Flywheel energy storage permanent magnet motor rotor position estimation method based on position winding
By establishing the data set and using the extended Kalman filtering algorithm, the problem of large position capture error of flywheel energy storage permanent magnet synchronous motor is solved, and the accuracy of position calculation and anti-interference ability are improved.
Patent Information
- Application Number
- CN202510108899.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-04-25
AI Technical Summary
In the prior art, the rotor position capture error of flywheel energy storage permanent magnet synchronous motor has a large rotor position capture, especially in the actual operating environment, which seriously interferes with and noise, affects the signal quality, and it is difficult to accurately extract the rotor position information.
By establishing a data set, the back EMF signal and feature vector of the permanent magnet motor are obtained, the current working state and working parameter information are matched, the rotor position is calculated using the extended Kalman filtering algorithm to improve the confidence probability of the original data and enhance the accuracy of position calculation.
It improves the accuracy of rotor position calculation, enhances position judgment ability in environments of severe signal interference and noise, and reduces the limitation of computing resources.
Smart Images
Figure CN119891850A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of motor operation and maintenance, and in particular to a method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding. Background Art
[0002] When a flywheel energy storage permanent magnet synchronous motor is in use, it is necessary to capture the rotor position information in real time in order to provide position guidance for solving the fault when a fault occurs later. In the prior art, the method for capturing the rotor position of a flywheel energy storage permanent magnet synchronous motor generally relies on a mechanical position sensor, which has a large position estimation error. Summary of the invention
[0003] In view of the shortcomings of the prior art, the present application provides a method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on a position winding. The embodiment of the present application establishes a data set, obtains the current working back electromotive force signal of the permanent magnet motor, matches the current working state and current working parameter information of the permanent magnet motor, and then performs position calculation based on the matched working parameter information, thereby improving the confidence probability of the original data used for position calculation, thereby improving the accuracy of subsequent rotor position calculation.
[0004] The above application objectives of this application are achieved through the following technical solutions: A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on a position winding comprises the following steps: Acquire a back electromotive force signal and a characteristic vector of a typical working state of the permanent magnet motor, wherein the elements in the characteristic vector are working parameters of the permanent magnet motor, and the typical working state includes a normal working state and a typical fault state; A data set is established according to the back electromotive force signal of the permanent magnet motor and its corresponding permanent magnet motor state label and feature vector; Get the current back electromotive force signal of the permanent magnet motor; Determine the current working state of the permanent magnet motor; When it is determined that the current working state of the permanent magnet motor is a fault state, a characteristic vector of the current fault state of the permanent magnet motor is obtained; The current rotor position of the permanent magnet motor is calculated according to the characteristic vector of the current fault state of the permanent magnet motor.
[0005] In summary, this application has the following beneficial technical effects: The embodiment of the present application establishes a data set, matches the current working state and current working parameter information of the permanent magnet motor by acquiring the current working back electromotive force signal of the permanent magnet motor, and then performs position calculation based on the matched working parameter information, thereby improving the confidence probability of the original data used for position calculation, thereby improving the accuracy of subsequent rotor position calculation. BRIEF DESCRIPTION OF THE DRAWINGS
[0006] Figure 1 is a schematic diagram of a simulation model of an implementation method of the present application; Figure 2 is a schematic diagram of observing speed in one embodiment of the present application; Figure 3 It is a schematic diagram of the observation position of one embodiment of the present application. DETAILED DESCRIPTION
[0007] The following is combined with Figure 1 -Attached Figure 3 This application is described in further detail.
[0008] In order to more clearly understand the technical solution presented in the embodiment of the present application, a brief introduction to the existing flywheel energy storage permanent magnet motor is first given.
[0009] Large flywheel energy storage systems are widely used as energy storage devices in the field of renewable energy storage. They include single flywheel energy storage units and array equipment. Since permanent magnet motors have the advantages of high power density, wide speed range and low loss, new flywheel energy storage systems based on permanent magnet motors have become the most common choice.
[0010] However, since mechanical position sensors are commonly used in flywheel permanent magnet synchronous motor control systems, mechanical sensors often have faults, including sensor damage, signal interference and installation problems, which will make it difficult for the motor control system to quickly and accurately obtain rotor position information. If the initial position is not in the correct position, the motor will cause jitter and reverse when starting, thus affecting the motor control and operation, and may cause the motor to fail to start normally, unstable speed or reduced energy conversion efficiency.
