Motor driving system bus capacitance monitoring method

Abstract

The invention relates to a motor driving system bus capacitance monitoring method in the technical field of crane driving monitoring, and the method comprises the steps: firstly, introducing noise self-adaption in Kalman filtering, and effectively solving the problem that the noise of a monitoring model cannot be measured; and secondly, in order to solve the problem of incomplete monitoring of the aging parameters of the aluminum electrolytic capacitor, pseudo measurement is introduced into the monitoring model, and all-parameter monitoring of the aging parameters of the capacitor is realized. Then, a digital band-pass filter is designed, the filtered capacitor voltage and the reconstructed current are used as input and output of a model, and the monitoring precision is improved; and finally, through comparison with a recursive least square method and Kalman filtering, the superiority of the proposed method is shown. Simulation and experimental verification show that the method meets the precision requirement of capacitor online monitoring.

Description

Technical Field

[0001] The invention relates to the technical field of crane drive monitoring, and in particular to a method for monitoring bus capacitance of a motor drive system. Background Art

[0002] With the rapid development of industrial automation and drive technology, cranes play an increasingly critical role in a wide range of industrial fields. Among them, the motor drive system realizes the precise control of the crane.

[0003] In the actual operation of the crane, the DC bus capacitor, as an indispensable energy storage component of the crane motor drive system, undertakes the key tasks of power balance and filtering. However, as the use time increases, the capacitor may be affected by factors such as mechanical vibration, thermal stress and electrical stress, and gradually age and wear out. This aging may cause changes in the electrical characteristics of the capacitor, such as a decrease in equivalent series capacitance (ESC) and an increase in equivalent series resistance (ESR), which may lead to consequences such as reduced power quality, increased energy loss, and system failure in actual applications.

[0004] According to statistics, DC bus capacitors are one of the most vulnerable components in converter systems, and power electronic converter failures caused by DC bus capacitor failure account for 30% of total failures. Therefore, in-depth research on DC capacitor status monitoring technology in motor drive systems and improving the accuracy of capacitor aging monitoring are of great significance for optimizing motor drive system design and improving crane reliability.

[0005] At present, the mainstream online monitoring methods can be divided into two solutions: circuit model-based and fuzzy model-based: In the scheme based on circuit model, it can be further divided into steady-state ripple-based and disturbance signal-based methods. The steady-state ripple method covers the ripple caused by the converter switching frequency and the line frequency ripple caused by the input power on the DC bus. The disturbance signal is divided into artificially injected disturbance signal and converter transient process disturbance signal. The steady-state ripple-based method usually requires high-precision sensors and occupies more computing resources. It is often difficult for online state monitoring technology based on disturbance signals to select injection signals of appropriate frequency and amplitude, because the injected signal may violate the prescribed technical standards and pollute the power quality of the power grid.

[0006] In the capacitance monitoring method based on fuzzy model, the estimation theory is used to reduce the requirements for sensor capabilities, but the traditional estimation method based on recursive least squares (RLS) only contains measurement equations and does not consider the model error. The ability to estimate parameters is limited, and problems such as non-convergence of aging parameters and incorrect convergence of aging parameters may occur.

[0007] To this end, in order to address the problems of the existing capacitor online state monitoring method based on system steady-state operation, such as the great influence of model noise on the monitoring results, unobservable aging parameters due to measurement loss, and incomplete parameter monitoring, we proposed an online state monitoring method for DC bus capacitor based on adaptive Kalman filter (AKF), which fully considered the dynamic adaptation of model noise and effectively solved the problem of unmeasurable monitoring model noise. Summary of the invention

[0008] In order to overcome the deficiencies in the background technology, the present invention discloses a method for monitoring bus capacitance of a motor drive system.

