Low-switching-frequency three-dimensional satisfaction space optimization model predictive control method and device

Abstract

The application discloses a low-switching-frequency three-dimensional satisfactory space optimization model predictive control method and device, and the method comprises the following steps: dividing the multi-target optimization of a frequency converter system into main control target optimization and auxiliary control target optimization; estimating the state of the main control target through a control system mathematical model; constructing a three-dimensional satisfactory space according to the tracking deviation of the main control target, and implementing satisfactory space optimization on the main control target to select a candidate switching vector; selecting a corresponding processing method according to the number of the candidate switching vectors, so as to obtain an optimized switching state; and the processing method comprises a maximum average residence time method and a minimum ordering average method. The main control target and the auxiliary control target are divided into two independent optimization levels, so that the complex weight coefficient design and debugging process are effectively avoided. In addition, through the two-level optimization strategy, high-performance control of the motor is realized, and the switching frequency is effectively reduced, so that the optimization balance of energy efficiency and performance is realized.

Description

Technical Field

[0001] This invention relates to the field of power electronic equipment control, and in particular to a predictive control method and apparatus for a three-dimensional satisfaction space optimization model at low switching frequency. Background Technology

[0002] In industrial applications, high-power electric motors are widely used, serving as the core power source for various heavy machinery and automated production lines. With technological advancements, the performance requirements for electric motor control systems are increasingly demanding. This involves not only improving motor control accuracy and response speed but also emphasizing energy reduction and system reliability. Currently, frequency conversion control technology for high-power electric motors primarily relies on two mainstream methods: direct torque control and vector control. While these existing technologies have demonstrated some effectiveness in high-power electric motor applications, they often struggle to achieve an ideal balance between higher control performance and reduced energy consumption. Model predictive control technology, with its advanced control strategies and excellent adaptability, offers a new approach to overcoming these limitations of traditional technologies.

[0003] Traditional model predictive control (MPC) techniques achieve multi-objective optimization by constructing complex cost functions, which typically include primary and secondary objectives. In three-level frequency converter (MLC) applications, primary control objectives usually include excitation current error, torque current error, and midpoint voltage balance. As a secondary objective, reducing the switching frequency aims to decrease energy consumption and extend equipment life. However, due to potential conflicts between these objectives, setting the weighting coefficients in the cost function becomes extremely complex, increasing the difficulty of system debugging and optimization. This limits the widespread application of traditional MPC in high-power motor control. Summary of the Invention

[0004] The purpose of this invention is to provide a predictive control method and apparatus for a three-dimensional satisfaction space optimization model with low switching frequency, in order to solve the above-mentioned problems.

[0005] This invention is implemented according to the following technical solution:

[0006] This invention provides a predictive control method for a three-dimensional satisfaction space optimization model with low switching frequency, the method comprising:

[0007] The multi-objective optimization of the frequency converter system is divided into primary control objective optimization and auxiliary control objective optimization.

[0008] The state of the main control target is predicted by using a mathematical model of the control system;

[0009] A satisfaction space is constructed based on the estimated tracking deviation of the main control objective, and the satisfaction space is optimized for the main control objective to select candidate switch vectors.

[0010] The appropriate processing method is selected based on the number of candidate switches to obtain the optimal switch state; the processing methods include the maximum average dwell time method and the minimum sorted average method.

[0011] In one implementation, the primary control objectives include excitation current error, torque current error, and midpoint voltage balance, while the auxiliary objectives include reducing the switching frequency to reduce energy consumption and extend equipment life.

[0012] In one implementation, the satisfaction space constructed based on the tracking deviation prediction of the main control objective is as follows: excitation current error is represented on the x-axis, midpoint potential error on the y-axis, and torque current error on the z-axis; the steps for optimizing the satisfaction space are as follows:

[0013] Set the number of candidate switches N to 0;

[0014] Configure the satisfactory range and calculate the deviation of the main control targets;

[0015] The deviations of the midpoint potential, torque current, and excitation current are sequentially evaluated to determine if they are satisfactory; if they are all satisfactory, they are considered as candidate switching states in the initial screening.