[0011] At present, the technical difficulties faced by the sensorless rotor position diagnosis of flywheel energy storage permanent magnet motors are: signal interference and noise in the actual operating environment seriously affect the signal quality used for diagnosis, making it difficult to accurately extract rotor position information; at the same time, motor parameters change with temperature, load, etc., affecting the accuracy of the diagnosis results; the signal-to-noise ratio is low during inertial operation, making it difficult to accurately determine the rotor position; finally, although complex algorithms can improve diagnostic accuracy, they are limited by computing resources.
[0012] In view of the above technical problems, an embodiment of the present application provides a method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on a position winding, comprising the following steps: S101, obtaining a back electromotive force signal and a characteristic vector of a historical typical working state of a permanent magnet motor (PMSM), wherein the elements in the characteristic vector are working parameters such as voltage, speed, harmonic content, and phase of the permanent magnet motor, and the typical working state includes a normal working state and several typical fault states; S102, establishing a data set based on historical typical back electromotive force signals of permanent magnet motors and their corresponding permanent magnet motor state labels and feature vectors, wherein the permanent magnet motor state labels include normal states and several medium fault states; S103, obtaining the current back electromotive force signal of the permanent magnet motor; S104, judging the current working state of the permanent magnet motor based on the data set in step S102; S105, when it is determined that the current working state of the permanent magnet motor is a fault state, obtaining a characteristic vector of the current fault state of the permanent magnet motor; S106. Calculate the current rotor position of the permanent magnet motor according to the characteristic vector of the current fault state of the permanent magnet motor.
[0013] The following is a further introduction based on specific usage scenarios.
[0014] First, step S101 is performed to obtain typical working state types of permanent magnet motors and back electromotive force signals and characteristic vectors corresponding to various typical working state types. The typical working state includes a normal working state and several typical fault states. The typical fault states of permanent magnet motors include winding short circuit, bearing fault, permanent magnet demagnetization, etc. The characteristic vector is a multidimensional vector. The elements in the characteristic vector are working parameters such as voltage, speed, harmonic content, and phase of the permanent magnet motor. The typical working state of the permanent magnet motor and its corresponding back electromotive force signal and characteristic vector can be extracted through ANSYS flywheel permanent magnet synchronous motor simulation; Then, step S102 is performed to establish a data set based on the historical typical back-electromotive force signal of the permanent magnet motor and its corresponding permanent magnet motor state label and feature vector. The aforementioned permanent magnet motor state label includes a normal state and several medium fault states. In the embodiment of the present application, the back-electromotive force signal of the typical working state of the permanent magnet motor can be converted into a frequency spectrum, and preliminary feature extraction is performed. According to the frequency spectrum of the permanent magnet motor and its corresponding permanent magnet motor state label and feature vector, a data set is established; Step S103, obtain the current back electromotive force signal of the permanent magnet motor, and convert the current back electromotive force signal of the permanent magnet motor into a spectrum diagram. In the embodiment of the present application, the current back electromotive force signal of the permanent magnet motor rotor can be obtained through the detection winding, and the detection winding is arranged on the rotor core of the permanent magnet motor. The detection winding on the rotor core is arranged with two phases, and there is a vector angle difference of 120 degrees between the two position windings. The diagnostic winding of each phase occupies three slots, and the position detection winding and the main winding are in the same slot together. The functions of the position detection winding and the main winding are independent of each other. The back electromotive force signal of the permanent magnet motor is obtained through the position detection winding. When the motor is running, the rotation of the rotor will induce a specific voltage signal in the position winding, and the main winding is responsible for the normal operation and power output of the generator. The position winding and the main winding co-slot design adopted in the present application will not interfere with the normal operation of the permanent magnet motor, and there is no need to add additional energy-consuming equipment. At the same time, compared with the split signal detection equipment in the prior art, the probability of damage to the position detection winding is relatively small.
