Method and system for implementing model-free predictive current control of synchronous reluctance motor based on neural network disturbance observer

Abstract

The application discloses a synchronous reluctance motor model-free predictive current control method based on a neural network disturbance observer, aims to solve the problems of slow speed response and poor control effect of the synchronous reluctance motor under complex working conditions, and realizes accurate control of the synchronous reluctance motor; meanwhile, the control method provides a neural network disturbance observer to observe disturbances caused by factors such as motor parameter mismatch, and compensates the disturbances to a prediction model as lumped disturbances, so that the parameter robustness of the control system is improved. The biggest feature of the neural network disturbance observer is to adopt the ideas of pure integration, compensation and transfer function type, and a super local model is introduced to construct a motor mathematical model, so that the robustness of the system is improved.

Description

Technical Field

[0001] This invention belongs to the field of motor control technology and relates to a method and system for realizing model-free predictive current control of synchronous reluctance motor based on a neural network interference observer. Background Technology

[0002] Synchronous reluctance motors (SynRMs), due to their unique rotor structure and the absence of permanent magnets (excluding permanent magnet assisted synchronous reluctance motors), offer high performance and are suitable for various harsh environments. They also outperform switched reluctance motors in terms of torque ripple, harmonic content, and power density, attracting increasing attention from researchers. Therefore, higher demands are placed on the control performance of SynRMs when facing complex operating conditions, parameter variations due to high temperatures, or external disturbances.

[0003] In recent years, synchronous reluctance motor control methods based on Model Predictive Control (MPC) have been regarded as an effective alternative to traditional SynRM control methods due to their intuitive concepts and ease of implementation. MPC achieves fast and stable control based on the mathematical model of the controlled object and a cost function composed of control constraints.

[0004] However, existing MPC-based synchronous reluctance motor control methods have some significant drawbacks:

[0005] First, its predictive control relies on the motor body model to construct predictive models for various control objectives, which is easily affected by the body parameters. At the same time, the magnetic saturation phenomenon of the synchronous reluctance motor leads to nonlinear changes in motor parameters. Both of these will cause motor parameter mismatch problems, which reduces the robustness of the control algorithm.

[0006] Secondly, MPC applies a single voltage vector and continuously optimizes it during the control cycle to give the motor good dynamic performance, but the motor's steady-state performance is poor. Summary of the Invention

[0007] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a method and system for model-free predictive current control of synchronous reluctance motors based on a neural network interference observer. The aim is to solve the technical problems of existing MPC-based synchronous reluctance motor control methods, which rely on a motor body model to construct predictive models for various control objectives, making them susceptible to the influence of body parameters; and the nonlinear changes in motor parameters caused by magnetic saturation of the synchronous reluctance motor, leading to motor parameter mismatch and reduced robustness of the control algorithm; as well as the poor steady-state performance of the motor.

[0008] To achieve the above objectives, according to one aspect of the present invention, a method for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer is provided, comprising the following steps:

[0009] (1) The three-phase voltage signal u of the synchronous reluctance motor in the kth sampling period is obtained by sampling with a voltage sensor. a (k), u b (k), u c (k), the position signal θ(k) of the synchronous reluctance motor in the kth sampling period is obtained through the position sensor, and the three-phase voltage signal u is adjusted according to the position signal θ(k). a (k), u b (k), u c (k) Perform coordinate transformation to obtain the d-axis and q-axis voltage signals u in the k-th sampling period. d (k), u q (k), the d-axis and q-axis voltage signals u in the kth sampling period d (k), u q (k) The inputs are weighted and processed by the hyperlocal model. The processing results are then compared with the d-axis and q-axis interference observations output by the neural network interference observer in the kth sampling period. Add them together to obtain the differential signals of the d-axis and q-axis currents in the k-th sampling period, respectively. and The first-order forward Euler discretization method was used to discretize the d-axis and q-axis current differential signals respectively. Combined with the current obtained by the current sensor sampling, the d-axis and q-axis discrete current prediction values ​​of the k-th sampling period output by the hyperlocal model were obtained respectively. Where k is a positive integer.

[0010] (2) Based on the predicted values ​​of the discrete currents i along the d and q axes of the hyperlocal model obtained in step (1) for the (k+1)th sampling period. d (k+1),i q (k+1) Obtain the first optimal voltage vector u in the kth sampling period. g1d (k), u g1q (k), based on the optimal voltage vector u g1d (k), u g1q (k) Calculate the voltage difference ΔE u1d (k), ΔE u1q (k) Obtain the second and third optimal voltage vectors u along the d and q axes by looking up the voltage difference in a table. g2d (k), u g2q (k) and u g2q (k), u g3q(k) uses the direct calculation method of duty cycle to obtain the d and q axis duty cycles d of the kth sampling period after the hyperlocal model optimization. nd (k), d nq (k), and based on the obtained duty cycle d nd (k) and d nq (k) The control pulse generator outputs a drive signal to the inverter so that the inverter outputs a voltage to the synchronous reluctance motor according to the drive signal, thereby driving the synchronous reluctance motor to run, where n∈[1,3].

[0011] (3) The d-axis and q-axis current signals i of the synchronous reluctance motor in the (k+1)th sampling period are obtained by sampling with a current sensor. d (k+1),i q (k+1), based on the d-axis and q-axis current signals i d (k+1),i q (k+1) Obtain the d-axis and q-axis current estimation errors e in the (k+1)th sampling period respectively. d (k+1) and e q (k+1) Based on the obtained d-axis and q-axis current estimation errors, obtain the d-axis and q-axis error output matrix for the (k+1)th sampling period. Estimation of the d-axis and q-axis weights for the (k+1)th sampling period The obtained error output matrix Estimation of weights Multiply to obtain an estimate of the interfering observations. The synchronous reluctance motor is controlled based on the estimated value of the disturbance observation.

[0012] Preferably, step (1) includes the following sub-steps:

[0013] (1-1) The three-phase voltage signal u of the synchronous reluctance motor in the kth sampling period is obtained by sampling with a voltage sensor. a (k), u b (k), u c (k), the position signal θ(k) of the synchronous reluctance motor in the kth sampling period is obtained through the position sensor, and the three-phase voltage signal u in the kth sampling period is calculated based on the position signal θ(k) of the kth sampling period. a (k), u b (k), u c (k) Perform coordinate transformation to obtain the d-axis and q-axis voltage signals u in the k-th sampling period. d (k), u q (k).

[0014] (1-2) The d-axis and q-axis voltage signals u obtained in step (1-1) for the kth sampling period d(k), u q (k) Use the corresponding scaling factor α d α q The values ​​are weighted separately and then compared with the interference observation values ​​output by the neural network interference observer in the kth sampling period. Add them together to obtain the d-axis and q-axis current differential signals of the k-th sampling period, respectively. and

[0015] (1-3) Based on the d-axis and q-axis currents of the k-th sampling period and the d-axis and q-axis voltage signals u obtained in step (1-1) of the k-th sampling period d (k), u q (k) and using the first-order forward Euler discretization method, the d-axis and q-axis current differential signals obtained in (1-2) for the kth sampling period are respectively processed. Discretization is performed to obtain the discrete current prediction values ​​of the d and q axes in the (k+1)th sampling period of the hyperlocal model output.