[0016] If the midpoint potential and torque current are satisfactory, but the excitation current is unsatisfactory, then the excitation current is relaxed. After calculating the satisfactory range of the relaxed excitation current, it is then determined whether the excitation current is satisfactory. If it is satisfactory, it is used as a preliminary candidate switching state.

[0017] After incrementing the initial number of candidate switch states N by 1, determine whether to iterate through all switch states that meet the conditions. If so, these will be the final selected candidate switch states.

[0018] In one embodiment, the satisfactory interval of the relaxation excitation current is k times the satisfactory interval of the excitation current, where 1 < k < 2.

[0019] In one implementation, the step of selecting the appropriate processing method based on the number of candidate switches to obtain the optimal switch state includes: determining that the number of candidate switches N is greater than or equal to 1; if N is greater than or equal to 1, then there exists a switch state that satisfies the above conditions; otherwise, it indicates that after satisfaction space optimization, there is no switch state that meets the conditions; when there is a candidate switch state, the maximum average dwell time method is used to continue optimizing the auxiliary control index - switch frequency; when there is no candidate switch state, the minimum sorted average method is used to reselect the optimal switch state as a transition.

[0020] In one embodiment, the maximum average residence time method includes:

[0021] Calculate the distance between the deviation of the main control objective and the satisfaction boundary;

[0022] The time to reach the boundary is calculated by measuring the distance of the main control target deviation from the satisfactory boundary and the rate of change of movement toward the satisfactory boundary.

[0023] Calculate the number of switch switching times after the candidate switch state is implemented;

[0024] The average time of torque current deviation, excitation current deviation and midpoint potential deviation is calculated based on the number of switching operations, and the minimum value of the three is taken as the dwell time of this candidate switching state in the satisfactory space.

[0025] After obtaining the dwell time of all candidate switch states, the candidate switch state with the longest dwell time is selected as the optimal switch state, and the optimization of the auxiliary control target layer ends.

[0026] In one implementation, the formula for calculating the number of switchovers is as follows:

[0027] nsw = |S a (k+1)-S a (k)|+|S b (k+1)-S b (k)|+|S c (k+1)-S c (k)|

[0028] Among them, S a (k), S b (k),S c (k) represents the switching state of one phase of the inverter at time k, S a (k+1),

[0029] S b (k+1),S c (k+1) represents the switching state of one phase of the inverter at time k+1, and nsw represents the total number of switching times for the three phases.

[0030] In one embodiment, the minimum sorted average method includes:

[0031] The tracking deviations of each switch state under a certain control target are sorted separately.

[0032] Calculate the average value of the sorted values ​​of each switch state under different control objectives;

[0033] Obtain the average ranking value of all candidate switch states, select the candidate switch state with the smallest average ranking value as the optimal switch state, and end the optimization of the auxiliary control target layer.

[0034] In one implementation, the method for sorting the tracking deviations of each switch state under a certain control target is as follows:

[0035] For each set of switching states, calculate the tracking error and obtain the excitation current deviation Δi. sd The array, torque current deviation Δi sq An array of midpoint potential deviations Δu o Array;

[0036] Sort the elements in the three arrays respectively to obtain the sorted excitation current deviation sorting array R. id Torque-current deviation sorting array R iq Midpoint potential deviation sorting array R uo .

[0037] The present invention also provides a low-switching-frequency three-dimensional satisfaction space optimization model predictive control device, the device comprising:

[0038] The objective partitioning module is used to divide the multi-objective optimization of the frequency converter system into primary control objective optimization and auxiliary control objective optimization.

[0039] The main control target prediction module is used to predict the state of the main control target through the mathematical model of the control system;

[0040] The candidate switch selection module is used to construct a satisfaction space based on the estimated value of the main control objective, and to optimize the satisfaction space for the main control objective to select candidate vectors.

