Improved observer-free deadbeat control method

Abstract

The invention discloses an improved observer-free and dead-beat control method, which comprises the following steps of: constructing a super local model, fully utilizing a known counter electromotive force term, and adopting an inductance extraction algorithm which only needs information of a previous period; and the inductance is accurately extracted by converging d-axis disturbance to zero. According to the method, complete model-free control is realized, and the method is completely independent of any motor parameters, so that the robustness of predictive current control is improved.

Description

Technical Field

[0001] The present invention relates to the field of observer-free technology, and in particular to an improved observer-free deadbeat control method. Background Art

[0002] The establishment process of the traditional model-free and observer-free (MF-OF) method is as follows:

[0003] The mathematical model of SPMSM (surface mounted permanent magnet synchronous motor) based on super local model is described as follows:

[0004]

[0005] i s Represents the stator current, α is the input coefficient (the inverse of the inductance, i.e. α = 1 / L c ),u s is the input voltage, F is the lumped disturbance, and t is the time;

[0006] Using Euler discretization (1):

[0007]

[0008] Where T is the control period and k is the kth time. Since the control period is very short, it can be assumed that α and F remain constant over several consecutive periods. Therefore, α can be expressed as α k-1 , α k-2 , α k+1 , etc. to replace it, and the same is true for F.

[0009] The model functions in several adjacent periods are listed below:

[0010]

[0011] By subtracting (3) from (4), we can get the expression for α as follows (F k-1 =F k-2 ):

[0012]

[0013] Substituting (7) into (4), we can get the expression of the perturbation F as follows:

[0014]

[0015] Taking into account digital delays, the reference voltage calculated from the algorithm is applied to the motor at time k, and the actual execution is not until time k+1. Therefore, it is necessary to look ahead two steps and calculate u s k+1 As the reference voltage at time k, u sk+1 It will be actually implemented at time k+1.

[0016] By adding (5) and (6), we can get the following relationship (F = F k =F k+1 ):

[0017]

[0018] In order to track the current during the two control cycles, i s k+2 Set to i s ref . In this way, α and F can be estimated without an observer. However, due to measurement noise and iteration reasons, high-frequency harmonics may appear in α and F. Therefore, a filter is needed to filter out the high-frequency components to obtain accurate estimates. α is actually the inverse of the inductance term, which changes very little when the motor is running, while the disturbance term F may change greatly. The control block diagram of the traditional model-free and observer-free method based on the above model is as follows Figure 1 shown.

[0019] Therefore, how to provide an improved observer-free deadbeat control method has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0020] In view of this, an object of the present invention is to provide an improved observer-free deadbeat control method, which realizes complete model-free control and is completely independent of any motor parameters, thereby improving the robustness of predictive current control.

[0021] The present invention solves the technical problem by adopting the following technical solution:

[0022] An improved observer-free deadbeat control method includes: constructing a super-local model, making full use of known back electromotive force terms, and adopting an inductance extraction algorithm that only requires information from the last cycle; and accurately extracting the inductance by making the d-axis disturbance converge to zero.

[0023] Furthermore, the hyperlocal model is as follows:

[0024]

[0025] Easy to get

[0026]

[0027] Among them, L m is the real inductance of SPMSM, is the real magnetic flux of the rotor, u d is the d-axis voltage, i d is the q-axis current, uq is the q-axis voltage, L c is the controller inductance, R s To define resistance, F d is the d-axis disturbance, F q is the q-axis disturbance, ω ε is the electrical angular velocity;

[0028] In order to implement the algorithm in DSP, the function should be converted into discrete form, and the equation expression at adjacent moments around time k is as follows:

[0029]

[0030] Furthermore, when the MTPA algorithm is implemented on the SPMSM, the d-axis current is set to zero, so formula (14) is simplified to:

[0031]

[0032] This means that if L m and L c There is an error between d will not be 0; that is, if F d Converges to 0, then the L used in the controller c The term will converge to L m ; Extract the true value of the inductor without using observers and other complex parameter identification methods; In formula (16), replace F d k-1 Set to zero; the following is the inductance extraction algorithm:

[0033]