[0015] Step S104, based on the data set in step S102, determine the current working state of the permanent magnet motor, and when the current working state of the permanent magnet motor is normal, return to step S103 to continue monitoring the back electromotive force signal of the permanent magnet motor in the current working state; Step S105: when it is determined that the current working state of the permanent magnet motor is a fault state, obtaining a feature vector corresponding to the current fault state label of the permanent magnet motor; Step S106, calculate the current rotor position of the permanent magnet motor according to the characteristic vector of the current fault state of the permanent magnet motor. In the embodiment of the present application, based on the characteristic vector of the current fault state of the permanent magnet motor obtained, the extended Kalman filter algorithm (EKF) is used to calculate the current rotor position of the permanent magnet motor, and the extended Kalman filter (EKF) algorithm of the permanent magnet synchronous motor (PMSM) is established in the stationary coordinate system. Combined with the characteristic vector of the permanent magnet synchronous motor, the position information of the rotor is calculated. In the embodiment of the present application, a two-pole, thirty-six-slot, three-phase star-connected permanent magnet synchronous motor is taken as an example. The voltage equation of the permanent magnet synchronous motor in the stationary coordinate system is: By transforming the above formula, we can get the current equation: Where R is the stator resistance, = = is the stator inductance; , , , , are the stator voltage and current; Considering the rotor's moment of inertia J to be infinite, we can get: Combined with the following formula: The state equation of the system can be obtained: in, , , Measurement equation: , Using the EKF algorithm, its discrete form is: Where W(k) is the system noise; V(k) is the measurement noise. Assuming that W(k) and V(k) are zero-mean white noise, the expectation is: System noise and measurement noise cannot be precisely controlled. In the EKF algorithm, uncertainty is generally measured by its own covariance matrix. The W covariance matrix Q and the V covariance matrix R are Assuming that W(k) and V(k) are unrelated, the initial position x(0) is a random vector and is also unrelated to W(k) and V(k), the state equation is linearized and Taylor expansion is performed at time t to obtain The corresponding Jacobian matrix is: Substitute this matrix into the Figure 1 The simulation model PMSM control block diagram is shown for calculation.
[0016] In general, the embodiment of the present application establishes a data set, matches the current working state and current working parameter information of the permanent magnet motor by acquiring the current working back electromotive force signal of the permanent magnet motor, and then performs position calculation based on the matched working parameter information, thereby improving the confidence probability of the original data used for position calculation, thereby improving the accuracy of subsequent rotor position calculation.
[0017] As a feasible specific implementation method of the embodiment of the present application, the determination of the current working state of the permanent magnet motor includes the following steps: Step S201: Establishing a fault detection model; Step S202: using the back electromotive force signal spectrum of the permanent magnet motor in the data set in the typical working state as input and the permanent magnet motor working state label in the data set as supervision to train the fault detection model; Step S203: Save the trained model parameters; Step S204: obtaining the current back electromotive force signal spectrum of the permanent magnet motor as input, and using the fault detection model to determine the current working state of the permanent magnet motor.
[0018] Step S205: based on the current working state of the permanent magnet motor, obtaining a characteristic vector of the current fault state of the permanent magnet motor.
[0019] The current working state of the permanent magnet motor is detected by means of a fault detection model, which further improves the confidence probability of the current working state and parameters of the permanent magnet motor, thereby improving the accuracy of subsequent rotor position calculations. At the same time, the use of a fault detection model will increase the speed of extracting the current working state and parameters of the permanent magnet motor.
[0020] In step S202, training the fault detection model includes the following steps: Step S2021: extracting a feature vector of a typical working state of the permanent magnet motor through a CNN convolutional neural network; Step S2022: calculating the Euclidean distance between the characteristic vectors of various working states of the permanent magnet motor; Step S2023: continuously selecting triples by updating the triplet loss function online; Step S2024: judging whether the fault detection model converges according to the triplet loss function; Step S2025: when the fault detection model does not converge, updating the network weights of the fault detection model by using the BP algorithm; Step S2026: When the fault detection model converges, the trained model parameters are saved.
[0021] By continuously comparing the characteristic vectors of various working states of the permanent magnet motor and selecting triplets, similar fault state types can be merged, the output speed of the fault detection model can be improved, and the calculation efficiency of the rotor position can be improved.
[0022] Based on the above description, in the embodiment of the present application, obtaining the characteristic vector of the current fault state of the permanent magnet motor includes the following steps: Step S2051: inputting the back electromotive force signal spectrum diagram of the permanent magnet motor in the current working state into the fault detection model; Step S2052: the fault detection model outputs the working state label and feature vector corresponding to the back electromotive force signal spectrum diagram of the permanent magnet motor in the current working state; Step S2053: The characteristic vector output by the fault detection model is used as the characteristic vector of the current fault state of the permanent magnet motor. Fault detection provides more accurate initial state estimation and system parameter estimation for the extended Kalman filter optimization algorithm, thereby improving the accuracy of position judgment.
[0023] As a feasible specific implementation method of the embodiment of the present application, the determination of the current working state of the permanent magnet motor includes the following steps: Step S301: establishing a template library according to the permanent magnet motor working state labels and corresponding feature vectors in the data set; Step S302: extracting a characteristic vector of the current working state of the permanent magnet motor through a fault detection model; Step S303: calculating the Euclidean distance between the current feature vector of the permanent magnet motor and the feature vector corresponding to each working state label in the template library; Step S304: Select the minimum value of the aforementioned Euclidean distance, and determine whether the minimum value of the Euclidean distance is greater than a threshold; Step S305: when the minimum value of the Euclidean distance is greater than the threshold, assign a new label to the fault type corresponding to the current feature vector of the permanent magnet motor, and add the new label and the corresponding feature vector to the template library; Step S306: When the minimum Euclidean distance is less than the threshold, the diagnosis result is output.