[0041] The optimal switch calculation module is used to select the appropriate processing method based on the number of candidate switches to obtain the optimal switch state; the processing methods include the maximum average dwell time method and the minimum sorted average method.

[0042] Compared with the prior art, the present invention has the following advantages:

[0043] This invention departs from traditional cost function optimization, instead dividing the primary control objective and auxiliary control objective into two independent optimization levels. At the primary control objective level, satisfactory optimization replaces traditional cost function optimization, thereby relaxing the strict constraints on the primary control objective and providing more possible switching states for achieving the auxiliary control objective. At the auxiliary control objective level, a maximum average dwell time strategy is employed to further optimize the candidate switching states. Through this two-level optimization strategy, high-performance motor control is achieved while effectively reducing the switching frequency, thus achieving an optimal balance between energy efficiency and performance. Attached Figure Description

[0044] The accompanying drawings, as part of this invention, are provided to further illustrate the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation thereof. Clearly, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0045] Figure 1 Here is a flowchart of a low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to an embodiment of the present invention;

[0046] Figure 2 A satisfactory space for construction provided in an embodiment of the present invention;

[0047] Figure 3 A flowchart for selecting candidate switch vectors by performing satisfaction space optimization on the main control objective, as provided in an embodiment of the present invention;

[0048] Figure 4 This is a flowchart illustrating a method for selecting the appropriate processing method based on the number of candidate switches, according to an embodiment of the present invention.

[0049] Figure 5 Flowcharts of the maximum average residence time method and the minimum sorted average method provided in an embodiment of the present invention;

[0050] Figure 6 This is a schematic diagram of a low-switching-frequency three-dimensional satisfaction space optimization model predictive control device according to an embodiment of the present invention.

[0051] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments will be clearly and completely described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0053] The optional embodiments of this disclosure are described in detail below with reference to the accompanying drawings.

[0054] Reference Figure 1 As shown, a predictive control method for a three-dimensional satisfaction space optimization model with low switching frequency is provided. The method includes the following steps:

[0055] Step S100: Divide the multi-objective optimization of the frequency converter system into primary control objective optimization and auxiliary control objective optimization.

[0056] Furthermore, the main control objectives in a three-level system include excitation current error, torque current error, and midpoint voltage balance, while auxiliary objectives include reducing the switching frequency to reduce energy consumption and extend equipment life.

[0057] Step S200: Predict the state of the main control target using the mathematical model of the control system.

[0058] Based on the collected current, speed, voltage, and continuous state equations of the motor and frequency converter, the discrete mathematical model of the control system is obtained by using the backward difference method.

[0059] The excitation current, torque current, and midpoint potential of the system at the next sampling time are estimated using a discrete mathematical model.

[0060] Specifically, taking an induction motor as an example, its control system mathematical model is as follows:

[0061]

[0062]

[0063]

[0064]

[0065] in u o For the midpoint potential deviation, S a ,S b ,S c For the open state corresponding to the bridge arm, i a i b i c For three-phase currents, Rs, Rr, Ls, Lr, and Lm represent the stator resistance, rotor resistance, stator inductance, rotor inductance, and mutual inductance, respectively. sd ,u sq i sd i sq These are the components of the stator voltage and current, ψ rd It is the flux linkage amplitude of the motor, ω r ω1 and ω2 are the rotor rotational angular velocity and stator flux linkage rotational angular velocity, respectively. It is the leakage flux coefficient.

[0066] The following can be obtained using the backward difference method:

[0067]

[0068]

[0069]

[0070] ψ rd (k+1)=k5T s i sd (k)+(1-k6T s )ψ rd (k).

[0071] Step S300: Construct a satisfaction space based on the estimated tracking deviation of the main control objective, and perform satisfaction space optimization on the main control objective to select candidate switch vectors.