[0016] Preferably, step (1-2) uses the following formula:

[0017]

[0018] Where α d =1 / L d and α q =1 / L q α represents the weighting factor. d =1 / L d and α q =1 / L q These values ​​were all obtained through experience-based debugging.

[0019] Steps (1-3) use the following formula:

[0020]

[0021] Where T s This represents the system sampling time.

[0022] Preferably, the d-axis and q-axis interference observations in the k-th sampling period It is obtained through the following sub-steps:

[0023] (a) Obtain the currents a0, a1, ..., a0 at different load startup times using a traditional model-predicted current control model. x And based on the currents a0, a1, ..., a xObtain the actual load disturbance current error set F = [d1, d2, ..., d], which consists of all actual load disturbance current errors generated after the start-up of all loads in the model predictive current control model. x ], where x represents the total number of loads in the model predictive current control model.

[0024] (b) Randomly select multiple load disturbance current errors from the load disturbance current error set F obtained in step (a) n times to form a data set. For each data set p (where p∈[1,3]), obtain all load disturbance current errors in the p-th data set. The mean of the values ​​is used as the center of the Gaussian function corresponding to the p-th data set. Where n is a positive integer less than or equal to x, m p This represents the total number of load disturbance current errors in the p-th data set;

[0025] (c) For each data set p obtained in step (b), obtain the center of the Gaussian function corresponding to each load disturbance current error in the data set p. The Euclidean distance between the data points is calculated, and the sum of all the Euclidean distances is taken as the average to obtain the Gaussian function center b corresponding to the p-th data set. p And obtain the load disturbance current error set consisting of all load disturbance current errors in the p-th data set obtained in step (1-2). The load disturbance current error of each data set corresponds to the center of the Gaussian function of the p-th data set. The difference between them is calculated, and the sum of all the differences is taken as the average to obtain the center width b of the neural network. n .

[0026] (d) For each data set p obtained in step (b), the m load disturbance current errors in the p-th data set are... Input a Gaussian function for calculation and obtain multiple error outputs. Construct the initial error output matrix H0;

[0027] (e) Obtain the ideal weights based on the initial error output matrix H0 obtained in step (d) and the actual load disturbance current error set F obtained in step (a).

[0028] (f) Based on the output matrix H0 obtained in step (d) and the ideal weights obtained in step (e) Estimate the interference observations for the first sampling period

[0029] (g) Estimation of the disturbance observations obtained in step (f) As initial values, the neural network interference observer is input for multiple iterations to obtain the interference observation value for the k-th sampling period.

[0030] Preferably, step (a) uses the following formula:

[0031]

[0032] Where a0 is the model predicted current control model current under no-load conditions, and d x For a x The corresponding load disturbance current error.

[0033] Step (e) uses the following formula:

[0034]

[0035] in, It is the transpose of the initial error output matrix H0.

[0036] Step (f) uses the following formula:

[0037]

[0038] Where ε is the estimation error of the neural network perturbation observer, and ε ≤ ε n , ε n This represents the estimation error of the neural network interference observer after n iterations.

[0039] Preferably, step (2) includes the following sub-steps:

[0040] (2-1) Based on the predicted values ​​of the discrete currents along the d and q axes in the (k+1)th sampling period of the hyperlocal model output obtained in step (1) Obtain the ideal voltage vectors along the d and q axes in the kth sampling period.

[0041] (2-2) Based on the ideal voltage vectors of the d and q axes obtained in step (2-1) for the kth sampling period And the position signal θ(k) of the kth sampling period, to obtain the voltage cost functions of the α and β axes.

[0042] (2-3) Substitute the effective voltage vector u(k) of the kth sampling period into the α and β axis voltage cost functions obtained in step (2-2) for calculation, and obtain the values ​​that make g α and g β The minimum voltage vector is taken as the first optimal voltage vector u along the d and q axes in the k-th sampling period. g1d (k), u g1q(k), the ideal voltage vector of the kth sampling period is compared with the first optimal voltage vector u on the d and q axes respectively. g1d (k), u g1q (k) Calculate the difference to obtain the voltage difference ΔE between the d and q axes in the kth sampling period. u1d (k), ΔE u1q (k).

[0043] (2-4) Divide the voltage vector distribution diagram of the pulse generator into 6 sectors, and based on the d-axis and q-axis voltage vector errors ΔE obtained in step (2-3). u1d (k), ΔE u1q (k) Find the corresponding sector in the voltage vector distribution diagram, and look up the corresponding second voltage vector u of the d and q axes in the pre-established second optimal voltage vector table for the kth sampling period based on the obtained sector. g2d (k), u g2q (k), and respectively query the corresponding d-axis and q-axis third voltage vector u of the k-th sampling period in the pre-established third optimal voltage vector table according to the obtained sector. g3d (k), u g3q (k).

[0044] (2-5) Based on the nth voltage vector u of the d and q axes obtained in step (2-4) for the kth sampling period respectively gnd (k) and u gnd (k) respectively obtains the predicted current values ​​i of the d and q axes for the kth sampling period corresponding to the nth voltage vector of the d and q axes. dn (k+1) and i qn (k+1).

[0045] (2-6) The d-axis and q-axis reference currents obtained in step (2-1) for the kth sampling period and The predicted current values ​​i on the d and q axes corresponding to the nth voltage vector on the d and q axes obtained in steps (2-5) during the kth sampling period are respectively compared with those on the d and q axes. dn (k+1) and i qn Divide by (k+1) to obtain the nth voltage vector u. gnd (k) and u gnd (k) represents the current error Δε along the d and q axes for the kth sampling period, respectively. dn (k) and Δε qn (k).

[0046] (2-7) Based on the nth voltage vector u obtained in step (2-6) gnd (k) and u gnd (k) represents the d-axis and q-axis current errors Δε corresponding to the kth sampling period. dn(k), Δε qn (k) Obtain the nth voltage vector u gnd (k) and u gnd (k) respectively correspond to the duty cycle d nd (k) and d nq (k).

[0047] (2-8) Based on the duty cycle d obtained in step (2-7) nd (k) and d nq (k) The control pulse generator outputs a drive signal to the inverter, so that the inverter outputs a voltage to the synchronous reluctance motor according to the drive signal, thereby driving the synchronous reluctance motor to run.

[0048] Preferably, step (2-1) uses the following formula:

[0049]

[0050] in, Let be the d-axis and q-axis reference current for the k-th sampling period, and have e d (k), e q (k) represents the estimation error of the d-axis and q-axis currents in the k-th sampling period, with an initial value of 0; m is the error coefficient, with a value of 19000.

[0051] Step (2-2) uses the following formula:

[0052]

[0053] Where u α (k), u β (k) is the effective voltage vector of the kth sampling period of the α and β axes.

