Permanent magnet synchronous motor parameter identification method based on adaptive forgetting factor

Abstract

The invention relates to the technical field of motor control and parameter identification, and discloses a permanent magnet synchronous motor parameter identification method based on a self-adaptive forgetting factor, which comprises the following steps of: acquiring an input signal and an output signal of a permanent magnet synchronous motor in real time, and constructing a regression vector by using the acquired input signal and output signal; calculating the information entropy of the current moment according to the constructed regression vector; performing normalization processing on the information entropy, and calculating a forgetting factor at the current moment according to the normalized information entropy; and substituting the forgetting factor into a parameter updating formula of a recursive least square method for parameter identification. According to the method disclosed by the invention, the adaptive capacity to dynamic change of motor parameters can be improved, the interference of noise on identification precision is effectively suppressed, the credibility of an identification result is improved, and the robustness and long-term stability of a control system are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control and parameter identification, and in particular to a permanent magnet synchronous motor parameter identification method based on an adaptive forgetting factor. Background Art

[0002] Permanent magnet synchronous motors have been widely used in industrial automation, new energy vehicles, aerospace and other fields due to their high efficiency, high power density, low noise and superior dynamic performance. In the control system of permanent magnet synchronous motors, accurate motor parameters are one of the key factors to achieve high-performance control. These parameters include stator resistance, inductance, flux linkage, etc. The accuracy of the parameters directly affects the operating characteristics of the motor and the effectiveness of the control algorithm. However, since the motor parameters will change dynamically due to external conditions such as temperature, load, and working conditions, static or fixed parameter setting methods are difficult to meet actual use requirements. Therefore, conducting real-time parameter identification research is of great significance to improving the control performance of the motor system.

[0003] In the prior art, the recursive least squares method (RLS) is a classic algorithm widely used in the identification of permanent magnet synchronous motor parameters. This method realizes real-time estimation of system parameters through recursive calculation, and has the characteristics of high computational efficiency and easy implementation. In order to enhance the ability of the RLS algorithm in processing dynamic systems, a forgetting factor is usually introduced, which can reduce the influence of historical data on the current estimation, thereby improving the sensitivity of the algorithm to parameter changes. However, in practical applications, the choice of the forgetting factor is usually a fixed value or set by experience. This method has obvious limitations when facing complex working conditions: if the forgetting factor is set too small, the algorithm will not respond quickly enough to parameter changes, resulting in an increase in identification error; and if the forgetting factor is set too large, the identification result will be more sensitive to noise, thereby reducing the stability of the system. In addition, the traditional RLS algorithm is difficult to achieve high-precision identification of motor parameters in the presence of measurement noise or interference signals.

[0004] The above problems mainly arise from the following reasons: First, the fixed setting method of the forgetting factor cannot dynamically adapt to the actual operating state of the system, and cannot strike a balance between real-time identification and robustness; second, the existing algorithms do not fully consider the uncertainty of information in the operating environment and the impact of noise on identification accuracy; finally, traditional algorithms usually lack the ability to fully respond to complex operating conditions and cannot meet the strict requirements of modern high-performance motor drive systems for parameter identification.

[0005] Therefore, in order to solve the problems in the existing technology, there is an urgent need for a new method that can adaptively adjust the forgetting factor, enhance the algorithm's ability to track dynamic changes in parameters, and improve noise resistance, thereby achieving higher precision and more robust motor parameter identification. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a permanent magnet synchronous motor parameter identification method based on an adaptive forgetting factor, so as to achieve the purpose of improving the adaptability to dynamic changes of motor parameters, effectively suppressing the interference of noise on identification accuracy, improving the credibility of identification results, and ensuring the robustness and long-term stability of the control system.

[0007] To achieve the above object, the technical solution of the present invention is as follows:

[0008] A method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor comprises the following steps:

[0009] Step 1, collecting input signals and output signals of the permanent magnet synchronous motor in real time, and constructing a regression vector with the collected input signals and output signals;

[0010] Step 2, calculating the information entropy at the current moment according to the constructed regression vector;

[0011] Step 3, normalize the information entropy, and calculate the forgetting factor at the current moment according to the normalized information entropy;

[0012] Step 4: Substitute the forgetting factor into the parameter update formula of the recursive least squares method to perform parameter identification, and the latest estimated motor parameters can be obtained.

[0013] In the above scheme, the input signal of the permanent magnet synchronous motor includes a stator voltage, and the stator voltage includes voltage components u of the d-axis and the q-axis. d and u q The output signal includes the d-axis and q-axis current i d and i q , and the motor speed ω m .