[0024] By collecting new fault parameters of the permanent magnet motor to enrich the template library, it can be adapted to the complex working conditions of the permanent magnet motor. The dynamic update of the template library can also enable the permanent magnet motor to quickly calculate and output the rotor position when the same fault occurs subsequently.
[0025] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.
Claims
1. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding, characterized in that: include: Acquire a back electromotive force signal and a characteristic vector of a typical working state of the permanent magnet motor, wherein the elements in the characteristic vector are working parameters of the permanent magnet motor, and the typical working state includes a normal working state and a typical fault state; A data set is established according to the back electromotive force signal of the permanent magnet motor and its corresponding permanent magnet motor state label and feature vector; Get the current back electromotive force signal of the permanent magnet motor; Determine the current working state of the permanent magnet motor; When it is determined that the current working state of the permanent magnet motor is a fault state, a characteristic vector of the current fault state of the permanent magnet motor is obtained; The current rotor position of the permanent magnet motor is calculated according to the characteristic vector of the current fault state of the permanent magnet motor.
2. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 1, characterized in that: A position detection winding is arranged on the permanent magnet motor rotor. The position detection winding has two phases. There is a vector angle difference of 120 degrees between the two position windings. The diagnostic winding of each phase occupies three slots. The position detection winding and the main winding are in the same slot. The functions of the position detection winding and the main winding are independent of each other. The back electromotive force signal of the permanent magnet motor is obtained through the position detection winding.
3. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 2, characterized in that: The calculation of the current rotor position of the permanent magnet motor comprises: Based on the acquired characteristic vector of the current fault state of the permanent magnet motor, the extended Kalman filter algorithm is used to calculate the current rotor position of the permanent magnet motor.
4. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 3, characterized in that: Also includes: The back electromotive force signal of the typical working state of the permanent magnet motor is converted into a frequency spectrum, and a data set is established according to the frequency spectrum of the permanent magnet motor and its corresponding permanent magnet motor state label and feature vector; The current back electromotive force signal of the permanent magnet motor is converted into a spectrum diagram.
5. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 4, characterized in that: Determining the current working state of the permanent magnet motor includes: Establish a fault detection model; The back-EMF signal spectrum of the permanent magnet motor in the typical working state in the dataset is used as input, and the working state label of the permanent magnet motor in the dataset is used as supervision to train the fault detection model; Save the trained model parameters; The current back-electromotive force signal spectrum of the permanent magnet motor is obtained as input, and the fault detection model is used to determine the current working state of the permanent magnet motor.
6. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 5, characterized in that: The training fault detection model also includes: The feature vector of the typical working state of the permanent magnet motor is extracted through a convolutional neural network; Calculate the Euclidean distance between the characteristic vectors of various working states of permanent magnet motors; Continuously select triplets by updating the triplet loss function online; Determine whether the fault detection model converges based on the triplet loss function; When the fault detection model does not converge, the network weights of the fault detection model are updated through the BP algorithm; When the fault detection model converges, the trained model parameters are saved.
7. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 6, characterized in that: Also includes: A template library is established based on the permanent magnet motor working state labels and corresponding feature vectors in the data set; Extract the characteristic vector of the current working state of the permanent magnet motor through the fault detection model; Calculate the Euclidean distance between the current feature vector of the permanent magnet motor and the feature vector corresponding to each working state label in the template library; Select the minimum value of the aforementioned Euclidean distance, and determine whether the minimum value of the Euclidean distance is greater than a threshold; When the minimum value of the Euclidean distance is greater than the threshold, the fault type corresponding to the current feature vector of the permanent magnet motor is assigned a new label, and the new label and the corresponding feature vector are added to the template library; When the minimum Euclidean distance is less than the threshold, the diagnosis result is output.
8. A method for estimating the rotor position of a flywheel energy storage permanent magnet motor based on position winding according to claim 6, characterized in that: The method of obtaining the characteristic vector of the current fault state of the permanent magnet motor includes: Input the back electromotive force signal spectrum of the permanent magnet motor in the current working state into the fault detection model; The fault detection model outputs the working state label and feature vector corresponding to the back electromotive force signal spectrum diagram of the permanent magnet motor in the current working state; The feature vector output by the fault detection model is used as the feature vector of the current fault state of the permanent magnet motor.