[0054] The calculation of the α-axis and β-axis voltage difference in the k-th sampling period in step (2-3) uses the following formula:

[0055]

[0056] Steps (2-5) involve calculating using the following formula:

[0057]

[0058] Steps (2-6) are calculated using the following formula:

[0059]

[0060] Steps (2-7) involve calculating using the following formula:

[0061]

[0062] Where d 1d (k) represents the d-axis u g1d (k) corresponds to the duty cycle, d 2d (k) represents u g2d (k) corresponds to the duty cycle, d 3d (k) represents u g3d (k) corresponds to the duty cycle, d 1q (k) represents the q-axis u g1q (k) corresponds to the duty cycle, d 2q (k) represents u g2q (k) corresponds to the duty cycle, d 3q (k) represents u g3q (k) corresponds to the duty cycle.

[0063] Preferably, step (3) includes the following sub-steps:

[0064] (3-1) Obtain the ABC three-phase current signal i in the (k+1)th sampling period after the synchronous reluctance motor starts running using a current sensor. a (k+1),i b (k+1),i c (k+1), and based on the position signal θ(k+1) of the synchronous reluctance motor in the (k+1)th sampling period obtained from the position sensor, obtain the d-axis and q-axis current signals i in the (k+1)th sampling period. d (k+1) and i q (k+1).

[0065] (3-2) Based on the ideal voltage vectors of the d and q axes obtained in step (2-1) for the kth sampling period and Obtain the d-axis and q-axis current estimates for the k-th sampling period respectively. and

[0066] (3-3) Obtain the d-axis and q-axis currents i in the (k+1)th sampling period from step (3-1). d (k+1) and i q (k+1) are compared with the d-axis and q-axis current estimates obtained in step (3-2) for the k-th sampling period. and The difference is calculated to obtain the d-axis and q-axis current estimation errors e in the (k+1)th sampling period, respectively. d (k+1) and e q (k+1).

[0067] (3-4) Determine the estimation errors e of the d-axis and q-axis currents obtained in step (3-3) for the (k+1)th sampling period. d (k+1),eq If all (k+1) are 0, it means that the synchronous reluctance motor is running smoothly and the process ends; otherwise, proceed to step (3-5).

[0068] (3-5) Estimating the d-axis and q-axis currents based on the values ​​obtained in step (3-3) for the (k+1)th sampling period, respectively. d (k+1) and e q (k+1), and use the Gaussian function to obtain the error output h corresponding to the d and q axes of the (k+1)th sampling period respectively. d (k+1) and h q (k+1), and output h using the error corresponding to the d and q axes. d (k+1) and h q (k+1) Obtain the error output matrix for the (k+1)th sampling period.

[0069] (3-6) Based on the error output matrix H(k+1) of the (k+1)th sampling period obtained in step (3-5), obtain the weight estimate for the (k+1)th sampling period.

[0070] (3-7) Combine the error output matrix H(k+1) of the (k+1)th sampling period obtained in step (3-5) with the weight estimate of the (k+1)th sampling period obtained in step (3-6). Multiply to obtain an estimate of the interference observations in the (k+1)th sampling period. Then return to step (3-1).

[0071] Preferably, step (3-1) uses the following formula:

[0072]

[0073] The d-axis and q-axis current estimates in step (3-2) are obtained using the following formulas:

[0074]

[0075] Step (3-3) involves calculating the current estimation error using the following formula:

[0076]

[0077] Steps (3-5) use the following formula:

[0078]

[0079] in, The centers of the Gaussian function on the d and q axes are respectively; b d b qThese represent the center widths of the neural network along the d and q axes, respectively.

[0080] Steps (3-6) involve using the following formula:

[0081]

[0082] in It is the derivative of the weight estimate for the (k+1)th sampling period; It is the derivative of the error in estimating the weights for the (k+1)th sampling period, and satisfies W * (k) is the ideal value of the weight in the k-th sampling period. It is an estimate of the weights for the (k+1)th sampling period; e T (k+1) is the transpose of e(k+1). γ is a positive constant.

[0083] According to another aspect of the present invention, a system for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer is provided, comprising:

[0084] The first module is used to sample and acquire the ABC three-phase voltage signal u of the synchronous reluctance motor in the kth sampling period through a voltage sensor. a (k), u b (k), u c (k), the position signal θ(k) of the synchronous reluctance motor in the kth sampling period is obtained through the position sensor, and the three-phase voltage signal u is adjusted according to the position signal θ(k). a (k), u b (k), u c (k) Perform coordinate transformation to obtain the d-axis and q-axis voltage signals u in the k-th sampling period. d (k), u q (k), the d-axis and q-axis voltage signals u in the kth sampling period d (k), u q (k) The inputs are weighted and processed by the hyperlocal model. The processing results are then compared with the d-axis and q-axis interference observations output by the neural network interference observer in the kth sampling period. Add them together to obtain the differential signals of the d-axis and q-axis currents in the k-th sampling period, respectively. and The first-order forward Euler discretization method was used to discretize the d-axis and q-axis current differential signals respectively. Combined with the current obtained by the current sensor sampling, the d-axis and q-axis discrete current prediction values ​​of the k-th sampling period output by the hyperlocal model were obtained respectively. Where k is a positive integer.

[0085] The second module is used to predict the discrete current values ​​i along the d and q axes in the (k+1)th sampling period of the hyperlocal model obtained from the first module. d (k+1),i q (k+1) Obtain the first optimal voltage vector u in the kth sampling period. g1d (k), u g1q (k), based on the optimal voltage vector u g1d (k), u g1q (k) Calculate the voltage difference ΔE u1d (k), ΔE u1q (k) Obtain the second and third optimal voltage vectors u along the d and q axes by looking up the voltage difference in a table. g2d (k), u g2q (k) and u g2q (k), u g3q (k) uses the direct calculation method of duty cycle to obtain the d and q axis duty cycles d of the kth sampling period after the hyperlocal model optimization. nd (k), d nq (k), and based on the obtained duty cycle d nd (k) and d nq (k) The control pulse generator outputs a drive signal to the inverter so that the inverter outputs a voltage to the synchronous reluctance motor according to the drive signal, thereby driving the synchronous reluctance motor to run, where n∈[1,3].

[0086] The third module is used to sample and acquire the d-axis and q-axis current signals i of the synchronous reluctance motor in the (k+1)th sampling period using a current sensor. d (k+1),i q (k+1), based on the d-axis and q-axis current signals i d (k+1),i q (k+1) Obtain the d-axis and q-axis current estimation errors e in the (k+1)th sampling period respectively. d (k+1) and e q (k+1) Based on the obtained d-axis and q-axis current estimation errors, obtain the d-axis and q-axis error output matrix for the (k+1)th sampling period. Estimation of the d-axis and q-axis weights for the (k+1)th sampling period The obtained error output matrix Estimation of weights Multiply to obtain an estimate of the interfering observations. The synchronous reluctance motor is controlled based on the estimated value of the disturbance observation.

[0087] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0088] (1) Since the present invention adopts step (1) combined with step (3), it uses a model-free control method to estimate and update the system disturbance in real time through the current estimation error. This method does not require identification of the specific motor parameters of the controlled object, so it can solve the problem of reduced robustness caused by motor parameter mismatch in the existing MPC-based synchronous reluctance motor control method.

