by sliding mode; and the energy recovery control algorithm is used in braking mode to coordinate motor power generation and inverter feedback.

6. The adaptive brushless motor control method of claim 1, wherein, Based on the trend prediction result and the selected control strategy, the core parameters of the control strategy are optimized online by introducing adaptive inertia weight and chaos mutation operator into the improved particle swarm optimization algorithm to obtain the optimal control parameter combination, specifically including: Based on the load and state prediction result, the selected control strategy is optimized by the improved particle swarm optimization algorithm to optimize the core parameters of the control strategy; The prediction time domain length, the control time domain length and the weight coefficient of the MPC algorithm are optimized, the d / q axis current loop PI regulator parameters of the vector control algorithm are optimized, and the PID gain, sliding mode surface coefficient and exponential reaching law parameters of the PID-sliding mode composite control algorithm are optimized. The improved PSO algorithm introduces adaptive inertia weight and chaos mutation operator, uses large inertia weight to carry out global search in the initial iteration, obtains a wide range of optimal solution, reduces inertia weight in the later iteration, focuses on local search, breaks the local optimal trap through chaos mutation, continuously approaches the target value, and obtains the optimal control parameter combination.

7. The adaptive brushless motor control method of claim 1, wherein, The corresponding control instruction is generated based on the optimal control parameter and the selected control strategy, the drive signal generated based on the control instruction is transmitted to the power drive module, and the three-phase sinusoidal current or braking feedback current is outputted, specifically including: For the vector control, MPC, PID-sliding mode compound control strategy, the adaptive SVPWM algorithm is adopted, the three-phase voltage value is calculated combined with the motor rotor position information and the optimal control parameter, the voltage vector sector is divided, and the high-frequency PWM drive signal is generated; For the braking energy recovery control strategy, the energy feedback control signal is directly generated, and the inverter is adjusted to the energy feedback working mode; The generated PWM drive signal or energy feedback control signal is transmitted to the power drive module, and the three-phase sinusoidal current or braking feedback current is outputted according to the preset switching frequency and duty ratio.

8. The adaptive brushless motor control method of claim 1, wherein, The motor parameter data is monitored in real time, and compared with the motor control target value, the speed error, current error and state evaluation deviation are calculated, the identification result is corrected based on the deviation, the control strategy switching threshold is adjusted, and the optimization target function is dynamically updated, specifically including: The motor monitoring data is collected in real time, and compared with the preset control target value, the speed error and current error are calculated respectively, and the comprehensive state evaluation deviation is calculated combined with the state evaluation index system; The calculated speed error, current error and state evaluation deviation are used as feedback signals for data feedback; According to the feedback deviation, the identification result is corrected, the strategy switching threshold is adjusted, and the optimization target function is dynamically updated, and the above deviation calculation, feedback transmission and adjustment process are repeated to realize continuous iteration optimization of control performance.

9. An adaptive brushless motor control system for implementing an adaptive brushless motor control method according to any one of claims 1-8, characterized by It includes: Parameter identification module: based on standardized monitoring data, the extended Kalman filter algorithm is used to establish equations to preliminarily estimate parameters, and the radial basis function neural network error correction is used to output real-time motor time-varying parameter identification results; State evaluation and trend prediction module: an evaluation index system is constructed, the comprehensive state evaluation value is calculated by using the analytic hierarchy process, and the future state trend is predicted by using the long short-term memory neural network to form complete evaluation and prediction results; Control strategy decision module: based on the comprehensive state evaluation value and the trend prediction result, a multi-modal control strategy library is constructed, and the control strategy is matched and determined according to the motor operation mode and health state; Control parameter optimization module: combined with the trend prediction result and the selected control strategy, the improved particle swarm optimization algorithm introducing adaptive inertia weight and chaos mutation operator is adopted to optimize the core parameters to obtain the optimal combination; Drive signal generation module: according to the optimal control parameter and the selected control strategy, the PWM drive signal or the energy feedback control signal is generated and transmitted to the power drive module to output the three-phase sinusoidal current or the braking feedback current; Feedback and iteration optimization module: real-time monitoring of motor parameters and comparison with target values to calculate deviations, correction of identification results based on deviations, adjustment of strategy switching threshold, and optimization of target function; Processor: the processor is used for processing calculation processes of respective formulas and construction calculation processes of respective models.

Images