[0072] Taking an induction motor as an example, such as Figure 2 As shown, the satisfaction space is constructed as follows: excitation current error is represented on the x-axis, midpoint potential error on the y-axis, and torque current error on the z-axis. Excitation current, torque current, and midpoint potential constitute the main control targets for the induction motor drive. (Refer to...) Figure 3 As shown, the specific steps for performing satisfaction space optimization to select candidate switch vectors for the main control objective are as follows:

[0073] S310: Set the number of candidate switches N to 0.

[0074] S320: Configure the satisfactory range and calculate the deviation of the main control target.

[0075] A vector is established with the main control target as its element. In this embodiment, the excitation current, midpoint voltage, and torque current are used as elements to establish the vector [i d u o i q Define a vector representing the deviation between the given value and the predicted value for each element. In this example, this deviation vector is [△i]. sd , △i sq , △u o A three-dimensional satisfaction space is established using the deviation vector of the main control objective. The allowable fluctuation range of this deviation is defined as the satisfaction interval, which then constitutes one dimension of the satisfaction space.

[0076] Taking torque current as an example, its satisfactory range is configured as [-δ T ,δ T This refers to the allowable range of torque current deviation fluctuation. The torque current setpoint is... The predicted feedback value is i sq (k+1), its tracking error is defined as Similarly, the excitation current deviation and the midpoint potential deviation Δi are calculated. sd ,△u o .

[0077] S330: Sequentially determine whether the deviations of the midpoint potential, torque current, and excitation current are satisfactory; if all are satisfactory, they are considered as candidate switch states for preliminary screening.

[0078] S331: Determine whether the midpoint potential is satisfactory.

[0079] Midpoint potential deviation In [-δ uo ,δ uo If the satisfaction level is 1, the midpoint potential satisfaction level is determined to be 1; otherwise, it is 0. If the satisfaction level is 1, the process proceeds to the torque current satisfaction level assessment; otherwise, the switch is deemed unsatisfactory with respect to the midpoint potential, and further assessment of the switch's state is abandoned.

[0080] S332: Determine whether the torque current is satisfactory.

[0081] Torque current deviation In [-δ T ,δ T If the satisfaction level is 1, the torque current satisfaction level is determined to be 1; otherwise, it is 0. If the satisfaction level is 1, the process proceeds to the excitation current satisfaction level assessment; otherwise, the switch is deemed unsatisfactory with respect to the midpoint potential, and further assessment of the switch state is abandoned.

[0082] S333: Determine if the excitation current is satisfactory.

[0083] deviation of excitation current In If the satisfaction level of the excitation current is 1, it is determined to be 0 otherwise. If the satisfaction level is 1, it is considered a candidate switch state; otherwise, it is determined that the switch is not satisfied with the excitation current, and the excitation current is relaxed. After calculating the satisfied range of the relaxed excitation current, it is determined whether the excitation current is satisfied again. If it is satisfied, it is considered a preliminary candidate switch state.

[0084] The calculation process for the satisfactory range of the relaxation excitation current is as follows: As shown in the following formula, there is an influence of the rotor time constant k5 between the flux linkage and the excitation current. Fluctuations in the excitation current, after filtering, affect fluctuations in the flux linkage. Therefore, utilizing this characteristic, when unsatisfactory excitation current deviations occur, the satisfactory range of the excitation current is further relaxed for a secondary judgment. The formula for estimating the flux linkage is as follows:

[0085] ψ rd (k+1)=k5T s i sd (k)+(1-k6T s )ψ rd (k)

[0086] The satisfactory range of the relaxation excitation current is k times the satisfactory range of the excitation current, where 1 < k < 2.

[0087] In a preferred embodiment, the satisfactory range of the excitation current is expanded by a factor of 1.5. The deviation of the excitation current is within... If the satisfaction level of the excitation current is 1, it is determined to be 0 otherwise. If the satisfaction level is 1, it is considered as a candidate switch state in the initial screening; otherwise, it is determined that the switch is still not satisfied with the excitation current and the switch state is abandoned.