[0009] In order to achieve the above-mentioned invention object, the present invention adopts the following technical scheme: A method for monitoring bus capacitance of a motor drive system comprises the following steps: Step 1: According to the state of the inverter side switch tube and the collected voltage and current signals, the current of the DC bus capacitor is Reconstructing it gives: ; In the formula, is the rectifier output current, which is collected by the sensor. The input current of the inverter is: ; In the formula, Represent the output current of the three phases ABC respectively, and Indicates the switching state of the inverter upper bridge arm IGBT, where "1" indicates the on state and "0" indicates the off state; Step 2: Construct a discrete model based on the capacitor equivalent circuit and convert it into a differential equation with unknown parameters as state variables; Step 3: Design a Butterworth bandpass filter to filter the DC bus voltage and reconstructed current to extract the 300Hz ripple on the DC side; Step 4: Propose an adaptive Kalman filter capacitance monitoring algorithm to dynamically adapt to the discrete model; Step 5: Convert the differential equation in step 2 into a measurement equation and construct a state equation with unknown parameters as state variables; Step 6: Propose an improved monitoring model based on the pseudo-measurement monitoring model to improve the convergence and accuracy of monitoring; Step 7: Substitute the improved monitoring model into the adaptive Kalman filter for solution.

[0010] Preferably, in step 2, according to the equivalent circuit of the capacitor, assuming that the input is , the output is , then the transfer function of the discrete model is: ; Where ESC is the equivalent series capacitance, ESR is the equivalent series resistance, and s is the complex frequency domain transformation operator; Discretize the transfer function and use the input and output to estimate the aging parameters of the capacitor.

[0011] Preferably, a bilinear transformation is used:

[0012] Where, T represents the artificially set sampling time, and z is the impedance transformation operator; Rewrite the transfer function of the capacitor discrete model: ; ; Convert the transfer function to a difference equation between the output and the input: ; In the formula, , is the voltage and current at the previous moment, , is the voltage and current at the next moment; use Calculate the aging parameters of the capacitor: .

[0013] Preferably, in step three, the Butterworth bandpass filter is get; in, and are the Butterworth bandpass filter coefficients.

[0014] Preferably, a Butterworth bandpass filter is used to filter the DC bus voltage Process and obtain Ripple at line frequency of 300Hz.

[0015] Preferably, in the Kalman filter, a hyperparameter is introduced As an adjustment update parameter, the forgetting factor b is introduced to adjust the importance of historical data; Among them, the adjustment update parameter The relationship between and the forgetting factor b is: ; The process noise Q and measurement noise R in the Kalman filter can be rewritten as: ; ; In the formula, is the process noise of the kth iteration, is the measurement noise at the kth iteration.

[0016] Preferably, in step five, As a state variable, and rewrite the difference equation between output and input: ; in,

[0017] It is the observation matrix constructed based on the signals collected by the sensor. It is a new state vector constructed by treating the unknown parameters of the system as unknown states. is the measurement error vector constructed to describe the error of the system sensor; Construct the state-space equations for the state variables: .

[0018] Preferably, the voltage measurement value at the last moment is placed in the reconstructed measurement matrix and the state variable is used accordingly. The measurement equation is rewritten as:

[0019] in, .

[0020] Due to the adoption of the technical solution as described above, the present invention has the following beneficial effects: after the pseudo-measurement is introduced into the DC bus capacitor aging parameter identification technology based on adaptive Kalman filtering, the dimension of the observation matrix is ​​increased from one dimension to two dimensions. This improvement solves the problem that the Kalman filter algorithm cannot converge due to the lack of sufficient measurement in the field of parameter identification, and significantly improves the convergence of the identification algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a topological diagram of the motor driver of the present invention; Figure 2 Schematic diagram of adaptive Kalman filter; Figure 3 The simulation results at different speeds are shown in Figure 2. Figure 4 The simulation results of different capacitor specifications are shown in Figure 2. Figure 5 It is a comparison between online condition monitoring based on AKF, KF and RLS; Figure 6 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0022] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0023] A method for monitoring bus capacitance of a motor drive system comprises the following steps: Step 1: According to the state of the inverter side switch tube and the collected voltage and current signals, the current of the DC bus capacitor is Refactoring Since the DC link is usually not equipped with a current sensor, the capacitor current can be obtained by reconstructing the signal: ; In the formula, is the output current of the rectifier, which is collected by the sensor; is the input current of the inverter, which can be expressed as: ; In this process, Represent the output current of the three phases ABC respectively, and Indicates the switching state of the inverter upper bridge arm IGBT, where "1" indicates the on state and "0" indicates the off state.