[0034] Assume that L c and F d It remains unchanged in several control cycles. According to formula (16), we get F d The expression is as follows:

[0035]

[0036] Furthermore, by calculating the reference voltage vector and delay compensation, the d-axis disturbance converges to zero and the inductance is accurately extracted:

[0037] The voltage equations for the two cycles are listed as follows:

[0038]

[0039]

[0040] Assumption F dq =F dqk =F dq k+1 , where F dq is the dq axis disturbance, F dq k is the dq axis disturbance at time k, F dq k+1 is the dq axis disturbance at time k+1; considering the digital delay, according to formulas (24) and (26), the current at the next moment is predicted to be:

[0041]

[0042] According to formulas (25) and (27), and i dq k+2 Replace with i dq ref , we get the expression of dq axis reference voltage:

[0043]

[0044] Among them, i dq k+2 is the dq axis current at time k+2, i dq ref is the dq axis reference current, u d k+1 is the d-axis voltage at time k+1, u q ref is the q-axis reference voltage, and the dq-axis voltage u at time k+1 dq k+1 is the dq axis reference voltage u that should be loaded on the motor at time k dq ref , due to the digital delay, it will actually be loaded at time k+1 to compensate.

[0045] The improved observer-free deadbeat control method disclosed in the present invention has the following beneficial effects:

[0046] 1. The present invention realizes complete model-free control and is completely independent of any motor parameters, thereby improving the robustness of predictive current control.

[0047] 2. The present invention has innovative observer-free control. Unlike the traditional observer-free method that requires information from two cycles to extract inductance, the proposed method constructs a hyperlocal model, fully utilizes the known back electromotive force (EMF) term, and adopts an inductance extraction algorithm that only requires information from the previous cycle. By making the d-axis disturbance converge to zero, the inductance is accurately extracted. Therefore, the reduction and full utilization of periodic information in observer control effectively reduces the computational burden, and can shorten the adjustment time of the q-axis current by nearly 50%, while improving the total harmonic distortion (THD), which is a significant improvement over the traditional MF-OF method. At the same time, in steady-state and transient experiments, the proposed MF-OF method and the MF-ESO method achieved similar performance in steady-state and dynamic performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is the control block diagram of the traditional model-free and observer-free method;

[0049] Figure 2 It is a control block diagram of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In order to reduce the dependence of predictive current control (DBPCC) on motor parameters, model-free methods based on hyperlocal models have attracted widespread attention. Different from the complex strategies of estimating disturbances and identifying parameters using observers or data-driven methods in model-free control, an improved deadbeat control method without observer (MF-OF) is proposed to simplify the algorithm and improve dynamic performance. Typical observer-free methods use information from two cycles to extract inductance, resulting in slower dynamic response. To address this problem, the proposed MF-OF method includes a new hyperlocal model that makes full use of the known back electromotive force (EMF) term to alleviate the pressure of disturbance estimation. At the same time, the inductance extraction algorithm in the proposed method only requires information from the previous cycle, which can achieve high-precision inductance estimation by making the d-axis disturbance converge to zero, reduce the computational burden, and improve the response speed. Finally, a comparative study is carried out on a 1kW surface-mount permanent magnet synchronous motor (SPMSM) experimental platform. Experimental results demonstrate that the proposed MF-OF method shortens the dynamic response time by nearly 50% compared with the traditional MF-OF method and successfully achieves a dynamic performance similar to that of the model-free extended state observer (MF-ESO) method.

[0052] Like MF-ESO and conventional MF-OF, the proposed observer-free method is a completely model-free method and does not use any motor parameters such as inductance terms. It can adapt the inductance term while still maintaining excellent tracking performance.

[0053] Unlike the conventional MF-OF, which builds an overly generalized hyperlocal model and requires information from the past two cycles to estimate the inductance, the proposed MF-OF adds the back electromotive force (EMF) term to the hyperlocal model and adopts a new inductance extraction algorithm that only requires information from the past cycle, thereby achieving a more accurate model and, in turn, better dynamic performance. More accurate inductance estimation means fewer current harmonics, which can be verified experimentally. Moreover, for SPMSM (surface mounted permanent magnet synchronous motor), the back EMF term does not include the inductance (L d =L q ), so the proposed MF-OF can add the back-EMF term without introducing additional parameters, making full use of the known information, making it have faster dynamic performance than the traditional MF-OF because it greatly reduces the computational pressure of disturbance estimation, especially at high-speed operation.