[0088] S334: After incrementing the count N of the initially selected candidate switch states by 1, determine whether to traverse all switch states that meet the conditions. If traversal is performed, the selected switch states will be used as the final candidate switch states.

[0089] Upon entering this step, the switch state is added to the candidate switch state list. Further optimization is then performed at the auxiliary target layer. Simultaneously, the number of candidate switch states is incremented by 1. The condition for meeting the criteria is that, under a given switch state, the selectable eligible switch states must satisfy the condition that the voltage jump is less than half of the DC bus voltage, in order to reduce the rate of change of bridge arm voltage and line voltage.

[0090] Step S400: Select the appropriate processing method based on the number of candidate switches to obtain the optimal switch state;

[0091] Furthermore, the processing methods include the maximum average residence time method and the minimum sorted average method.

[0092] In this embodiment, after satisfaction space optimization, it is determined whether a switching state meeting the conditions has been selected. Here, N is greater than or equal to 1, indicating that a switching state meeting the above conditions exists; otherwise, it indicates that no switching state meeting the conditions has been selected after satisfaction space optimization. When a candidate switching state exists, the process proceeds to S410 to continue optimizing the auxiliary control index—switching frequency. When no candidate switching state exists, the process proceeds to S420 to reselect the optimal switching state as a transition. Figure 4 The diagram shows a flowchart of the predictive control method for a three-dimensional satisfaction space optimization model at low switching frequency, in which the appropriate processing method is selected based on the number of candidate switches.

[0093] Step S410: Continue to optimize the auxiliary control index - switching frequency using the maximum average residence time method.

[0094] For reference Figure 5As shown, this step belongs to the optimization of the auxiliary control target layer. The auxiliary control target can be common-mode voltage, switching frequency, etc. In this embodiment, the auxiliary control target is set to the switching frequency. One or more candidate switching states are selected through satisfaction space optimization. In order to select the optimal switching state that minimizes the system switching frequency from these switching states, this invention proposes a maximum dwell time optimization method. As the name suggests, it is the method that maximizes the average dwell time for one switch change within the satisfaction space. The specific implementation steps are as follows:

[0095] Step S411: Calculate the distance between the deviation of the main control target and the satisfaction boundary.

[0096] Specifically, the main control target moves within the satisfaction boundary according to a certain rate of change. Taking torque current as an example, the rate of change can be obtained by the following formula:

[0097]

[0098] Then the distance L from the boundary:

[0099]

[0100] Similarly, the distances from the excitation current deviation and the midpoint potential deviation to their respective satisfactory boundaries can be obtained: L u .

[0101] Step S412: Calculate the time to reach the boundary by measuring the distance of the main control target deviation from the satisfactory boundary and the rate of change of movement toward the satisfactory boundary;

[0102] Specifically, by controlling the distance of the target deviation from the satisfactory boundary and the rate of change of movement towards the satisfactory boundary, the time to reach the satisfactory boundary can be obtained. Taking torque current as an example, the time for the torque current deviation to reach the satisfactory boundary is:

[0103]

[0104] The denominator is the discretized result of the continuous rate of change formula. Similarly, the time for the excitation current deviation to reach the satisfactory boundary and the time for the midpoint potential deviation to reach the satisfactory boundary can be obtained. t uo .

[0105] Step S413: Calculate the number of switch switching times after the candidate switch state is implemented.

[0106] Furthermore, calculate the number of switch switching operations after this candidate switch state is implemented. The formula is as follows:

[0107] nsw = |S a (k+1)-S a (k)|+|Sb (k+1)-S b (k)|+|S c (k+1)-S c (k)|

[0108] Among them, S a (k), S b (k),S c (k) represents the switching state of one phase of the inverter at time k, S a (k+1),

[0109] S b (k+1),S c (k+1) represents the switching state of one phase of the inverter at time k+1, and nsw represents the total number of switching times for the three phases.