[0014] In the above scheme, the constructed regression vector φ k for:

[0015] φ k =[u d (k)u q (k)i d (k)i q (k)ω m (k)] T

[0016] Among them, k represents the current moment.

[0017] In the above scheme, the specific method of step 2 is as follows:

[0018] (1) Prepare sliding window data:

[0019] Assume that the sliding window length is N, and the sliding window data at the current moment is:

[0020] X k ={x k-N+1 ,x k-N+2 ,…,x k}

[0021] Among them, x k represents the q-axis data sampled at time k;

[0022] Each time a new data point x is collected k After that, the sliding window data slides forward and adds x k As the latest data, and remove the oldest data x k-N ;

[0023] (2) Estimate the probability distribution of sliding window data:

[0024] First, the continuous data in the sliding window is discretized:

[0025] Divide the data values ​​into several intervals [b1, b2, …, b i ,…,b m ], b i is the boundary of the i-th interval, and counts the number of data points n in each interval i ;n i For falling on [b i-1 ,b i )The number of data points in the interval;

[0026] Secondly, the kernel density estimation method is used to smooth the data probability distribution:

[0027]

[0028] Where h is the bandwidth parameter; K(·) is the kernel function, x j is the voltage, current, speed and other parameters collected at time j, p(x i ) The data probability distribution corresponding to the data x;

[0029] (3) Calculate the current information entropy value:

[0030] Through the probability distribution p(x i ) Calculate the information entropy value of the sliding window data:

[0031]

[0032] Here, m represents the number of intervals.

[0033] In the above scheme, in step 3, the formula for normalizing information entropy is as follows:

[0034]

[0035] Among them, H max is the theoretical maximum value of the information entropy of the system operation, H k is the information entropy value at the current moment, is the normalized information entropy value at the current moment.

[0036] In the above scheme, in step 3, the formula for calculating the forgetting factor at the current moment according to the normalized information entropy is as follows:

[0037]

[0038] Among them, λ min and λ max are the upper and lower limits of the forgetting factor, respectively. is the normalized information entropy value at the current moment.

[0039] In the above scheme, in step 4, the parameter update formula of the recursive least squares method is as follows:

[0040]

[0041] Among them, P k is the gain matrix at time k, φ k is the regression vector at time k, λ k is the forgetting factor at time k, θ k is the parameter of the identified k-th moment, To optimize the residual, the calculation formula is as follows:

[0042]

[0043] Among them, w k is the weight, α is the adjustment factor used to balance the sensitivity to the residual amplitude, e k To identify the error.

[0044] Through the above technical solution, the method for parameter identification of a permanent magnet synchronous motor based on an adaptive forgetting factor provided by the present invention has the following beneficial effects:

[0045] 1. Aiming at the problem of insufficient dynamic parameter identification accuracy, the present invention introduces an adaptive forgetting factor adjustment mechanism based on information entropy, dynamically adjusts the forgetting factor size according to the amount of information of motor parameter changes, so that the algorithm can maintain a high identification accuracy under both slowly changing and rapidly changing working conditions. This improvement significantly improves the adaptability to dynamic changes in motor parameters and can meet the requirements of motor control systems for parameter identification real-time and accuracy under different working conditions.

[0046] 2. In view of the problem of insufficient anti-noise performance of the prior art, the present invention adopts information entropy as a tool to measure the uncertainty of system information, and uses it to guide the dynamic adjustment of the forgetting factor. When the information entropy is high (when the noise or disturbance is large), the system will increase the adjustment of the forgetting factor to enhance the suppression of noise; when the information entropy is low, the system will reduce the adjustment amplitude to improve the sensitivity of parameter identification. Through this adaptive adjustment strategy, the present invention can effectively enhance the robustness of parameter identification in a noisy environment and improve the credibility of the identification results.

[0047] 3. The present invention aims to solve the problem of the difficulty in balancing the real-time performance and stability in the prior art by using an adaptive forgetting factor mechanism based on information entropy to balance the contradiction between real-time performance and stability. During the operation of the system, information entropy provides dynamic quantification of the parameter change rate and the amount of information. By adaptively adjusting the forgetting factor, it is possible to quickly respond to parameter changes while avoiding excessive response to noise or algorithm oscillation, thereby ensuring the stability and reliability of the system operation. This method not only improves the efficiency and accuracy of parameter identification, but also effectively ensures the robustness and long-term stability of the control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0049] Figure 1 The present invention discloses a method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0051] The present invention provides a method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor, such as Figure 1 As shown, the following steps are included:

[0052] Step 1, collecting input signals and output signals of the permanent magnet synchronous motor in real time, and constructing a regression vector with the collected input signals and output signals;

[0053] Step 2, calculating the information entropy at the current moment according to the constructed regression vector;

[0054] Step 3, normalize the information entropy, and calculate the forgetting factor at the current moment according to the normalized information entropy;

[0055] Step 4: Substitute the forgetting factor into the parameter update formula of the recursive least squares method to perform parameter identification.