[0054] Unlike MF-ESO, the proposed MF-OF does not contain ESO, thus avoiding complex parameter adjustment and model construction. At the same time, it can also achieve similar motor performance. Therefore, the proposed MF-OF combines the advantages of MF-ESO and traditional MF-OF, and can achieve fast dynamic performance without ESO.

[0055] refer to Figure 2 The present invention discloses an improved observer-free deadbeat control method, comprising: constructing a super-local model, making full use of the known back electromotive force term, and adopting an inductance extraction algorithm that only requires information from the previous cycle; and accurately extracting the inductance by making the d-axis disturbance converge to zero.

[0056] Based on the traditional model-free and observer-free (MF-OF) method, an improved MF-OF method is proposed. The traditional observer-free method requires the information of the previous two cycles to extract the inductance, so the dynamic performance is poor. Compared with the traditional method, the proposed MF-OF method has excellent dynamic performance, which mainly comes from the following two improvements. First, the hyperlocal model includes the back electromotive force (EMF) term to fully utilize the known terms. At the same time, the inductance extraction algorithm is improved by converging the d-axis perturbation to zero based on the new hyperlocal model, using only the information of the previous cycle.

[0057] In order to illustrate the proposed method, the mathematical model of SPMSM is first established. For the traditional deadbeat predictive control, the dq axis voltage equation in the synchronous rotating coordinate system can be expressed as:

[0058]

[0059] L m is the real inductor of SPMSM, is the real magnetic flux of the rotor.

[0060] In order to accurately estimate the inductance term and the disturbance term, a new hyperlocal model is adopted. The hyperlocal model is shown as follows:

[0061]

[0062] Easy to get

[0063]

[0064] Among them, L m is the real inductance of SPMSM, is the real magnetic flux of the rotor, u d is the d-axis voltage, i d is the q-axis current, u q is the q-axis voltage, L c is the controller inductance, R s To define resistance, F d is the d-axis disturbance, F q is the q-axis disturbance, ω ε is the electrical angular velocity;

[0065] In order to implement the algorithm in DSP, the function should be converted into discrete form, and the equation expression at adjacent moments around time k is as follows:

[0066]

[0067] Inductance and disturbance estimation:

[0068] In formula (14), when the MTPA algorithm is implemented on the SPMSM, the d-axis current is set to zero, so formula (14) is simplified to:

[0069]

[0070] This means that if L m and L c There is an error between d will not be 0; that is, if F d Converges to 0, then the L used in the controller c The term will converge to L m ; Therefore, the true value of the inductance can be extracted without using observers and other complex parameter identification methods; For convenience, in formula (16), F dk-1 Set to zero; the following is the inductance extraction algorithm:

[0071]

[0072] Same as the traditional MF-OF method, assuming that L c and F d remains unchanged in several control cycles, then, according to formula (16), it is easy to get F d The expression is as follows:

[0073]

[0074] And Lc has been obtained in formula (21). Similarly,

[0075]

[0076] By calculating the reference voltage vector and delay compensation, the d-axis disturbance converges to zero and the inductance is accurately extracted:

[0077] The voltage equations for the two cycles are listed as follows:

[0078]

[0079] Assumption F dq =F dq k =F dq k+1 , where F dq is the dq axis disturbance, F dq k is the dq axis disturbance at time k, F dq k+1 is the dq axis disturbance at time k+1; considering the digital delay, according to formulas (24) and (26), the current at the next moment can be predicted as:

[0080]

[0081] According to formulas (25) and (27), and i dq k+2 Replace with i dq ref , we can get the expression of dq axis reference voltage:

[0082]

[0083] Among them, i dq k+2 is the dq axis current at time k+2, i dq ref is the dq axis reference current, u dk+1 is the d-axis voltage at time k+1, u q ref is the q-axis reference voltage, and the dq-axis voltage u at time k+1 dq k+1 is the dq axis reference voltage u that should be loaded on the motor at time k dq ref , due to the digital delay it will actually be loaded at time k+1, so this has been compensated.