[0110] Step S414: Calculate the average switching time of torque current deviation, excitation current deviation and midpoint potential deviation based on the number of switching cycles, and take the minimum value of the three as the dwell time of this candidate switching state in the satisfactory space.

[0111] For a given candidate switching state, the average switching time of its torque current deviation is:

[0112] T av1 =nsw / t T ,

[0113] Its excitation current deviation average switching time is:

[0114]

[0115] The average switching time of the point potential deviation is:

[0116] T av3 =nsw / t uo

[0117] The minimum value of the three is taken as the residence time of the error of this candidate switch state in the satisfactory space.

[0118] T avi =max{T av1 ,T av2 T av3}

[0119] Step S415: After obtaining the dwell time of all candidate switch states, select the candidate switch state with the longest dwell time as the optimal switch state, and end the optimization of the auxiliary control target layer.

[0120] In step S300, when the number of candidate switch states N is 0, it indicates that the deviation of all switch states exceeds the limit of the satisfaction space. At this time, no candidate switch states will enter the optimization layer of the auxiliary control target. This state is defined as an out-of-control state. In order to quickly escape this out-of-control state, this invention proposes a minimum average sorting value optimization strategy.

[0121] Step S420: Using the minimum sorting average method, reselect the optimal switching state as the transition. (Refer to...) Figure 5 As shown, the specific steps are as follows:

[0122] Step S421: Sort the tracking deviations of each switch state under a certain control target.

[0123] The tracking error Δi of the main control target has been obtained in the above steps. sd ,△i sq ,△u o For each set of switch states, calculate Δi. sd ,△i sq ,△u o And thus obtain the excitation current deviation Δi sd The array, torque current deviation Δi sq An array of midpoint potential deviations Δu o The elements in the three arrays are sorted respectively to obtain the sorted excitation current deviation sorting array R. id Torque-current deviation sorting array R iq Midpoint potential deviation sorting array R uo .

[0124] Step S422: Calculate the average value of the sorted values ​​of each switch state under different control objectives.

[0125] Calculate the average value of the switch state across the three sets of deviation sorting arrays to obtain an array Rank[i] of the average deviation sorting values ​​for all selectable switch states.

[0126] Step S423: Obtain the average ranking value of all candidate switch states, select the candidate switch state with the smallest average ranking value as the optimal switch state, and end the optimization of the auxiliary control target layer.

[0127] Specifically, the switch state corresponding to the minimum value in the array Rank[i] is selected as the optimal switch state. Then the minimum average sort value optimization ends.

[0128] Finally, the optimal switching state selected above is output to the power semiconductor device. Optionally, the output method can be electrical signal transmission or optical signal transmission, ultimately controlling the power semiconductor device to turn on or off.

[0129] This invention provides a three-dimensional satisfactory space optimization model predictive control method with low switching frequency. By pre-selecting switching states, the computational load of system optimization is reduced. Multi-objective optimization is divided into primary control objective optimization and auxiliary control objective optimization. By implementing satisfactory optimization control for the primary control objective, the complex weight coefficient design process is avoided. By optimizing the auxiliary control objective, the switching frequency of the system is further reduced.

[0130] In one embodiment, a low-switching-frequency three-dimensional satisfaction space optimization model predictive control device is proposed. For example... Figure 6 As shown, the structure of the device includes,

[0131] The objective partitioning module is used to divide the multi-objective optimization of the frequency converter system into primary control objective optimization and auxiliary control objective optimization.

[0132] The main control target prediction module is used to predict the state of the main control target through the mathematical model of the control system;

[0133] The candidate switch selection module is used to construct a satisfaction space based on the estimated value of the main control objective, and to optimize the satisfaction space for the main control objective to select candidate vectors.

[0134] The optimal switch calculation module is used to select the appropriate processing method based on the number of candidate switches to obtain the optimal switch state; the processing methods include the maximum average dwell time method and the minimum sorted average method.

