Sector prediction optimization control method based on switched reluctance motor

Abstract

The invention discloses a sector prediction optimization control method based on a switched reluctance motor. Aiming at the problems of more switch states and larger calculation burden in divided sectors in the traditional finite set model predictive torque control, the method comprises the following steps of: firstly, re-dividing a commutation interval of a traditional sector according to the tracking condition of each phase actual torque to a reference torque in the commutation interval; thirdly, combining the influence of the switching state in the whole phase conduction domain on the torque tracking effect, and reformulating the candidate switching state of the improved sector; through a series of optimization measures, on the premise of ensuring the operation stability of the system, the number of candidate switch states in the sector is greatly reduced, and the calculation pressure of the controller is effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the field of motor control and proposes a sector prediction optimization control method based on a switched reluctance motor to reduce candidate switching states of model predictive torque control. Background Art

[0002] Switched reluctance motors (SRMs) are gaining increasing attention in industrial applications due to their simple structure, lack of permanent magnets, high starting torque, low starting current, high reliability, low manufacturing cost, and strong fault tolerance. However, their highly nonlinear electromagnetic characteristics and doubly salient pole structure result in high torque ripple and noise, which limits their application in high-performance applications such as automobiles. To better adapt SRMs to high-performance applications, it is crucial not only to suppress high torque ripple but also to consider the motor's speed regulation performance. In recent years, research on reducing torque ripple and improving the motor's dynamic performance has received increasing attention.

[0003] In the field of motor control, the use of advanced control strategies to suppress torque ripple and improve dynamic performance has become a mainstream research direction. Traditional methods are mainly based on direct torque control (DTC) and indirect torque control (ITC) architectures. With technological advancements, the introduction of the torque split function (TSF) has significantly improved the problem of motor torque ripple. Current research trends show that modern control theory has successfully integrated intelligent algorithms such as adaptive control, sliding mode variable structure control, BP neural networks, iterative learning control, and model predictive control (MPC) into torque control systems. Of particular note is model predictive control technology, which is becoming a research hotspot in the field of motor control due to its advantages such as simple algorithm structure and strong adaptability to nonlinear and multivariable systems.

[0004] Currently, there is a significant body of research on applying MPC to the inner torque loop, leveraging its ability to process nonlinear models to suppress SRM torque pulsation. However, traditional Finite Control Set Model Predictive Torque Control (FCS-MPTC) suffers from the high number of switching states within the sector, resulting in a heavy computational burden. This significant computational burden severely impacts the controller's runtime, leading to high switching frequency and a significant decrease in control effectiveness. This high computational burden has always been a significant challenge for model-based predictive torque control.

[0005] In this paper, a model predictive torque control sector optimization method based on the switched reluctance motor is proposed to address the problem that the traditional FCS-MPTC has a large number of switching states in the divided sectors and a heavy computational burden. This method improves the traditional sector division and re-formulates the switching states in the sector. While ensuring system stability, it effectively reduces the number of candidate switching states in the sector and reduces the computational burden of the controller. Summary of the Invention

[0006] To address the computational burden of numerous switching states within the traditional FCS-MPTC sector, this paper proposes a sector optimization method for model-predictive torque control based on switched reluctance motors. This method first improves on the traditional sector division based on the specific effect of switching states within the entire phase conduction domain on the tracking of the phase actual torque to the phase reference torque. The candidate switching states within the improved sectors are then redefined. This effectively reduces the number of candidate switching states within the sector while ensuring system stability, thereby reducing the computational burden on the controller.