[0024] Step 2: Construct a discrete model based on the capacitor equivalent circuit and convert it into a differential equation with unknown parameters as state variables; Specifically, according to the equivalent circuit of the capacitor, assuming the input is , the output is , then the transfer function of the capacitor model is: ; Where ESC is the equivalent series capacitance, ESR is the equivalent series resistance, and s is the complex frequency domain transformation operator; Discretizing the transfer function, the input and output can be used to estimate the aging parameters of the capacitor. Here, bilinear transformation is used: ; Among them, T represents the artificially set sampling time, and z is the impedance transformation operator. Therefore, the transfer function of the capacitor model can be rewritten as: ; ; Convert this to a difference equation between the output and the input: ; In the formula, , is the voltage and current at the previous moment, , is the voltage and current at the next moment; When the sampling time T is determined and the voltage and current of the previous sampling and the voltage and current of this sampling are known, we can use To calculate the aging parameters of the capacitor: .

[0025] Step 3: Design a Butterworth bandpass filter to filter the DC bus voltage and reconstructed current to extract the 300Hz ripple on the DC side; Specifically, when the input and output of the capacitor are AC, the impedance equivalent model best fits the physical characteristics of the capacitor. In order to obtain the input and output sampling of the capacitor under the ripple signal, a Butterworth bandpass filter is designed to filter the DC bus voltage. Process and obtain Ripple at line frequency (300Hz), filter passes Design acquisition; in and are the filter coefficients, which can be calculated by the MATLAB filter design toolbox.

[0026] Step 4: Propose an adaptive Kalman filter capacitance monitoring algorithm to dynamically adapt to the discrete model; Specifically, in Kalman filtering, setting process noise Q and measurement noise R is a challenge, and these two values ​​are usually difficult to determine, which is related to the tracking ability of the algorithm. Specifically, process noise is related to the accuracy of the model, while measurement noise is related to the accuracy of the sensor used. Usually, designers need to determine the values ​​of Q and R offline and assign them to the program. In order to improve this situation, the noise adaptive Kalman filtering method is adopted. By introducing the forgetting factor, the initial values ​​of the measurement noise and process noise are given, and the two values ​​are dynamically adjusted to ensure that they are continuously updated during the iteration process, adapted to the convergence of the algorithm, and finally approach the real measurement noise and process noise.

[0027] In order to achieve noise adaptation, a hyperparameter is introduced , as the adjustment update parameter, and introduce the forgetting factor b to adjust the importance of historical data. Adjust the update parameter The relationship between and the forgetting factor b is: ; Rewrite Q and R in the Kalman filter as: ;

[0028] In the formula, is the process noise of the kth iteration, is the measurement noise at the kth iteration.

[0029] Step 5: Convert the differential equation in step 2 into a measurement equation and construct a state equation with unknown parameters as state variables; Specifically, in order to estimate ESC and ESR, is considered as a state variable. Therefore, the difference equation between output and input can be rewritten as: ; in, ; It is the observation matrix constructed based on the signals collected by the sensor. It is a new state vector constructed by treating the unknown parameters of the system as unknown states. is a measurement error vector constructed to describe the error of the system sensor.

[0030] Since the aging speed of the capacitor parameters is much slower than the iteration speed of the adaptive Kalman filter, it can be approximately considered that the state value at the previous moment is equal to the state value at the next moment, thereby constructing the state space equation of the state variable: ; The new state quantity is obtained by estimating the previous state quantity and process noise. The process noise of each state equation will change with the increase of the number of iterations and is used to describe the error of the previous and next state estimates.

[0031] Step 6: Propose an improved monitoring model based on the pseudo-measurement monitoring model to improve the convergence and accuracy of monitoring; Specifically, in Kalman filtering, the traditional observation model assumes that the measurement of state quantities is directly observable, but in practical applications, sometimes we encounter situations where some important state quantities cannot be directly measured. By introducing pseudo-measurements, we can indirectly obtain the estimated values ​​of these state quantities that are not directly measured, thereby improving the recognition ability of the filter. Pseudo-measurements can be calculated based on a combination of known related state quantities and sensor measurements, and then passed as input to the Kalman filter to optimize the performance of the filter. Therefore, the use of pseudo-measurement technology is an effective method to improve the recognition accuracy of the Kalman filter.