[0135] In this embodiment, the main control targets include the midpoint potential, magnetizing current, and torque current. Alternatively, it could be torque, flux linkage, or midpoint potential.

[0136] Furthermore, the candidate switch selection module includes a satisfaction space optimization module and a candidate switch quantity determination module. The satisfaction space optimization module implements the satisfaction space optimization strategy and selects candidate switch states. The candidate switch quantity determination module determines how many candidate switch states the satisfaction space optimization strategy selects, and then determines the implementation strategy for entering the auxiliary control target optimization layer.

[0137] Furthermore, the optimal switching calculation module includes a maximum average dwell time optimization module and a minimum average ranking value optimization module. The maximum average dwell time optimization module is used to further optimize the auxiliary control target switching frequency when the number of candidate switch states is greater than or equal to 1. The minimum average ranking value optimization module is used to optimize the runaway state when the number of candidate switch states is greater than or equal to 0.

[0138] It should be noted that the low-switching-frequency three-dimensional satisfaction space optimization model predictive control device provided in the above embodiments is only illustrated by the division of the above functional modules when executing the low-switching-frequency three-dimensional satisfaction space optimization model predictive control method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the equipment can be divided into different functional modules to complete all or part of the functions described above. In addition, the low-switching-frequency three-dimensional satisfaction space optimization model predictive control device and the low-switching-frequency three-dimensional satisfaction space optimization model predictive control method embodiments provided in the above embodiments belong to the same concept, and its implementation process is detailed in the low-switching-frequency three-dimensional satisfaction space optimization model predictive control method embodiments, which will not be repeated here.

[0139] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0140] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features found in other embodiments but not others, combinations of features from different embodiments are also within the scope of protection of this invention and form different embodiments. For example, in the embodiments described above, those skilled in the art can use them in combination based on known technical solutions and the technical problems to be solved by this application.

[0141] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-described technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A predictive control method for a three-dimensional satisfaction space optimization model with low switching frequency, characterized in that, The method includes: The multi-objective optimization of the three-level frequency converter system is divided into primary control objective optimization and auxiliary control objective optimization; the primary control objective includes excitation current error, torque current error and neutral point voltage balance, and the auxiliary control objective includes reducing the switching frequency; The state of the main control target is predicted by using a mathematical model of the control system; The mathematical model of the control system is obtained in the following way: based on the collected current, speed, voltage, and continuous state equations of the motor and frequency converter, the discrete mathematical model of the control system is obtained by using the backward difference method. A satisfaction space is constructed based on the estimated tracking deviation of the main control objective, and the satisfaction space is optimized for the main control objective to select candidate switch vectors. The appropriate processing method is selected based on the number of candidate switches to obtain the optimal switching state. The processing method includes the maximum average dwell time method and the minimum sorted average method. When there are candidate switching states, the maximum average dwell time method is used to optimize the switching frequency of the auxiliary control target. When there are no candidate switching states, the minimum sorted average method is used to select the optimal switching state as the transition.

2. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 1, characterized in that, The satisfaction space constructed based on the tracking deviation prediction of the main control objective is as follows: the excitation current error is represented on the x-axis, the midpoint potential error on the y-axis, and the torque current error on the z-axis; the steps for optimizing the satisfaction space for the main control objective and selecting candidate switching vectors are as follows: Set the number of candidate switches N to 0; Configure the satisfactory range and calculate the deviation of the main control targets; The deviations of the midpoint potential, torque current, and excitation current are sequentially evaluated to determine if they are satisfactory; if they are all satisfactory, they are considered as candidate switching states in the initial screening. If the midpoint potential and torque current are satisfactory, but the excitation current is unsatisfactory, then the excitation current is relaxed. After calculating the satisfactory range of the relaxed excitation current, it is then determined whether the excitation current is satisfactory. If it is satisfactory, it is used as a preliminary candidate switching state. After incrementing the initial candidate switch states by 1, determine whether to iterate through all the switch states that meet the conditions. If so, these will be the final selected candidate switch states.

3. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 2, characterized in that, The satisfactory range of the relaxation excitation current is k times the satisfactory range of the excitation current, where 1 < k < 2.

4. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 2, characterized in that, The step of selecting the appropriate processing method based on the number of candidate switches to obtain the optimal switching state includes: determining whether the number of candidate switches N is greater than or equal to 1; if N is greater than or equal to 1, then there exists a switching state that meets the above conditions; otherwise, it means that after the satisfaction space optimization, there is no switching state that meets the conditions; when there is a candidate switching state, the maximum average dwell time method is used to continue optimizing the switching frequency of the auxiliary control target; when there is no candidate switching state, the minimum sorted average method is used to reselect the optimal switching state as the transition.

5. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 4, characterized in that, The maximum average residence time method includes: Calculate the distance between the deviation of the main control objective and the satisfaction boundary; The time to reach the boundary is calculated by measuring the distance of the main control target deviation from the satisfactory boundary and the rate of change of movement toward the satisfactory boundary. Calculate the number of switch switching times after the candidate switch state is implemented; The average time of torque current deviation, excitation current deviation and midpoint potential deviation is calculated based on the number of switching operations, and the minimum value of the three is taken as the residence time of the error of this candidate switching state within the satisfactory space. After obtaining the dwell time of all candidate switch states, the candidate switch state with the longest dwell time is selected as the optimal switch state, and the optimization of the auxiliary control target layer ends.

6. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 5, characterized in that, The formula for calculating the number of switch switching operations is as follows: ， in, , , , respectively the first The switching status of one phase of the frequency converter at all times. , , The first The switching status of one phase of the frequency converter at all times. This represents the total number of switching operations for the three phases.

7. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 4, characterized in that, The minimum sorted average method includes: The tracking deviations of each switch state under each main control objective are sorted. Calculate the average value of the sorted values ​​of each switch state under different control objectives; Obtain the average ranking value of all candidate switch states, select the candidate switch state with the smallest average ranking value as the optimal switch state, and end the optimization of the auxiliary control target layer.

8. The low-switching-frequency three-dimensional satisfaction space optimization model predictive control method according to claim 7, characterized in that, The method for sorting the tracking deviations of each switch state under each main control objective is as follows: For each set of switching states, calculate the tracking error and obtain the excitation current deviation Δi. sd The array, torque current deviation Δi sq An array of midpoint potential deviations Δu o Array; Sort the elements in the three arrays respectively to obtain the sorted excitation current deviation sorting array R. id Torque-current deviation sorting array R iq Midpoint potential deviation sorting array R uo .

9. A predictive control device for a low-switching-frequency three-dimensional satisfaction space optimization model, characterized in that, The device includes: The target partitioning module is used to divide the multi-objective optimization of the three-level frequency converter system into primary control target optimization and auxiliary control target optimization; the primary control targets include excitation current error, torque current error and neutral point voltage balance, and the auxiliary control targets include reducing the switching frequency. The main control target prediction module is used to predict the state of the main control target through the mathematical model of the control system. The mathematical model of the control system is obtained by means of the following method: based on the collected current, speed, voltage, and continuous state equations of the motor and frequency converter, the discrete mathematical model of the control system is obtained by using the backward difference method. The candidate switch selection module is used to construct a satisfaction space based on the estimated value of the main control objective, and to optimize the satisfaction space for the main control objective to select candidate vectors. The optimal switch calculation module is used to select the appropriate processing method based on the number of candidate switches, thereby obtaining the optimal switch state. The processing method includes the maximum average dwell time method and the minimum sorted average method. When there are candidate switch states, the maximum average dwell time method is used to optimize the switching frequency of the auxiliary control target. When there are no candidate switch states, the minimum sorted average method is used to select the optimal switch state as the transition.

Images