[0007] The technical solution of the present invention is:

[0008] A sector predictive optimization control method based on a switched reluctance motor is proposed, which includes the following steps:

[0009] Step 1: Analyze the switching state of the asymmetric half-bridge leg;

[0010] Step 2: Derived the prediction model based on the system model of the switched reluctance motor, and the input of the prediction model is the switching state u ph (k);

[0011] Step 3: Divide the sectors of the model predictive torque control based on the torque distribution function;

[0012] Step 4: Based on the torque distribution function, study the switching state of each phase, and finally combine the three-phase switching states to form a candidate switching state combination;

[0013] Step 5: Analyze the number of switching state combinations under the traditional asymmetric half-bridge topology;

[0014] Step 6: Reduce the candidate switching states of the model predictive torque control according to the actual switching conditions of the switched reluctance motor, which are called traditional candidate switching states;

[0015] Step 7: Reduce the candidate switching state combinations in this interval based on the demagnetization characteristics of the inductor in the phase-leading torque decreasing region of the commutation region;

[0016] Step 8: Reduce the candidate switching state combinations in this interval based on the influence of the switching states in the leading torque rising region and the single-phase conduction region on the actual torque tracking reference torque in the commutation region;

[0017] Step 9: Analyze the actual torque tracking reference torque according to the motor characteristics and divide the commutation interval into sections;

[0018] Step 10: Improve the traditional sector division according to the division of the commutation interval;

[0019] Step 11: Re-formulate the switch states in the improved sector according to the specific effect of the actual torque tracking the reference torque, and further reduce the candidate switch state combinations;

[0020] Beneficial effects

[0021] This invention discloses a sector-specific predictive optimization control method for a switched reluctance motor. This method re-defines the candidate switching states within a sector based on the specific effects of the switching states within the entire phase conduction domain on the tracking of the phase actual torque with the phase reference torque. This effectively reduces the number of candidate switching states within a sector while ensuring system stability, thereby reducing the computational burden on the controller. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Asymmetric half-bridge topology

[0023] Figure 2 Working principle diagram of asymmetric half-bridge topology

[0024] Figure 3 Cosine torque distribution diagram

[0025] Figure 4 Sector division diagram of rotor position

[0026] Figure 5 Diagram of actual torque and reference torque of phase A under traditional candidate switching state

[0027] Figure 6 Commutation region [θ on +θ s ,θ off ] Phase A actual torque and reference torque diagram when the candidate switch state is {1, 0}

[0028] Figure 7 Commutation region [θ on +θ s ,θ off ] Phase A actual torque and reference torque diagram when the candidate switch state is {1, -1}

[0029] Figure 8 Phase A control diagram

[0030] Figure 9 Actual torque and reference torque diagram of phase A when the candidate switch state is {1, 0} in the single-phase conduction region

[0031] Figure 10 Actual torque and reference torque diagram of phase A when the candidate switch state is {1, -1} in the single-phase conduction region

[0032] Figure 11 Commutation region [θ on ,θ on +θ ov ]Actual torque and reference torque diagram of phase A when the candidate switch state is {1, 0}

[0033] Figure 12 Commutation region [θ on ,θ on +θ ov ]Actual torque and reference torque diagram of phase A when the candidate switch state is {1, -1}

[0034] Figure 13 Improved control diagram of phase A

[0035] Figure 14 Phase actual torque tracks phase reference torque diagram

[0036] Figure 15 Commutation area division diagram

[0037] Figure 16 Improved sector division diagram DETAILED DESCRIPTION

[0038] The following further describes examples of the present invention in conjunction with the accompanying drawings.

[0039] This example proposes a sector prediction optimization control method based on a switched reluctance motor, which includes the following steps:

[0040] Step 1: According to Figure 1 Asymmetric half-bridge topology and Figure 2 The working principle of the asymmetric half-bridge topology shows the relationship between the switching state and the working mode of the asymmetric half-bridge as shown in Table 1.

[0041] Table 1 Relationship between the switch state and working mode of AHBC

[0042]

[0043] Step 2: Model prediction The prediction model of torque control is shown in Equation 1, where the phase torque T ph (k) can be calculated based on current and position. And only u ph (k) is the unknown input quantity.