[0084] The control block diagram of the improved MF-OF method is shown in Figure 2 As shown. Figure 1 Compared with the traditional MF-OF method in , it not only differs in the inductance extraction and disturbance estimation formulas, but also adds a current prediction part due to the addition of the ωe term.

[0085] It is important to note that although equations (17) and (19) still appear in the paper, just like equation (3), they are not actually used. In other words, the traditional MF-OF method uses the information of the previous two cycles to estimate the inductance, while the proposed MF-OF method only relies on the information of the previous cycle, thus achieving a faster response time.

[0086] The present invention realizes complete model-free control, which is completely independent of any motor parameters, thereby improving the robustness of predictive current control. The present invention has innovative observer-free control. Unlike the traditional observer-free method that requires two cycles of information to extract inductance, the proposed method constructs a super-local model, fully utilizes the known back electromotive force (EMF) term, and adopts an inductance extraction algorithm that only requires information from the previous cycle. By making the d-axis disturbance converge to zero, the inductance is accurately extracted. Therefore, the reduction and full utilization of periodic information in observer control effectively reduces the computational burden, and can shorten the adjustment time of the q-axis current by nearly 50%, while improving the total harmonic distortion (THD), which is significantly improved compared to the traditional MF-OF method. At the same time, in steady-state and transient experiments, the proposed MF-OF method and the MF-ESO method achieved similar performance in steady-state and dynamic performance.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An improved observer-free deadbeat control method, characterized in that: include: A hyperlocal model is constructed to fully exploit the known back-EMF term and adopt an inductance extraction algorithm that only requires information from the last cycle; the inductance is accurately extracted by converging the d-axis perturbation to zero.

2. The improved observer-free deadbeat control method according to claim 1, characterized in that: The hyperlocal model looks like this: Easy to get Among them, L m is the real inductance of SPMSM, is the real magnetic flux of the rotor, u d is the d-axis voltage, i d is the q-axis current, u q is the q-axis voltage, L c is the controller inductance, R s To define resistance, F d is the d-axis disturbance, F q is the q-axis disturbance, ω ε is the electrical angular velocity; In order to implement the algorithm in DSP, the function should be converted into discrete form, and the equation expression at adjacent moments around time k is as follows:

3. The improved observer-free deadbeat control method according to claim 2, characterized in that: When implementing the MTPA algorithm on the SPMSM, the d-axis current is set to zero, so equation (14) is simplified to: This means that if L m and L c There is an error between d will not be 0; that is, if F d Converges to 0, then the L used in the controller c The term will converge to L m ; Extract the true value of the inductor without using observers and other complex parameter identification methods; In formula (16), replace F d k-1 Set to zero; the following is the inductance extraction algorithm: Assume that L c and F d It remains unchanged in several control cycles. According to formula (16), we get F d The expression is as follows:

4. The improved observer-free deadbeat control method according to claim 3, characterized in that: By calculating the reference voltage vector and delay compensation, the d-axis disturbance converges to zero and the inductance is accurately extracted: The voltage equations for the two cycles are listed as follows: Assumption F dq =F dq k =F dq k+1 , where F dq is the dq axis disturbance, F dq k is the dq axis disturbance at time k, F dq k+1 is the dq axis disturbance at time k+1; considering the digital delay, according to formulas (24) and (26), the current at the next moment is predicted to be: According to formulas (25) and (27), and i dq k+2 Replace with i dq ref , we get the expression of dq axis reference voltage: Among them, i dq k+2 is the dq axis current at time k+2, i dq ref is the dq axis reference current, u d k+1 is the d-axis voltage at time k+1, u q ref is the q-axis reference voltage, and the dq-axis voltage u at time k+1 dq k+1 is the dq axis reference voltage u that should be loaded on the motor at time k dq ref , due to the digital delay, it will actually be loaded at time k+1 to compensate.

Images