[0056] To more clearly illustrate the method of the present invention, MATLAB is used for simulation, which includes the following steps:

[0057] S1: Initialization parameters

[0058] (1) System parameter initialization

[0059] Set the initial parameter vector θ0 of the recursive least squares method. The initial value can be set according to the rated parameters or empirical values ​​of the motor;

[0060] Set the initial covariance matrix P0, which is usually a diagonal matrix with a large initial value to reduce the impact of the initial error on subsequent estimates;

[0061] Define the range of the forgetting factor [λ min ,λ max ], take λ min =0.65,λ max =1.00.

[0062] (2) Initialization of variables related to sampling window and information entropy value calculation

[0063] Set the sliding window length N, which is usually an integer multiple of the sampling frequency, to balance real-time performance and stability;

[0064] Initialize the information entropy value calculation related variables, including the sliding window data X = [x1, x2, ..., x N ], used to record the q-axis current and the probability density estimation function p(x) of the data distribution.

[0065] S2: Real-time acquisition of input and output signals of permanent magnet synchronous motor:

[0066] (1) Motor input signal

[0067] For permanent magnet synchronous motors, the usual input signals include the stator voltage u d and u q (voltage components of the d-axis and q-axis). These signals are usually provided by the motor driver. The control system calculates the corresponding voltage input in real time based on the required torque and speed and applies it to the motor stator.

[0068] In the mathematical model of the motor, the input signal includes not only the voltage, but also the electrical characteristics of the motor, such as inductance, resistance, permanent magnet flux, etc. The voltage equations of the motor's d-axis and q-axis are:

[0069]

[0070] Among them, R s is the stator resistance; L d ,Lq are the inductances of the d-axis and q-axis respectively; i d ,i q are the currents of the d-axis and q-axis respectively; ω is the angular velocity of the motor; λ m is the motor flux.

[0071] (2) Motor output signal

[0072] The motor output signal is mainly the current i of the d-axis and q-axis d and i q And the motor speed ω m (or angular position). These signals are usually measured in real time by current sensors and speed sensors and fed back to the motor control system.

[0073] Current signal i d and i q It is a direct reflection of motor drive control and is also a key output signal for motor parameter identification.

[0074] Speed ​​signalω m Describes the rotation speed of the motor, which is an important indicator of the motor's operating status and affects the output of the motor's torque.

[0075] The output signal is not only directly affected by the input signal, but also by the motor rotor dynamics and electrical parameters. e The relationship with the current signal can be expressed by the following formula:

[0076]

[0077] Where p is the number of motor pole pairs; λ m is the motor flux; L d ,L q is the inductance of the d-axis and q-axis; i d ,i q are the currents of the d-axis and q-axis respectively.

[0078] There may be noise or interference signals in the collected raw data, especially when there are high-frequency dynamic changes or when the motor is subject to external interference during operation. Therefore, in actual implementation, the collected input and output signals need to use a low-pass filter method to remove high-frequency noise to ensure the smoothness of the signal.

[0079] In order to use these acquired data in the recursive least squares (RLS) algorithm, the acquired signal needs to be converted into a regression vector φ k The regression vector consists of a combination of input and output signals, and usually includes the following parts:

[0080] The input signal at the current moment: u d(k),u q (k);

[0081] Output signal at the current moment: i d (k),i q (k);

[0082] Motor speed or other related status signals: ω m (k).

[0083] The constructed regression vector φ k The form is:

[0084] φ k =[u d (k) u q (k) i d (k) i q (k) ω m (k)] T

[0085] S3: Calculating Information Entropy

[0086] (1) Definition of information entropy

[0087] Information entropy is a concept in Shannon's information theory, which is used to measure the uncertainty of random variables. For a discrete random variable X, its information entropy H(X) is defined as:

[0088]

[0089] Among them, x i represents the possible value of a random variable; p(x i ) represents x i The probability of occurrence; n is the total number of possible values ​​of the random variable.

[0090] The larger the information entropy, the higher the uncertainty of the random variable; the smaller the information entropy, the more concentrated the distribution of the random variable.