[0044]

[0045] Where: i ph (k) is the phase current, θ ph (k) is the phase position, R is the load resistance, T s is the control period, ψ ph (k) is the phase flux, u ph (k) is the winding voltage.

[0046] Step 3: According to Figure 3 The cosine torque distribution diagram is shown in Equation (2). The rotor position is divided into sectors. The division method is evolved based on the torque-angle characteristics of the switched reluctance motor. The sector division of the rotor position is as follows: Figure 4 As shown in the figure, sector 1 refers to the commutation overlap region of phases C and A; sector 2 refers to the independent conduction region of phase A; sector 3 refers to the commutation overlap region of phases A and B; sector 4 refers to the independent conduction region of phase B; sector 5 refers to the commutation overlap region of phases B and C; sector 6 refers to the independent conduction region of phase C.

[0047]

[0048] Where: T ref is the total reference torque, θ A is the rotor position of phase A, θ al is the rotor angular period, θ ov is the commutation overlap angle, θ s represents the angular displacement between adjacent phases, where θ on +θ s =θ off -θ ov , but in order to make it easier to express each angle in the entire rotor angle cycle, we use the unified θ on +θ s To express.

[0049] Step 4: Based on the torque distribution function, the switching state of each phase is studied, and finally the three-phase switching states are combined to form a candidate switching state combination.

[0050] Step 5: In the three-phase switched reluctance motor system, according to the analysis in step 1, the switching state of each phase has three values: {1}, {-1} and {0}. It can be deduced that the total number of switching state combinations of the motor is 3 3= 27, see Table 2 for details. During the operation of the model predictive control algorithm, in order to screen out the optimal switching state, these 27 switching states need to be tested one by one. The specific implementation method is to substitute different switching states into the cost function for calculation, and by comparing the calculation results, finally determine the optimal switching state under the current working conditions. In each subsequent sampling cycle, the traversal process of the above 27 switching states needs to be repeated. It is worth noting that if the prediction step size is further increased, the corresponding number of switch combinations will increase rapidly in the form of a power series. This means that in the actual control process, traversing all switch combinations will significantly increase the computational load of the control system, and put forward higher requirements for hardware performance and algorithm efficiency.

[0051] Table 2 Three-phase AHBC working state set

[0052]

[0053]

[0054] Step 6: The traditional switch state combination formulation is based on the above rotor position division. For the 27 switch state combinations of the asymmetric half-bridge topology, in order to exclude switch state combinations that will not appear, the traditional switch state formulation rules are as follows:

[0055] 1) In the single-phase conduction region, only one phase provides all the output torque. Therefore, in the single-phase conduction range (sectors 2, 4, and 6), the conducting phase has three switching states: {1, 0, and -1}. The other two phases are in the off state and have only one switching state: {-1}. Therefore, there are three possible switching state combinations.

[0056] 2) In the commutation region, adjacent phases simultaneously provide output torque. Therefore, in the commutation interval (sectors 1, 3, and 5), the commutating phases have three switching states: {1, 0, and -1}, while the off-phase phase has only one switching state: {-1}. Therefore, there are nine possible switching state combinations.

[0057] In summary, the traditional candidate switching states for the asymmetric half-bridge topology are detailed in Table 3. The valid switching state combinations in this table are obtained by eliminating switching state combinations that are unlikely to occur in actual operation. As can be seen from Table 3, while it is not necessary to traverse all 27 switching state combinations at any given time, up to nine switching state combinations must still be traversed during the commutation sector. Even with high-precision prediction models, this large number of state traversals still imposes a significant computational burden on the controller.