[0089] (2) Because the present invention adopts step (2), the inverter output voltage under the three-vector MPC strategy is closer to the ideal voltage vector, thus solving the problem of poor motor steady-state performance of the existing MPC-based synchronous reluctance motor control method.

[0090] (3) This invention employs sub-steps (2-3) to (2-5) to improve steady-state performance by applying three voltage vectors under the three-vector MPC strategy. By designing the voltage cost function and re-dividing the vector sectors, the selection process of the three-vector combination is optimized, reducing the complexity of algorithm implementation.

[0091] (4) Since the present invention employs sub-steps (1-3), if the sampling time is very short, the hyperlocal model can be approximated as a real-time updated system model, thus ensuring the continuity of system operation. Attached Figure Description

[0092] Figure 1 This is a control block diagram of the method for realizing model-free predictive current control of synchronous reluctance motor based on neural network interference observer according to the present invention;

[0093] Figure 2 The voltage vector distribution diagram obtained in steps (2-4) of the method of the present invention is shown;

[0094] Figure 3 This is a flowchart of the method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer, according to the present invention.

[0095] Figure 4 This is a flowchart of the offline training process for a neural network interference observer. Detailed Implementation

[0096] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0097] This invention discloses a model-free predictive current control method for synchronous reluctance motors (SRRMs) based on a neural network disturbance observer. Its aim is to address the problems of slow speed response and poor control performance of SRRMs under complex operating conditions, thereby achieving accurate control of the SRRM. Simultaneously, this control method provides a neural network disturbance observer to observe disturbances caused by factors such as motor parameter mismatch, and incorporates these disturbances as lumped disturbances into the predictive model to improve the parameter robustness of the control system. The most significant feature of the neural network disturbance observer is its use of pure integral, compensation, and transfer function types, and the introduction of a hyperlocal model to construct the mathematical model of the motor, further enhancing system robustness.

[0098] This invention provides a method for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer, which is applied to... Figure 1 The control system of the synchronous reluctance motor shown includes a signal detection circuit, a main circuit, and a control circuit. The main circuit includes a three-phase inverter and a synchronous reluctance motor, with the three-phase inverter primarily used to drive the synchronous reluctance motor. The signal detection circuit includes a current detection circuit and a rotary encoder, mainly used to detect the current and rotor position signals of the synchronous reluctance motor. The control circuit includes a coordinate transformation module, a neural network interference observer, a hyperlocal model prediction module, an optimal vector selection module, a duty cycle calculation module, and a pulse generator module, mainly used to process the signals obtained from the signal detection circuit to obtain the control signals for the main circuit.

[0099] The rotary encoder detects the rotor position angle θ of the synchronous reluctance motor, and the mechanical angular velocity ω of the motor is obtained by differentiation. m ; Motor mechanical angular velocity setpoint and the mechanical angular velocity feedback value ω of the motor m The q-axis current setpoint is obtained after subtraction and passing through a PI controller. The control circuit uses the current detection circuit to detect the three-phase current i obtained from the synchronous reluctance motor. a i b i c After processing by the coordinate transformation module, i is obtained from the rotor position angle θ. d i q i d i q After processing by the neural network disturbance observer, the lumped perturbation component F in the hyperlocal model is obtained. d F q Observed values; current i d i q Total disturbance F d F qThe predicted current value is obtained after processing by the hyperlocal model prediction module; the predicted dq-axis current value and the reference dq-axis current value are substituted into the optimal vector selection module to obtain the optimal voltage vector u. g1 u g2 The duty cycles d1 and d2 of the corresponding voltage vectors are calculated through the duty cycle; finally, the switching sequence of the three-phase inverter is obtained through the pulse generator module to control the synchronous reluctance motor.

[0100] like Figure 3 As shown, this invention provides a method for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer, comprising the following steps:

[0101] (1) The three-phase voltage signal u of the synchronous reluctance motor in the kth sampling period is obtained by sampling with a voltage sensor. a (k), u b (k), u c (k), the position signal θ(k) of the synchronous reluctance motor in the kth sampling period is obtained through the position sensor, and the three-phase voltage signal u is adjusted according to the position signal θ(k). a (k), u b (k), u c (k) Perform coordinate transformation to obtain the d-axis and q-axis voltage signals u in the k-th sampling period. d (k), u q (k), the d-axis and q-axis voltage signals u in the kth sampling period d (k), u q (k) The inputs are weighted and processed by the hyperlocal model. The processing results are then compared with the d-axis and q-axis interference observations output by the neural network interference observer in the kth sampling period. Add them together to obtain the differential signals of the d-axis and q-axis currents in the k-th sampling period, respectively. and The first-order forward Euler discretization method was used to discretize the d-axis and q-axis current differential signals respectively. Combined with the current obtained by the current sensor sampling, the d-axis and q-axis discrete current prediction values ​​of the k-th sampling period output by the hyperlocal model were obtained respectively. Where k is a positive integer.

[0102] Specifically, the hyperlocal model in this step is detailed in the paper "Model-free control-based vector control of synchronous reluctance motor" published in the journal Dynamics and Control by Belkacem Selma and Elhadj Bounadja et al. in November 2023.

[0103] This step includes the following sub-steps:

[0104] (1-1) The three-phase voltage signal u of the synchronous reluctance motor in the kth sampling period is obtained by sampling with a voltage sensor. a (k), u b (k), u c (k), the position signal θ(k) of the synchronous reluctance motor in the kth sampling period is obtained through the position sensor, and the three-phase voltage signal u in the kth sampling period is calculated based on the position signal θ(k) of the kth sampling period. a (k), u b (k), u c (k) Perform coordinate transformation to obtain the d-axis and q-axis voltage signals u in the k-th sampling period. d (k), u q (k).

[0105] (1-2) The d-axis and q-axis voltage signals u obtained in step (1-1) for the kth sampling period d (k), u q (k) Use the corresponding scaling factor α d α q The values ​​are weighted separately and then compared with the interference observation values ​​output by the neural network interference observer in the kth sampling period. Add them together to obtain the d-axis and q-axis current differential signals of the k-th sampling period, respectively. and

[0106] Specifically, this step uses the following formula:

[0107]

[0108] In the formula, α d =1 / L d and α q =1 / L q α represents the weighting factor. d =1 / L d and α q =1 / L q These values ​​were all obtained through experience-based debugging.

[0109] like Figure 4 As shown, specifically, the d-axis and q-axis interference observations in the k-th sampling period It is obtained through the following sub-steps:

[0110] (a) Obtain the currents a0, a1, ..., a0 at different load startup times using a traditional model-predicted current control model. xAnd based on the currents a0, a1, ..., a x Obtain the actual load disturbance current error set F = [d1, d2, ..., d], which consists of all actual load disturbance current errors generated after the start-up of all loads in the model predictive current control model. x ], where x represents the total number of loads in the model predictive current control model.

[0111] Specifically, this step uses the following formula:

[0112]

[0113] Where a0 is the model predicted current control model current under no-load conditions, and d x For a x The corresponding load disturbance current error.