[0091] During motor operation, changes in input and output data (such as voltage, current, and speed) reflect the dynamic characteristics of the system. By calculating the entropy value of these data, the magnitude of changes in the current operating conditions can be evaluated:

[0092] High entropy value: indicates that the system data changes dramatically, indicating that parameter identification requires a quick response;

[0093] Low entropy value: indicates that the system data is relatively stable, indicating that parameter identification can enhance the retention of historical data.

[0094] (2) Detailed steps for calculating information entropy

[0095] a. Prepare sliding window data

[0096] Sliding window is a commonly used data processing method, which is used to dynamically capture the data characteristics within a certain period of time. Assume that the window length is N, and the sliding window data at the current moment is:

[0097] X k ={x k-N+1 ,x k-N+2 ,…,x k}

[0098] Among them, x k Represents the q-axis data sampled at time k.

[0099] Each time a new data point x is collected k After that, the window data slides forward: add x k As the latest data, and remove the oldest data x k-N .

[0100] b. Estimate the probability distribution of sliding window data

[0101] The continuous data in the sliding window needs to be discretized. Divide the data values ​​into several intervals [b1, b2, …, b m ], b i is the boundary of the i-th interval, and counts the number of data points n in each interval i , n i For falling on [b i-1 ,b i ) is the number of data points in the interval.

[0102] In order to improve the accuracy of probability distribution, the kernel density estimation (KDE) method is used to smooth the data probability distribution:

[0103]

[0104] Where h is the bandwidth parameter; K(·) is the kernel function, x j is the voltage, current, speed and other parameters collected at time j, p(x i ) The data probability distribution corresponding to data x

[0105] c. Calculate the current information entropy value

[0106] Through the probability distribution p(x i ) Calculate the information entropy value of the sliding window data:

[0107]

[0108] Here, m represents the number of intervals.

[0109] The entropy value reflects the uniformity of the sliding window data distribution:

[0110] When p(x i ) is more evenly distributed (data varies greatly and there is no significant pattern), the entropy value is higher;

[0111] When p(x i ) are concentrated in certain specific intervals (the data are relatively stable and have obvious patterns), the entropy value is low.

[0112] S4: Adaptive adjustment of the forgetting factor

[0113] In the permanent magnet synchronous motor parameter identification based on the information entropy adaptive forgetting factor recursive least squares method, the forgetting factor λ determines the degree of memory of the algorithm for historical data. By adaptively adjusting λ, attention to new data can be enhanced when the motor changes drastically, and more historical information can be retained when the system is stable, ensuring a balance between identification accuracy and real-time performance.

[0114] (1) The role of the forgetting factor

[0115] The forgetting factor λ is a key parameter in the recursive least squares method (RLS), and its value range is 0<λ≤1:

[0116] λ→1: Pay more attention to historical data, which is suitable for situations where the system status is relatively stable.

[0117] λ→0: Pay more attention to new sampled data, which is suitable for situations where the system state changes dramatically.

[0118] By adjusting λ, the algorithm's reliance on new and old data can be changed dynamically.

[0119] (2) Principle of adaptive adjustment of forgetting factor

[0120] a. Dependence on information entropy H k

[0121] Information entropy H k Measures the uncertainty and change trend of input and output data at the current moment:

[0122] High entropy value: The data distribution changes greatly, and λ needs to be reduced to improve the responsiveness to new data.

[0123] Low entropy value: The data distribution is relatively stable, and λ needs to be increased to enhance the utilization of historical information.

[0124] b. Mapping relationship

[0125] Forgetting factor λ and information entropy H k The mapping relationship is often expressed as:

[0126]

[0127] Among them, λ min and λ max are the upper and lower limits of the forgetting factor (set to 0.65 and 1.0); H max It is the theoretical maximum value of information entropy (determined by the range and distribution of input and output data).

[0128] (3) Specific steps for adaptively adjusting the forgetting factor

[0129] a. Get the current information entropy value H k

[0130] Obtain the information entropy H of the current sliding window data from the calculation in step 3 k , reflecting the change range of the current system state.

[0131] b. Normalized information entropy

[0132] In order to uniformly handle entropy values ​​of different ranges, H k Normalization:

[0133]

[0134] Among them, H max It is the theoretical maximum value of the information entropy of the system operation.

[0135] c. Calculate the forgetting factor λ at the current moment based on the normalized information entropy k :

[0136]

[0137] when (Data changes dramatically): λ k →λ min , enhancing sensitivity to new data.