[0058] Table 3 Traditional candidate switch states

[0059]

[0060] Step 7: By analyzing the demagnetization problem of the previous phase winding in the commutation zone of the switched reluctance motor and taking into account the effect of the switching state on the actual torque tracking the reference torque, the switching state is further optimized to effectively reduce the number of switching state combinations. Figure 5 The actual torque and reference torque diagram of phase A in the traditional candidate switching state are shown in the figure. on +θ s ,θ off ]'s switch states to be selected are {1, 0, -1}. Figure 6 shows the commutation region [θ on +θ s ,θ off ] is the actual torque and reference torque of phase A when the candidate switch state is {1, 0}. Figure 7 shows the commutation region [θ on +θ s ,θ off ], the actual torque and reference torque of phase A when the candidate switching state is {1, -1}. By comparing the actual torque of phase A tracking the reference torque in the commutation region, it can be seen that the torque tracking effect and torque ripple when the candidate switching state is {1, 0, -1} are basically the same as those when the candidate switching state is {1, -1}. Compared with the candidate switching state {1, 0}, the torque tracking effect and torque ripple suppression when the candidate switching state is {1, -1} are more significant, and less negative torque is generated.

[0061] The reason for the above is that during the actual operation of the motor, the rate of change of the winding inductance of the phase in the commutation zone is relatively large, and the actual torque value generated by the phase winding is also relatively large. It is usually impossible to track the reference torque value of the phase. on +θ s ,θ off ] is indeed unable to track the reference torque. Therefore, when formulating the switching state, the switching state of the previous phase in the commutation area is set to {1, -1}, and the strong demagnetization ability of the "-1" switching state is used to make the actual torque of the motor track the reference torque as much as possible. The switching state of the leading phase in the commutation area is still set to {1, 0, -1}, and the switching state of other phases is closed, that is, the switching state is {-1}. Therefore, the switching state formulation rules are as follows:

[0062] 1) In the single-phase conduction region, only one phase provides all the output torque. Therefore, in the single-phase conduction range (sectors 2, 4, and 6), the conducting phase has three switching states: {1, 0, and -1}. The other two phases are in the off state and have only one switching state: {-1}. Therefore, there are three possible switching state combinations.

[0063] 2) In the commutation area, the two adjacent phases need to output torque at the same time. In order to ensure that the actual torque of the previous phase in the commutation area can drop rapidly and accurately track the reference torque, the switching state of the previous phase is set to {1, -1} in the commutation interval (sectors 1, 3, and 5). "-1" is selected for rapid demagnetization instead of "0" because the demagnetization effect of "0" is relatively weak. The switching state of the next phase in the commutation area remains at {1, 0, -1}, and the shutdown phase has only one switching state {-1}. Based on the above settings, the number of switch state combinations to be selected is reduced to 6. The optimized candidate switch state table is shown in Table 4, and the control diagram of phase A is shown in Figure 8 shown.

[0064] Table 4 Optimized candidate switch states

[0065]

[0066] Step 8: Step 7 has been done for the interval [θ on +θ s ,θ off ] has been studied, so step 8 only needs to study the phase A in the interval [θ on ,θ on +θ ov ] and [θ on +θ ov ,θ on +θ s ] to adjust the switch status.

[0067] First, analyze the single-phase conduction area [θ on +θ ov ,θ on +θ s ] affects the tracking of actual torque to reference torque using different switching states within this region. Within this region, increasing torque uses a switching state of "1," while there are two switching states for decreasing torque: "-1" and "0." Therefore, we investigated these two switching states to determine the most appropriate one. Figure 9 Shows the single-phase conduction region [θ on +θ ov ,θ on +θ s ] is the actual torque and reference torque of phase A when the candidate switch state is {1, 0}. Figure 10 Shows the single-phase conduction region [θ on +θ ov ,θ on +θ s ] is the actual torque and reference torque of phase A when the candidate switch state is {1, -1}. Among them, the waveform of the actual torque of phase A tracking the reference torque when the candidate switch state in the single-phase conduction area is {1, 0, -1} can be referred to Figure 5 . By comparing the actual torque of phase A tracking the reference torque in the single-phase conduction area in the figure, it can be found that when the candidate switch state is {1, 0, -1}, the torque tracking effect and torque pulsation are basically consistent with the candidate switch state {1, 0}. Compared with the candidate switch state {1, -1}, the torque tracking effect and torque pulsation suppression when the candidate switch state is {1, 0} are more superior. This shows that although all the switch states {1, 0, -1} are selected as candidate switch states in the single-phase conduction area, through the online optimization characteristics of the cost function of the model predictive control, the system mainly selects the "0" state to reduce the torque, and does not select the "-1" state to reduce the torque, because the use of the "-1" switch state will lead to an increase in torque pulsation.