[0114] (b) Randomly select multiple load disturbance current errors from the load disturbance current error set F obtained in step (a) n times to form a data set (thus obtaining n data sets). For each data set p (where p∈[1,3]), obtain all load disturbance current errors in the p-th data set. The mean of the values ​​is used as the center of the Gaussian function corresponding to the p-th data set. Where n is a positive integer less than or equal to x, m p This represents the total number of load disturbance current errors in the p-th data set;

[0115] (c) For each data set p obtained in step (b), obtain the center of the Gaussian function corresponding to each load disturbance current error in the data set p. The Euclidean distance between the data points is calculated, and the sum of all the Euclidean distances is taken as the average to obtain the Gaussian function center b corresponding to the p-th data set. p And obtain the load disturbance current error set consisting of all load disturbance current errors in the p-th data set obtained in step (1-2). The load disturbance current error of each data set corresponds to the center of the Gaussian function of the p-th data set. The difference between them is calculated, and the sum of all the differences is taken as the average to obtain the center width b of the neural network. n .

[0116] (d) For each data set p obtained in step (b), the m load disturbance current errors in the p-th data set are... Input a Gaussian function for calculation and obtain multiple error outputs. Construct the initial error output matrix H0;

[0117] (e) Obtain the ideal weights based on the initial error output matrix H0 obtained in step (d) and the actual load disturbance current error set F obtained in step (a).

[0118] Specifically, this step uses the following formula:

[0119]

[0120] in, It is the transpose of the initial error output matrix H0.

[0121] (f) Based on the output matrix H0 obtained in step (d) and the ideal weights obtained in step (e) Estimate the interference observations for the first sampling period

[0122] Specifically, this step uses the following formula:

[0123]

[0124] Where ε is the estimation error of the neural network perturbation observer, and ε ≤ ε n , ε n This represents the estimation error of the neural network interference observer after n iterations.

[0125] (g) Estimation of the disturbance observations obtained in step (f) As initial values, the neural network interference observer is input for multiple iterations to obtain the interference observation value for the k-th sampling period.

[0126] (1-3) Based on the d-axis and q-axis currents of the k-th sampling period and the d-axis and q-axis voltage signals u obtained in step (1-1) of the k-th sampling period d (k), u q (k) and using the first-order forward Euler discretization method, the d-axis and q-axis current differential signals obtained in (1-2) for the kth sampling period are respectively processed. Discretization is performed to obtain the discrete current prediction values ​​of the d and q axes in the (k+1)th sampling period of the hyperlocal model output.

[0127] Specifically, this step uses the following formula:

[0128]

[0129] In the formula, T s This represents the system sampling time.

[0130] The advantage of this sub-step is that if the sampling time is very short, the hyperlocal model can be approximated as a real-time updated system model.

[0131] (2) Based on the predicted values ​​of the discrete currents i along the d and q axes of the hyperlocal model obtained in step (1) for the (k+1)th sampling period. d (k+1),i q (k+1) Obtain the first optimal voltage vector u in the kth sampling period. g1d (k), u g1q (k), based on the optimal voltage vector u g1d (k), u g1q (k) Calculate the voltage difference ΔE u1d (k), ΔE u1q (k) Obtain the second and third optimal voltage vectors u along the d and q axes by looking up the voltage difference in a table. g2d (k), u g2q (k) and u g2q (k), u g3q (k) uses the direct calculation method of duty cycle to obtain the d and q axis duty cycles d of the kth sampling period after the hyperlocal model optimization. nd (k), d nq (k), and based on the obtained duty cycle d nd (k) and d nq (k) The control pulse generator outputs a drive signal to the inverter so that the inverter outputs a voltage to the synchronous reluctance motor according to the drive signal, thereby driving the synchronous reluctance motor to run, where n∈[1,3].

[0132] The advantage of this step is that the inverter output voltage under the three-vector MPC strategy is closer to the ideal voltage vector, thus solving the problem of poor motor steady-state performance in existing MPC-based synchronous reluctance motor control methods.

[0133] This step includes the following sub-steps:

[0134] (2-1) Based on the predicted values ​​of the discrete currents along the d and q axes in the (k+1)th sampling period of the hyperlocal model output obtained in step (1) Obtain the ideal voltage vectors along the d and q axes in the kth sampling period.

[0135] Specifically, this step uses the following formula:

[0136]

[0137] in, Let be the d-axis and q-axis reference current for the k-th sampling period, and have e d (k), eq (k) represents the estimation error of the d-axis and q-axis currents in the k-th sampling period, with an initial value of 0; m is the error coefficient, with a value of 19000.

[0138] (2-2) Based on the ideal voltage vectors of the d and q axes obtained in step (2-1) for the kth sampling period And the position signal θ(k) of the kth sampling period, to obtain the voltage cost functions of the α and β axes.

[0139] Specifically, the obtained α and β axis voltage cost functions are expressed using the following formulas:

[0140]

[0141] Where u α (k), u β (k) is the effective voltage vector of the kth sampling period of the α and β axes.

[0142] (2-3) The eight basic voltage vectors obtained from the inverter's on / off state are used as the effective voltage vectors u of the α and β axes in the kth sampling period. α (k), u β (k) Substitute the α and β axis voltage cost functions obtained in step (2-2) into the calculations to obtain the values ​​that make g α and g β The minimum voltage vector is taken as the first optimal voltage vector u along the d and q axes in the k-th sampling period. g1d (k), u g1q (k), the ideal voltage vector of the kth sampling period is compared with the first optimal voltage vector u on the d and q axes respectively. g1d (k), u g1q (k) Calculate the difference to obtain the voltage difference ΔE between the d and q axes in the kth sampling period. u1d (k), ΔE u1q (k).

[0143] Specifically, the voltage difference between the α and β axes in the kth sampling period is calculated using the following formula:

[0144]

[0145] (2-4) Divide the voltage vector distribution diagram of the pulse generator into 6 sectors (e.g., Figure 2 As shown), based on the d-axis and q-axis voltage vector errors ΔE obtained in step (2-3) u1d (k), ΔE u1q (k) Find the corresponding sector in the voltage vector distribution diagram, and look up the corresponding second voltage vector u of the d and q axes in the pre-established second optimal voltage vector table for the kth sampling period based on the obtained sector. g2d(k), u g2q (k), and respectively query the corresponding d-axis and q-axis third voltage vector u of the k-th sampling period in the pre-established third optimal voltage vector table according to the obtained sector. g3d (k), u g3q (k).

[0146] The difference ΔE between the first optimal vector and the ideal voltage vector in the k-th sampling period can be used as a reference. u1 (k) The second and third voltage vectors are obtained by looking up the table in the sector where (k) is located.

[0147] Specifically, the selection of the second and third voltage vectors is shown in Table 1 below.

[0148] Table 1

[0149]

[0150] The advantage of the above sub-steps (2-3) to (2-5) is that, under the three-vector MPC strategy, by designing the voltage cost function and re-dividing the vector sectors, the selection process of the three-vector combination is optimized, and the complexity of the algorithm implementation is reduced.

[0151] (2-5) Based on the nth voltage vector u of the d and q axes obtained in step (2-4) for the kth sampling period respectively gnd (k) and u g x d (k) respectively obtains the predicted current values ​​i of the d and q axes for the kth sampling period corresponding to the nth voltage vector of the d and q axes. dn (k+1) and i qn (k+1).