[0138] when (Data is stationary): λ k →λ max , retain more historical data.

[0139] d. Update the recursive formula

[0140] The adaptive forgetting factor λ k Substitute into the parameter update formula of the recursive least squares method to obtain the parameter θ at time k k :

[0141]

[0142] Among them, P k is the gain matrix at time k, φ kis the regression vector at time k, λ k is the forgetting factor at time k, θ k is the parameter of the identified k-th moment, To optimize the residual, the calculation formula is as follows:

[0143]

[0144] Among them, w k is the weight, α is the adjustment factor used to balance the sensitivity to the residual amplitude, e k To identify the error,

[0145] S5: Utilize the statistical characteristics of the residual to filter the noise through dynamic weighting. Replace the original residual e k Update the identification parameters to solve e k Problems such as poor noise suppression and poor robustness.

[0146] When the residual amplitude is large (possibly due to noise or abnormal sampling), the weight w k It will decrease, and the impact on abnormal points will be weakened.

[0147] S6: Output parameter identification results

[0148] Repeat S2 to S5 to update the motor parameters in real time, including stator resistance, inductance, flux linkage, etc. The identification results are passed to the motor controller for online adjustment of the control algorithm (such as vector control, direct torque control, etc.).

[0149] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor, characterized in that: The steps include: Step 1, collecting input signals and output signals of the permanent magnet synchronous motor in real time, and constructing a regression vector with the collected input signals and output signals; Step 2, calculating the information entropy at the current moment according to the constructed regression vector; Step 3, normalize the information entropy, and calculate the forgetting factor at the current moment according to the normalized information entropy; Step 4: Substitute the forgetting factor into the parameter update formula of the recursive least squares method to perform parameter identification, and the latest estimated motor parameters can be obtained.

2. A method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor according to claim 1, characterized in that: The input signal of the permanent magnet synchronous motor includes a stator voltage, which includes voltage components u of the d-axis and the q-axis. d and u q The output signal includes the d-axis and q-axis current i d and i q , and the motor speed ω m .

3. A permanent magnet synchronous motor parameter identification method based on adaptive forgetting factor according to claim 2, characterized in that: The constructed regression vector φ k for: φ k =[u d (k)u q (k)i d (k)i q (k)ω m (k)] T Among them, k represents the current moment.

4. A method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor according to claim 1, characterized in that: The specific method of step 2 is as follows: (1) Prepare sliding window data: Assume that the sliding window length is N, and the sliding window data at the current moment is: X k ={x k-N+1 ,x k-N+2 ,…,x k } Among them, x k represents the q-axis data sampled at time k; Each time a new data point x is collected k After that, the sliding window data slides forward and adds x k As the latest data, and remove the oldest data x k-N ; (2) Estimate the probability distribution of sliding window data: First, the continuous data in the sliding window is discretized: Divide the data values ​​into several intervals [b1, b2, …, b i ,…,b m ], b i is the boundary of the i-th interval, and counts the number of data points n in each interval i ;n i For falling on [b i-1 ,b i )The number of data points in the interval; Secondly, the kernel density estimation method is used to smooth the data probability distribution: Where h is the bandwidth parameter; K(·) is the kernel function, x j is the voltage, current, speed and other parameters collected at time j, p(x i ) The data probability distribution corresponding to the data x; (3) Calculate the current information entropy value: Through the probability distribution p(x i ) Calculate the information entropy value of the sliding window data: Here, m represents the number of intervals.

5. The method for parameter identification of a permanent magnet synchronous motor based on an adaptive forgetting factor according to claim 1, characterized in that: In step 3, the formula for normalizing information entropy is as follows: Among them, H max is the theoretical maximum value of the information entropy of the system operation, H k is the information entropy value at the current moment, is the normalized information entropy value at the current moment.

6. A method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor according to claim 1, characterized in that: In step 3, the formula for calculating the forgetting factor at the current moment based on the normalized information entropy is as follows: Among them, λ min and λ max are the upper and lower limits of the forgetting factor, respectively. is the normalized information entropy value at the current moment.

7. The method for identifying parameters of a permanent magnet synchronous motor based on an adaptive forgetting factor according to claim 1, characterized in that: In step 4, the parameter update formula of the recursive least squares method is as follows: Among them, P k is the gain matrix at time k, φ k is the regression vector at time k, λ k is the forgetting factor at time k, θ k is the parameter of the identified k-th moment, To optimize the residual, the calculation formula is as follows: Among them, w k is the weight, α is the adjustment factor used to balance the sensitivity to the residual amplitude, e k To identify the error.

Images