[0068] Commutation region [θ on ,θ on +θ ov ]The influence of different switch states within the reference torque on the actual torque tracking the reference torque is analyzed. Similarly, the switch state problem of torque reduction is studied, and the switch state of torque increase is still "1". Figure 11 shows the commutation region [θ on ,θ on +θ ov ] Actual torque and reference torque diagram of phase A when the candidate switch state is {1, 0}. Figure 12 shows the commutation region [θ on ,θ on +θ ov ] The actual torque and reference torque diagram of phase A when the candidate switch state is {1, -1}. By comparing the commutation area [θ on ,θ on +θ ov ], the actual torque of phase A tracks the reference torque. It can be seen that when the candidate switch states are {1, 0, -1}, the torque tracking effect and torque ripple are basically the same as when the candidate switch state is {1, -1}. Compared with the candidate switch state {1, 0}, the torque tracking effect and torque ripple suppression are superior when the candidate switch state is {1, -1}. When the candidate switch state is {1, 0}, the torque tracking effect is poor and the torque ripple is large, which clearly shows that the torque reduction effect of using the "0" state is not good.

[0069] Through the above analysis of the entire phase opening area [θ on ,θ off The influence of the switch state in ] on the actual torque tracking the reference torque is obtained, and the improved switch state formulation rules are as follows:

[0070] 1) In the single-phase conduction region, only one phase provides all the output torque. Therefore, in the single-phase conduction range (sectors 2, 4, and 6), the conducting phase has only two switching states: {1, 0}. Switching state "-1" is not used in this stage because state "0" reduces both torque and switching frequency. The other two phases are in the off state, with only one switching state, {-1}. Therefore, there are a total of two possible switching state combinations.

[0071] 2) In the commutation region, two adjacent phases must output torque simultaneously. To ensure that the actual torque of the output phase during the commutation process decreases rapidly and accurately matches the reference torque, the switching state of the output phase is set to {1, -1} in the commutation interval (sectors 1, 3, and 5). The "-1" state is chosen rather than the "0" state to reduce torque because the torque regulation effect of the "0" state is poor and cannot meet the requirements of rapid adjustment. For the leading phase, the switching state is also set to {1, -1}. Analysis found that using the "0" state to reduce torque would result in reduced torque tracking accuracy and generate large torque ripple. The only switching state for the off phase is {-1}. Based on these optimized settings, the number of switch state combinations to be selected is reduced to four.

[0072] Table 5 Improved candidate switch states

[0073]

[0074] In summary, the improved candidate switch states of each phase are shown in Table 4. In the single-phase conduction area, there are two candidate switch state combinations; in the commutation area, there are four candidate switch state combinations. The improved control diagram of phase A is shown in Figure 13 It can be seen that the switch state combinations in the improved sector are reduced, which greatly reduces the operating burden of the controller and optimizes the switch state in the sector.