[0032] State Vector is a third-order column vector, and the measurement equation is first-order. When the order of the state matrix is ​​greater than the order of the measurement equation, some state quantities may be unobservable. Therefore, pseudo-measurements need to be introduced to improve the convergence of the algorithm.

[0033] The voltage measurement value at the previous moment is placed in the reconstructed measurement matrix and the state variable is used accordingly. However, The state unknown quantity that is not composed of ESR and ESC is a pseudo-measurement state quantity whose value is always 1. Therefore, the measurement equation can be rewritten as: ; in, .

[0034] After pseudo-measurement is introduced into the busbar capacitor aging parameter identification technology based on adaptive Kalman filtering, the dimension of the observation matrix is ​​increased from one dimension to two dimensions.

[0035] This improvement solves the problem of Kalman filtering algorithm failure to converge due to lack of sufficient measurements in the field of parameter identification, and significantly improves the convergence of the identification algorithm.

[0036] Step 7: Substitute the improved monitoring model into the adaptive Kalman filter for solution.

[0037] like Figure 1 As shown in the figure, a DC bus capacitor aging monitoring method based on adaptive Kalman filter motor drive system is applicable to the following topologies: Typical topologies of motor drive systems are as follows: Figure 1 During the normal operation of the motor drive system, the sensor collects and provides the DC bus voltage. , Rectifier side output current 、Motor angular velocity and angle , and the three-phase output current on the inverter side , these data are used in the system control loop. At the same time, the control loop can obtain the switching status of the inverter.

[0038] Figure 2 The schematic diagram of adaptive Kalman filter is shown. Kalman filter is a commonly used estimation technique that combines observation data with system model to estimate and predict unknown state variables. It can reduce the estimation error by weighted averaging and correcting the observation data, and can adaptively adjust the weights to adapt to different noise characteristics and system models. Kalman filter is widely used in control, signal processing, navigation, robotics and other fields. One of the most important applications is to achieve high-precision target tracking and attack in missile guidance. Its core idea is based on Bayesian filtering theory, and the optimal state estimation result is obtained by probabilistic inference of the uncertainty of the current state and prior information. In practical applications, Kalman filter needs to configure parameters and select models according to specific problems to achieve the best estimation effect. Kalman filter assumes that the system has linear properties, the noise follows Gaussian distribution, and the parameters of the dynamic model and observation model of the system are known. Although it performs well in many cases, it is limited when facing nonlinear systems, unknown noise and model uncertainty.

[0039] The adaptive Kalman filter is a filter that can automatically adjust parameters. It can adaptively update the filter parameters according to real-time observation data and system response during the filtering process. The adaptive Kalman filter can estimate and adjust the parameters of the system model and measurement model in real time by using recursive least squares or other adaptive mechanisms to better adapt to complex and changeable systems, unknown noise and model uncertainty. Compared with traditional Kalman filters, adaptive Kalman filters are more flexible and adaptable, and can provide more accurate state estimation and prediction in dynamic environments.

[0040] Figure 3 The simulation results of the proposed method at different motor speeds are shown. It can be observed that at different speeds, the monitoring program can stably converge to the actual value of the capacitor aging parameter. There are slight differences in the monitoring errors at different speeds, where the monitoring error at 500rpm is slightly higher than that at other speeds, but this does not affect the final monitoring results. In general, the monitoring error of ESR is slightly higher than that of ESC. The monitoring error of ESR at each speed is less than 2%, while the monitoring error of ESC is less than 0.5%.

[0041] Figure 4 The online monitoring results of the DC bus capacitance and equivalent resistance of three common motor drive systems are shown. According to the experimental results, the ESR monitoring error of different capacitors is less than 1.5%, while the monitoring error of ESC is less than 0.6%.

[0042] Figure 5 The online monitoring results based on AKF, KF and RLS are shown. The online monitoring program is started when k=50000, and each program converges from the initial value to the estimated value.

[0043] according to Figure 5 It can be observed in (a) that the proposed method is smoother, converges faster, and has smaller spikes than the other two methods, showing stronger robustness. In actual debugging, it is found that the results of RLS and KF are easily affected by the initial parameters. When the initial state value deviates greatly from the true value, RLS cannot converge effectively or converges incorrectly. KF will not converge incorrectly, but the monitoring results will have large burrs, leading to incorrect judgment of the degree of capacitor aging.