[0152] Specifically, this step involves calculating using the following formula:

[0153]

[0154] (2-6) The d-axis and q-axis reference currents obtained in step (2-1) for the kth sampling period and The predicted current values ​​i on the d and q axes corresponding to the nth voltage vector on the d and q axes obtained in steps (2-5) during the kth sampling period are respectively compared with those on the d and q axes. dn (k+1) and i qn Divide by (k+1) to obtain the nth voltage vector u. gnd (k) and u gnd (k) represents the current error Δε along the d and q axes for the kth sampling period, respectively. dn (k) and Δε qn (k).

[0155] Specifically, this step involves calculation using the following formula:

[0156]

[0157] (2-7) Based on the nth voltage vector u obtained in step (2-6) gnd (k) and u gnd (k) represents the d-axis and q-axis current errors Δε corresponding to the kth sampling period. dn (k), Δε qn (k) Obtain the nth voltage vector u gnd (k) and u gnd (k) respectively correspond to the duty cycle d nd (k) and d nq (k).

[0158] Specifically, this step involves calculating using the following formula:

[0159]

[0160] Where d 1d (k) represents the d-axis u g1d (k) corresponds to the duty cycle, d 2d (k) represents u g2d (k) corresponds to the duty cycle, d 3d (k) represents u g3d (k) corresponds to the duty cycle, d 1q (k) represents the q-axis u g1q (k) corresponds to the duty cycle, d 2q (k) represents u g2q (k) corresponds to the duty cycle, d 3q (k) represents u g3q (k) corresponds to the duty cycle.

[0161] (2-8) Based on the duty cycle d obtained in step (2-7) nd (k) and d nq (k) The control pulse generator outputs a drive signal to the inverter, so that the inverter outputs a voltage to the synchronous reluctance motor according to the drive signal, thereby driving the synchronous reluctance motor to run.

[0162] (3) The d-axis and q-axis current signals i of the synchronous reluctance motor in the (k+1)th sampling period are obtained by sampling with a current sensor. d (k+1),i q (k+1), and compare them with the d-axis and q-axis current estimates calculated by the neural network interference observer obtained in step (1-3). The difference is calculated to obtain the d-axis and q-axis current estimation errors e in the (k+1)th sampling period, respectively. d(k+1) and e q (k+1) Based on the obtained d-axis and q-axis current estimation errors, obtain the d-axis and q-axis error output matrix for the (k+1)th sampling period. Estimation of the d-axis and q-axis weights for the (k+1)th sampling period The obtained error output matrix Estimation of weights Multiply to obtain an estimate of the interfering observations. The synchronous reluctance motor is controlled based on the estimated value of the disturbance observation.

[0163] The advantage of steps (1) to (3) is that the system disturbance is estimated and updated in real time through the current estimation error. This method does not require identification of the specific motor parameters of the controlled object, so it can solve the problem of reduced robustness caused by motor parameter mismatch in the existing MPC-based synchronous reluctance motor control method.

[0164] Specifically, this step includes the following sub-steps:

[0165] (3-1) Obtain the ABC three-phase current signal i in the (k+1)th sampling period after the synchronous reluctance motor starts running using a current sensor. a (k+1),i b (k+1),i c (k+1), and based on the position signal θ(k+1) of the synchronous reluctance motor in the (k+1)th sampling period obtained from the position sensor, obtain the d-axis and q-axis current signals i in the (k+1)th sampling period. d (k+1) and i q (k+1).

[0166] Specifically, this step uses the following formula:

[0167]

[0168] (3-2) Based on the ideal voltage vectors of the d and q axes obtained in step (2-1) for the kth sampling period and Obtain the d-axis and q-axis current estimates for the k-th sampling period respectively. and

[0169] Specifically, the d-axis and q-axis current estimates in this step are obtained using the following formulas:

[0170]

[0171] (3-3) Obtain the d-axis and q-axis currents i in the (k+1)th sampling period from step (3-1). d (k+1) and iq (k+1) are compared with the d-axis and q-axis current estimates obtained in step (3-2) for the k-th sampling period. and The difference is calculated to obtain the d-axis and q-axis current estimation errors e in the (k+1)th sampling period, respectively. d (k+1) and e q (k+1).

[0172] Specifically, this step involves calculating the current estimation error using the following formula:

[0173]

[0174] (3-4) Determine the estimation errors e of the d-axis and q-axis currents obtained in step (3-3) for the (k+1)th sampling period. d (k+1),e q If all (k+1) are 0, it means that the synchronous reluctance motor is running smoothly and the process ends; otherwise, proceed to step (3-5).

[0175] (3-5) Estimating the d-axis and q-axis currents based on the values ​​obtained in step (3-3) for the (k+1)th sampling period, respectively. d (k+1) and e q (k+1), and use the Gaussian function to obtain the error output h corresponding to the d and q axes of the (k+1)th sampling period respectively. d (k+1) and h q (k+1), and output h using the error corresponding to the d and q axes. d (k+1) and h q (k+1) Obtain the error output matrix for the (k+1)th sampling period.

[0176] Specifically, this step uses the following formula:

[0177]

[0178] in, The centers of the Gaussian function on the d and q axes are respectively; b d b q These represent the center widths of the neural network along the d and q axes, respectively.

[0179] (3-6) Based on the error output matrix H(k+1) of the (k+1)th sampling period obtained in step (3-5), obtain the weight estimate for the (k+1)th sampling period.

[0180] Specifically, this step uses the following formula:

[0181]

[0182] In the formula, It is the derivative of the weight estimate for the (k+1)th sampling period; It is the derivative of the error in estimating the weights for the (k+1)th sampling period, and satisfies W * (k) is the ideal value of the weight in the k-th sampling period. It is an estimate of the weights for the (k+1)th sampling period; e T (k+1) is the transpose of e(k+1). γ is a positive constant.

[0183] (3-7) Combine the error output matrix H(k+1) of the (k+1)th sampling period obtained in step (3-5) with the weight estimate of the (k+1)th sampling period obtained in step (3-6). Multiply to obtain an estimate of the interference observations in the (k+1)th sampling period. Then return to step (3-1).

[0184] In summary, this invention combines model-free predictive current control with a neural network disturbance observer, introducing a hyperlocal model into the predictive control of a synchronous reluctance motor and improving it into model-free predictive control. The unknown parts in the hyperlocal model are observed through the neural network disturbance observer. By improving the cost function structure and the duty cycle calculation method based on current vector error, the optimal three-vector combination and the corresponding duty cycle are obtained, enabling the synchronous reluctance motor to maintain its stable output under complex operating conditions, and significantly improving the robustness of the system.

[0185] This invention addresses the problems of low type, poor convergence, and low estimation accuracy when using extended observers to estimate unknown parts in hyperlocal models. This invention employs a neural network interference observer for estimation. The neural network interference observer utilizes the concepts of pure integration, compensation, and transfer function type estimation, modifying the structure of the extended observer. This results in the neural network interference observer having two higher type levels than the extended observer, leading to higher accuracy and stronger convergence.