[0075] Step 9: Figure 14 It can be seen from the phase actual torque tracking phase reference torque diagram. At the beginning of the commutation stage, due to the relatively small inductance change rate in the leading phase motor model, the actual torque value generated by the phase winding is small, and it is impossible to track the reference torque value of the phase. At the end of the commutation stage, due to the relatively large inductance change rate in the outgoing phase motor model, the actual torque value generated by the phase winding is large, and it is also impossible to track the reference torque value of the phase, resulting in an increase in the total torque pulsation of the motor in the commutation interval. As the motor speed increases, the current change rate of the phase winding decreases, that is, the torque tracking effect decreases, and the torque pulsation in the commutation area increases. In response to the above problems, the motor commutation interval is divided into two parts, namely interval 1 and interval 2. The present invention sets the separation point of these two intervals to the relative position of the rotor corresponding to the moment when the actual output torque of the latter phase is equal to the reference torque. The separation point can also be adjusted online according to different operating conditions. The commutation area is divided as follows Figure 15 As shown, the separation point of CA commutation interval is x a , the separation point of AB commutation interval is x b , the separation point of BC commutation interval is x c .

[0076] Step 10: Figure 16 As shown in Figure 1, the improved sector division method is based on the traditional sector division method and is optimized based on the principle of commutation interval division. This method divides the commutation zone sector into two sub-sectors. Then, according to the specific situation of the actual torque of each phase tracking the reference torque, corresponding switching rules are formulated to achieve more precise control effects.

[0077] Step 11: Formulate rules for the switch states in the improved sectors as follows: In the single-phase conduction interval (sectors 2, 4, 6), the candidate switch states are still {1, 0}, "1" is used to increase the torque, and "0" is used to reduce the torque.

[0078] The commutation interval consists of two subintervals, sectors x.1 and x.2. The switching state rules for these subintervals are defined as follows. Phase A is used as an example for analysis, and the analysis for other phases is similar. In sector 1.1, phase A's rotor position is near its minimum inductance, resulting in weak torque generation. The actual torque cannot track the reference torque of phase A. Therefore, only switching state "1" is used to increase torque, and no other switching states are used. Therefore, the candidate switching state is {1}. In sector 1.2, phase A has strong torque generation and tracking capabilities, so the previously improved switching state {1, -1} is used to maintain its torque tracking capability. In sector 3.1, phase A has strong torque tracking capability, so the previously improved switching state {1, -1} is also used to maintain good torque tracking capability. In sector 3.2, phase A's rotor position is near its maximum inductance, resulting in strong torque generation. Under normal circumstances, it is difficult for the actual torque to track the reference torque. Therefore, only the "-1" state is used to reduce torque. The improved sector division and switch states to be selected are shown in Table 6.

[0079] Table 6 Improved sector division and switch status to be selected

[0080]

[0081]

[0082] (Continued)

[0083]

[0084] The table above shows that during the single-phase conduction period, the number of switching state combinations remains at two. However, during the commutation period, there are only two switching state combinations, regardless of sector x.1 or sector x.2. This maintains only two switching state combinations across the entire sector, significantly reducing the controller's operational burden.

Claims

1. A sector prediction optimization control method based on a switched reluctance motor, comprising the following steps: Step 1: Analyze the switching state of the asymmetric half-bridge leg; Step 2: Derived the prediction model based on the system model of the switched reluctance motor, and the input of the prediction model is the switching state u ph (k); Step 3: Divide the sectors of the model predictive torque control based on the torque distribution function; Step 4: Based on the torque distribution function, study the switching state of each phase, and finally combine the three-phase switching states to form a candidate switching state combination; Step 5: Analyze the number of switching state combinations under the traditional asymmetric half-bridge topology; Step 6: Reduce the candidate switching states of the model predictive control according to the actual conditions of the switch reluctance motor on and off, which are called traditional candidate switching states; Step 7: Further reduce the candidate switching state combinations based on the demagnetization characteristics of the inductor in the phase-leading torque decreasing region of the commutation region; Step 8: Further reduce the candidate switching state combinations based on the influence of the switching states in the leading torque rising region and the single-phase conduction region in the commutation region on the actual torque tracking reference torque; Step 9: Analyze the actual torque tracking reference torque according to the motor characteristics and divide the commutation interval into sections; Step 10: Improve the traditional sector division according to the division of the commutation interval; Step 11: Re-formulate the switch states in the improved sector according to the specific effect of the actual torque tracking the reference torque, and further reduce the candidate switch state combinations; 2. The sector prediction optimization control method based on the switched reluctance motor according to claim 1, characterized in that: By analyzing the effect of switching states across the entire phase conduction domain on tracking the actual torque and reference torque, the candidate switching states within each sector were re-optimized. While ensuring system stability, the number of candidate switching states within the sector was effectively reduced, thereby reducing the computational burden on the controller.