[0044] Figure 5(b) in the figure shows the change of errors of the three methods after the system starts monitoring. It can be seen that when the state monitoring is started at k=50000, KF converges faster and has smaller errors than RLS. However, after k=52000, KF and RLS take longer to stabilize than AKF. In general, the proposed online monitoring method based on AKF converges faster than KF and RLS, and has smaller monitoring errors. In summary, the online state monitoring method based on AKF has better convergence.

[0045] The parts of the present invention that are not described in detail are prior art. It is obvious to those skilled in the art that the present invention is not limited to the details of the above-mentioned exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention; therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive, and it is intended that all changes that fall within the meaning and scope of equivalent elements are included in the present invention.

Claims

1. A method for monitoring bus capacitance of a motor drive system, characterized in that: The following steps are involved: Step 1: According to the state of the inverter side switch tube and the collected voltage and current signals, the current of the DC bus capacitor is Reconstructing it gives: ; In the formula, is the output current of the rectifier, The input current of the inverter is: ; In the formula, Represent the output current of the three phases ABC respectively, and Indicates the switch state of the inverter upper bridge arm IGBT, where "1" indicates the on state and "0" indicates the off state; Step 2: Construct a discrete model based on the capacitor equivalent circuit and convert it into a differential equation with unknown parameters as state variables; Step 3: Design a Butterworth bandpass filter to filter the DC bus voltage and reconstructed current to extract the 300Hz ripple on the DC side; Step 4: Propose an adaptive Kalman filter capacitance monitoring algorithm to dynamically adapt to the discrete model; Step 5: Convert the differential equation in step 2 into a measurement equation and construct a state equation with unknown parameters as state variables; Step 6: Propose an improved monitoring model based on the pseudo-measurement monitoring model to improve the convergence and accuracy of monitoring; Step 7: Substitute the improved monitoring model into the adaptive Kalman filter for solution.

2. The method for monitoring bus capacitance of a motor drive system according to claim 1, characterized in that: In step 2, based on the equivalent circuit of the capacitor, assume that the input is , the output is , then the transfer function of the discrete model is: ; Where ESC is the equivalent series capacitance, ESR is the equivalent series resistance, and s is the complex frequency domain transformation operator; Discretize the transfer function and use the input and output to estimate the aging parameters of the capacitor.

3. The method for monitoring bus capacitance of a motor drive system according to claim 2, characterized in that: Using a bilinear transformation: Where, T represents the artificially set sampling time, and z is the impedance transformation operator; Rewrite the transfer function of the capacitor discrete model: ; ； Convert the transfer function to a difference equation between the output and the input: ; In the formula, , is the voltage and current at the previous moment, , is the voltage and current at the next moment; use Calculate the aging parameters of the capacitor: .

4. The method for monitoring bus capacitance of a motor drive system according to claim 1, characterized in that: In step three, the Butterworth bandpass filter is used get; in, and are the Butterworth bandpass filter coefficients.

5. The method for monitoring bus capacitance of a motor drive system according to claim 4, characterized in that: Using Butterworth bandpass filter to filter DC bus voltage Process and obtain Ripple at line frequency of 300Hz.

6. The method for monitoring bus capacitance of a motor drive system according to claim 1, characterized in that: In Kalman filtering, a hyperparameter is introduced As an adjustment update parameter, the forgetting factor b is introduced to adjust the importance of historical data; Among them, the adjustment update parameter The relationship between and the forgetting factor b is: ; The process noise Q and measurement noise R in the Kalman filter can be rewritten as: ; ； In the formula, is the process noise of the kth iteration, is the measurement noise at the kth iteration.

7. The method for monitoring bus capacitance of a motor drive system according to claim 1, characterized in that: In step five, As a state variable, and rewrite the difference equation between output and input: ; in, It is the observation matrix constructed based on the signals collected by the sensor. It is a new state vector constructed by treating the unknown parameters of the system as unknown states. is the measurement error vector constructed to describe the error of the system sensor; Construct the state-space equations for the state variables: .

8. The method for monitoring bus capacitance of a motor drive system according to claim 1, characterized in that: The voltage measurement value at the previous moment is placed in the reconstructed measurement matrix and the state variable is used accordingly. The measurement equation is rewritten as: in, .

Images