[0186] This invention addresses the issue that the steady-state performance of the motor is unsatisfactory when a single voltage vector is applied within the control cycle in classic MPC (Multi-Process Control) systems. To address this, it proposes using three voltage vectors within the control cycle to improve steady-state performance. By designing a voltage cost function and re-dividing the vector sectors, the selection process for the three-vector combination is optimized, improving the system's steady-state performance and reducing the algorithm's implementation complexity. Furthermore, it proposes using the duty cycle calculation based on the current vector error, making the duty cycle calculation unaffected by perturbations in the motor's intrinsic parameters, thus improving system robustness.

[0187] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer, characterized in that, Includes the following steps: (1) Sampling and acquisition of the first synchronous reluctance motor Each sampling period Three-phase voltage signal The position sensor obtains the position of the synchronous reluctance motor. Position signal for each sampling period According to the position signal For three-phase voltage signals , , Perform coordinate transformations to obtain the first... Each sampling period shaft voltage signal , , will the Each sampling period shaft voltage signal , The inputs to the hyperlocal model are weighted and processed separately. The processed results are then compared with the output of the neural network interference observer. Output per sampling period Axis interference observations , Add them together to obtain the first one. Each sampling period Axis current differential signal and And the first-order forward Euler discretization method was used to respectively... The differential signal of the axis current is discretized and combined with the current sampled by the current sensor to obtain the first output of the hyperlocal model. Each sampling period Shaft Discrete Current Prediction , ,in, It is a positive integer; (2) Based on the hyperlocal model obtained in step (1), the first Predicted discrete current values ​​of the d and q axes for each sampling period , Get the The first optimal voltage vector for each sampling period , According to this optimal voltage vector , Calculate the voltage difference , The second and third optimal voltage vectors along the d and q axes are obtained by looking up the table based on the voltage difference. , and , Using the duty cycle direct calculation method, the optimized hyperlocal model is obtained. Each sampling period Axis duty cycle , And based on the obtained duty cycle and The control pulse generator outputs a drive signal to the inverter, which then outputs a voltage to the synchronous reluctance motor based on the drive signal, thereby driving the synchronous reluctance motor to run. ; (3) The synchronous reluctance motor is sampled by a current sensor at the first... Each sampling period shaft current signal According to this shaft current signal The first one was obtained respectively Each sampling period Shaft current estimation error and According to the obtained Obtaining the first shaft current estimation error Each sampling period Axis error output matrix and the Each sampling period Axis weight estimation The resulting error output matrix Estimation of weights Multiply to obtain an estimate of the interfering observations. And control the synchronous reluctance motor based on the estimate of the disturbance observation; step (3) includes the following sub-steps: (3-1) The current sensor is used to obtain the number of seconds after the synchronous reluctance motor starts running. Each sampling period Three-phase current signal And based on the position sensor, the synchronous reluctance motor's first... Position signal for each sampling period , obtain the Each sampling period shaft current signal and ; (3-2) The output of the hyperlocal model obtained in step (1) Each sampling period Shaft Discrete Current Prediction , , obtain the Each sampling period , Axial ideal voltage vector , According to the obtained first Each sampling period Axial ideal voltage vector and Obtain the first Each sampling period , Axis current estimation and ; (3-3) Obtain the sampling from step (3-1) Each sampling period , shaft current and The results obtained in step (3-2) are respectively compared with the first Each sampling period , Axis current estimation and Divide the values ​​to obtain the first value respectively. Each sampling period , Shaft current estimation error and ; (3-4) Determine the first step obtained in step (3-3). Each sampling period , Shaft current estimation error , If all values ​​are 0, it indicates that the synchronous reluctance motor is running smoothly and the process ends; otherwise, proceed to step (3-5). (3-5) Based on the results obtained in step (3-3), the first... Each sampling period , Shaft current estimation error and And use the Gaussian function to obtain the first... Each sampling period Error output corresponding to the axis and and use Error output corresponding to the axis and Get the Error output matrix for each sampling period Steps (3-5) involve using the following formula: ； in, , They are respectively Center of the Gaussian function on the axis; , They are respectively The center width of the axial neural network; (3-6) The first one obtained from step (3-5) Error output matrix for each sampling period Get the Estimation of weights for each sampling period Steps (3-5) involve using the following formula: ； in, , They are respectively Center of the Gaussian function on the axis; , They are respectively The center width of the axial neural network; Steps (3-6) involve using the following formula: ； in It is the first The derivative of the weight estimate for each sampling period; It is the first The derivative of the error in estimating the weights for each sampling period, and satisfying ; It is the first The ideal value of the weight for each sampling period. It is the first Estimation of the weights for each sampling period; yes transpose, , It is a positive number; (3-7) The first one obtained in step (3-5) Error output matrix for each sampling period The first obtained in step (3-6) Estimation of weights for each sampling period Multiply to obtain the first... Estimation of interference observations for each sampling period Then return to step (3-1).

2. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer according to claim 1, characterized in that, Step (1) includes the following sub-steps: (1-1) The voltage sensor samples the first voltage of the synchronous reluctance motor. Each sampling period Three-phase voltage signal The position sensor obtains the position of the synchronous reluctance motor. Position signal for each sampling period According to the Position signal for each sampling period For the first Three-phase voltage signal per sampling period Perform coordinate transformations to obtain the first... d-axis and q-axis voltage signals for each sampling period ; (1-2) The first step obtained in step (1-1) d-axis and q-axis voltage signals for each sampling period Use the corresponding scaling factor , The values ​​are weighted separately and then compared with the output of the neural network interference observer. Interference observations for each sampling period Add them together to obtain the first one. Differential signals of d-axis and q-axis currents for each sampling period and ; (1-3) According to the first Each sampling period The shaft current and the first obtained in step (1-1) Each sampling period shaft voltage signal And the first-order forward Euler discretization method is used to discretize the first-order Euler discretization method obtained by (1-2) respectively. Each sampling period Axis current differential signal , Discretization is performed to obtain the first output of the hyperlocal model. Predicted discrete current values ​​of the d and q axes for each sampling period , .

3. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer according to claim 1 or 2, characterized in that, Step (1-2) uses the following formula: ； in and Indicates the weighting factor. and These values ​​were all obtained through experience-based debugging. Steps (1-3) use the following formula: ； in This represents the system sampling time.

4. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer according to claim 3, characterized in that, No. d-axis and q-axis interference observations for each sampling period It is obtained through the following sub-steps: (a) Obtaining the current of a traditional model predictive current control model at different load startup times. And according to the current Obtain the set of actual load disturbance current errors, which consists of all actual load disturbance current errors generated after the start-up of all loads in the model predictive current control model. , where x represents the total number of loads in the model predictive current control model; (b) in sequence The set of load disturbance current errors obtained from step (a) Randomly select multiple load disturbance current errors to form a data set, and for each data set In other words, to obtain the first All load disturbance current errors in the data set The mean of the first value is used as the average of the second value. The center of the Gaussian function corresponding to each data set ,in Is it less than or equal to? positive integers, Indicates the first The total number of load disturbance current errors in each dataset. ; (c) For each dataset obtained in step (b) In other words, to obtain this dataset Each load disturbance current error is related to the first... The center of the Gaussian function corresponding to each data set The Euclidean distance between them is calculated, and the sum of all the obtained Euclidean distances is taken as the average to obtain the first Euclidean distance. The center of the Gaussian function corresponding to each data set and obtain the first The load disturbance current error set consisting of all load disturbance current errors in a data set. Each load disturbance current error is related to the first... The center of the Gaussian function corresponding to each data set The difference between them is calculated, and the sum of all the differences is taken as the average to obtain the center width of the neural network. ; (d) For each dataset obtained in step (b) In other words, the first In a data set Load disturbance current error Input a Gaussian function for calculation and obtain multiple error outputs. Construct the initial error output matrix ; (e) Based on the initial error output matrix obtained in step (d) and the actual load disturbance current error set obtained in step (a) Obtaining ideal weights ; (f) The output matrix obtained from step (d) and the ideal weights obtained in step (e) Estimate the interference observations for the first sampling period ; (g) Estimation of disturbance observations obtained in step (f) As initial values, the neural network interference observer is input for multiple iterations to obtain the first value. Interference observations for each sampling period .

5. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer according to claim 4, characterized in that, Step (a) is to use the following formula: ； in, To predict the current of the model and control the current of the model under no-load conditions, for The corresponding load disturbance current error; Step (e) uses the following formula: ； in, It is the initial error output matrix transpose; Step (f) is to use the following formula: ； Where ε is the estimation error of the neural network interference observer, and satisfies , This represents the estimation error of the neural network interference observer after n iterations.

6. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer according to claim 5, characterized in that, Step (2) includes the following sub-steps: (2-1) The output of the hyperlocal model obtained in step (1) Each sampling period Shaft Discrete Current Prediction , , obtain the Each sampling period , Axial ideal voltage vector , ; (2-2) The first one obtained from step (2-1) Each sampling period Axial ideal voltage vector , and the Position signal for each sampling period , obtain Voltage cost function of the axis and ; (2-3) The eight basic voltage vectors obtained from the inverter's on / off states are used as the first... Each sampling period Effective voltage vector of the axis , Substituting the result obtained in step (2-2) The axis voltage cost function is calculated to obtain the values ​​that make the axis voltage cost function. and The minimum voltage vector is taken as the first Each sampling period First optimal voltage vector of axis , , will the The ideal voltage vector for each sampling period is respectively related to First optimal voltage vector of axis , Divide the values ​​to obtain the first value respectively. Each sampling period shaft voltage difference , ; (2-4) Divide the voltage vector distribution diagram of the pulse generator into 6 sectors, based on the data obtained in step (2-3). Axis voltage vector error , In the voltage vector distribution map, the corresponding sector is selected, and the corresponding sector is queried in the pre-established second optimal voltage vector table. Each sampling period Second voltage vector of axis , And respectively, based on the obtained sector, look up the corresponding first optimal voltage vector in the pre-established third optimal voltage vector table. Each sampling period Third-axis voltage vector , ; (2-5) Based on the results obtained in step (2-4), the first... Each sampling period Axis No. Voltage vector and Obtain the respective Axis No. The voltage vector corresponding to the first Each sampling period Shaft current prediction and ; (2-6) The first step obtained in step (2-1) Each sampling period Shaft reference current and Compared with the results obtained in steps (2-5) Axis No. The voltage vector corresponding to the first Each sampling period Shaft current prediction and Divide the difference to obtain the first one. Voltage vector and The corresponding first Each sampling period Shaft current error and ; (2-7) Based on the nth voltage vector obtained in step (2-6) and The corresponding first Each sampling period Shaft current error , Get the first Voltage vector and The corresponding duty cycles and ; (2-8) Duty cycle obtained from step (2-7) and The control pulse generator outputs a drive signal to the inverter, so that the inverter outputs a voltage to the synchronous reluctance motor according to the drive signal, thereby driving the synchronous reluctance motor to run.

7. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network interference observer according to claim 6, characterized in that, Step (2-1) uses the following formula: ； in, , For the first Each sampling period Shaft reference current, and has , ; , For the first Each sampling period The estimation error of the shaft current is initially set to 0. This is the error coefficient, and its value is 19000; Step (2-2) uses the following formula: ； in , for Axis No. The effective voltage vector for each sampling period; In step (2-3) Each sampling period The shaft voltage difference is calculated using the following formula: ； Steps (2-5) involve calculating using the following formula: ； Steps (2-6) are calculated using the following formula: ； Steps (2-7) involve calculating using the following formula: ； in express axis The corresponding duty cycle, express The corresponding duty cycle, express The corresponding duty cycle, express axis The corresponding duty cycle, express The corresponding duty cycle, express The corresponding duty cycle.

8. The method for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer according to claim 7, characterized in that, Step (3-1) uses the following formula: ； The d-axis and q-axis current estimates in step (3-2) are obtained using the following formulas: ； Step (3-3) involves calculating the current estimation error using the following formula: 。 9. A system for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer, which is implemented based on the method for model-free predictive current control of a synchronous reluctance motor based on a neural network disturbance observer as described in claim 1, characterized in that... The system includes: The first module is used to sample and obtain the voltage of the synchronous reluctance motor via a voltage sensor. Each sampling period Three-phase voltage signal The position sensor obtains the position of the synchronous reluctance motor. Position signal for each sampling period According to the position signal For three-phase voltage signals Perform coordinate transformations to obtain the first... Each sampling period shaft voltage signal , , will the Each sampling period shaft voltage signal , The inputs to the hyperlocal model are weighted and processed separately. The processed results are then compared with the output of the neural network interference observer. Output per sampling period Axis interference observations , Add them together to obtain the first one. Each sampling period Axis current differential signal and And the first-order forward Euler discretization method was used to respectively... The differential signal of the axis current is discretized and combined with the current sampled by the current sensor to obtain the first output of the hyperlocal model. Each sampling period Shaft Discrete Current Prediction , ,in, It is a positive integer; The second module is used to determine the hyperlocal model obtained from the first module. Predicted discrete current values ​​of the d and q axes for each sampling period , Get the The first optimal voltage vector for each sampling period , According to this optimal voltage vector , Calculate the voltage difference , The second and third optimal voltage vectors along the d and q axes are obtained by looking up the table based on the voltage difference. , and , Using the duty cycle direct calculation method, the optimized hyperlocal model is obtained. Each sampling period Axis duty cycle , And based on the obtained duty cycle and The control pulse generator outputs a drive signal to the inverter, which then outputs a voltage to the synchronous reluctance motor based on the drive signal, thereby driving the synchronous reluctance motor to run. ; The third module is used to sample and obtain the synchronous reluctance motor data in the first stage via a current sensor. Each sampling period shaft current signal According to this shaft current signal The first one was obtained respectively Each sampling period Shaft current estimation error and According to the obtained Obtaining the first shaft current estimation error Each sampling period Axis error output matrix and the Each sampling period Axis weight estimation The resulting error output matrix Estimation of weights Multiply to obtain an estimate of the interfering observations. The synchronous reluctance motor is controlled based on the estimated value of the disturbance observation.

Images