3. According to step 6 of claim 1, in order to exclude the switch state combination that will not occur, the switch state is formulated as follows: 1) In the single-phase conduction region, only one phase provides all the output torque. Therefore, in the single-phase conduction range (sectors 2, 4, and 6), the conducting phase has three switching states: {1, 0, and -1}. The other two phases are in the off state and have only one switching state: {-1}. Therefore, there are three possible switching state combinations. 2) In the commutation region, adjacent phases simultaneously provide output torque. Therefore, in the commutation interval (sectors 1, 3, and 5), the commutating phases have three switching states: {1, 0, and -1}, while the off-phase phase has only one switching state: {-1}. Therefore, there are nine possible switching state combinations. According to the above switching state formulation rules, it can be seen that the switching state combinations are reduced from 27 to 9. The candidate switching states for model predictive torque control are effectively reduced.

4. According to step 7 of claim 1, by analyzing the demagnetization problem of the previous phase winding in the commutation zone of the switched reluctance motor and taking into account the effect of the switching state on the actual torque tracking the reference torque, the switching state is further optimized, effectively reducing the number of switching state combinations. The reason for the above is that during the actual operation of the motor, the rate of change of the winding inductance of the phase in the commutation zone is relatively large, and the actual torque value generated by the phase winding is also relatively large. It is usually impossible to track the reference torque value of the phase. The interval [θ on +θ s ,θ off ] cannot track the reference torque. Therefore, when defining the switching state, the switching state of the preceding phase in the commutation region is defined as {1, -1}. The strong demagnetization capability of the "-1" switching state is utilized to ensure that the motor's actual torque tracks the reference torque as closely as possible. The switching state of the leading phase in the commutation region is also defined as {1, 0, -1}, and all other phases are closed, i.e., the switching state is {-1}. Therefore, the switching state definition rules are as follows: 1) In the single-phase conduction region, only one phase provides all the output torque. Therefore, in the single-phase conduction range (sectors 2, 4, and 6), the conducting phase has three switching states: {1, 0, and -1}. The other two phases are in the off state and have only one switching state: {-1}. Therefore, there are three possible switching state combinations. 2) In the commutation region (sectors 1, 3, and 5), two adjacent phases jointly contribute to torque output. To ensure that the actual torque of the preceding phase rapidly follows the decrease in reference torque, its switching state is set to {1, -1}, with "-1" accelerating demagnetization (which is more effective than "0" demagnetization). The succeeding phase still uses the conventional {1, 0, -1} switching state; the off phase is fixed to {-1}. This reduces the number of possible switching state combinations from nine to six, significantly reducing control complexity. The optimized switching states are shown in Table 1. Table 1 Optimized candidate switch states 5. According to step 8 in claim 1, step 7 is further optimized and improved, and the phase A is studied in the interval [θ on ,θ on +θ ov ] and [θ on +θ ov ,θ on +θ s ] to indicate the switch status. First, analyze the single-phase conduction area [θ on +θ ov ,θ on +θ s ] on the actual torque tracking reference torque. In this region, the switching state "1" is used to increase the torque, while there are two switching states for reducing the torque: "-1" and "0". Through analysis, it is found that when the candidate switching states are {1, 0, -1}, the torque tracking effect and torque ripple are basically consistent with those when the candidate switching state is {1, 0}. Compared with the candidate switching states {1, -1}, the torque tracking effect and torque ripple suppression when the candidate switching state is {1, 0} are more superior. This shows that although all the switching states {1, 0, -1} are selected as candidate switching states in the single-phase conduction region, the system mainly selects the "0" state to reduce the torque, rather than the "-1" state, due to the online optimization characteristics of the cost function of the model predictive control, because the use of the "-1" switching state will lead to increased torque ripple. Commutation region [θ on ,θ on +θ ov ] on the actual torque tracking reference torque. Similarly, the study focused on the switching state that reduces torque, while the switching state that increases torque is still "1." The analysis shows that when {1,-1} and {1,0,-1} are used as candidate switching states, the torque tracking performance and torque ripple suppression are essentially equivalent. However, compared to {1,0}, {1,-1} performs better in terms of torque tracking accuracy and ripple suppression. Specifically, when {1,0} is selected, the demagnetization effect of the "0" state is poor, resulting in torque tracking lag and significantly increased ripple. Therefore, {1,-1} is the more optimal switching state selection. Through the above analysis of the entire phase opening area [θ on ,θ off The influence of the switch state in ] on the actual torque tracking the reference torque is obtained, and the improved switch state formulation rules are as follows: 1) In the single-phase conduction region, only one phase provides all the output torque. Therefore, in the single-phase conduction range (sectors 2, 4, and 6), the conducting phase has only two switching states: {1, 0}. Switching state "-1" is not used in this stage because state "0" reduces both torque and switching frequency. The other two phases are in the off state, with only one switching state, {-1}. Therefore, there are two possible switching state combinations. 2) In the commutation region, two adjacent phases simultaneously provide output torque. However, to rapidly reduce the actual torque of the output phase so that it tracks the reference torque, the output phase switching states are specified as {1, -1} within the commutation interval (sectors 1, 3, and 5). The "-1" state is used to rapidly reduce torque, and the "0" state is not used, primarily because the torque reduction effect of "0" is poor. The leading phase switching state is also {1, -1}. Analysis shows that using "0" to reduce torque in the leading phase results in poor torque tracking and large torque ripple. The off phase only has one switching state, {-1}. Therefore, the number of candidate switching state combinations is reduced to four. The improved candidate switching states are shown in Table 2. Table 2 Improved candidate switch states 6. According to the method described in step 10 of claim 1, the conventional sector division is optimized and improved: based on the commutation interval division principle proposed above, each commutation sector is further divided into two sub-sectors. Based on this, the corresponding switching control strategy is formulated in combination with the tracking of the reference torque by the actual torque of each phase.

7. According to step 11 of claim 1, the improved candidate switch states in the sector are formulated, and the formulation rules are as follows: In the single-phase conduction interval (sectors 2, 4, and 6), the candidate switch states are still {1, 0}. "1" is used to increase the torque, and "0" is used to reduce the torque. There are two sub-intervals in the commutation interval, sector x.1 and sector x.

2. The switching states of the sub-intervals are governed by the following rules. Phase A is used as an example for analysis, and the analysis of other phases is similar. In sector 1.1, phase A's rotor position is near the minimum inductance position, so its torque-generating ability is weak. The actual torque cannot track the reference torque of phase A. Therefore, only switching state "1" is used to increase the torque, and no other switching states are used. Therefore, the candidate switching state is {1}. In sector 1.2, phase A has good torque generation and tracking capabilities, so the previously improved switching state {1, -1} is used to maintain phase A's torque tracking capability. In sector 3.1, phase A has good torque tracking capability, so the previously improved switching state {1, -1} is also used to maintain good torque tracking capability. In sector 3.2, phase A's rotor is near its maximum inductance position, and phase A has strong torque generation capabilities. Under normal circumstances, it is difficult for the actual torque to track the reference torque, so only the "-1" state is used to reduce torque. The improved sector division and candidate switching states are shown in Table 3. There are only two switching state combinations within the entire sector, which greatly reduces the operating burden of the controller. Table 3 Improved sector division and candidate switch status

Images