Improved optimal switching sequence model predictive control strategy for loss optimization of voltage source inverter

Abstract

The invention provides an improved optimal switching sequence model predictive control strategy for loss optimization of a voltage source type inverter. The method comprises the following steps: firstly, calculating a clamping phase and a clamping state according to a large-current clamping method; based on a conventional optimal switch sequence model design thought, designing a new symmetric five-segment switch sequence set; according to the newly generated switching sequence set, constructing a value function of an error sum of squares of an output voltage predicted value and a reference value based on the voltage gradient, and solving a partial derivative of the value function to obtain an analytic expression of action time of each voltage vector in a period; calculating the sum of squares of voltage errors under the action of each segment of vector in the sequence set, and selecting a switching sequence enabling the value function to be minimum as an optimal switching sequence; and outputting the optimal switching sequence and the corresponding action time, comparing the optimal switching sequence and the corresponding action time with a carrier, and sending a PWM signal. The method has the advantages that conventional optimal switch sequence model prediction control is improved by introducing a large current clamping method, on the premise that low output voltage ripples of the system are guaranteed, switch switching of a power switch tube at the large current position is reduced, and the switch loss of a power device is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of power electronic conversion, and in particular to an improved optimal switching sequence model predictive control strategy for loss optimization of a voltage source inverter. Background Art

[0002] Model Predictive Control (MPC) has been widely used in the field of power electronics conversion due to its advantages such as multi-objective optimization and rapid response. However, existing MPC research has largely focused on optimizing control performance, without comprehensively considering how thermal stress caused by power loss in the inverter can accelerate the performance degradation of key power devices such as IGBTs, thereby reducing overall system reliability. To address device losses, some researchers have proposed a finite set MPC strategy based on high-current clamping. By comparing the relationship between the three-phase reference voltage and output current, this strategy clamps the high-current phase, reducing the number of device switching operations under high current, thereby reducing switching losses and balancing junction temperature distribution. However, this strategy still suffers from the non-fixed switching frequency and high output current ripple of traditional finite control set MPC (FCS-MPC). Therefore, further reducing device switching losses while achieving low ripple and a fixed switching frequency is of great significance for improving and optimizing current MPC strategies. Summary of the Invention

[0003] To address the shortcomings of existing model predictive control (MPC) strategies, this paper proposes an improved optimal switching sequence MPC strategy for loss optimization in voltage source inverters. Building on the conventional optimal switching sequence MPC, this strategy incorporates a high-current clamping approach to obtain a new set of preselected voltage vectors. Combined with these preselected voltage vectors, the switching sequence order and action time are rearranged. Compared to conventional optimal switching sequence MPC (OSS-MPC) and high-current clamp-based FCS-MPC, this strategy reduces power losses and balances loss distribution while maintaining low output voltage ripple.

[0004] To achieve the above functions, the present invention involves the following technical steps:

[0005] Step 1: Establish a discrete mathematical model of a three-phase two-level voltage source inverter with an LC filter;

[0006] Step 2: Calculate the clamping phase and clamping state according to the high current clamping method;

[0007] Step 3: As described in claim S1, based on the current clamping bridge arm device, a new symmetrical five-stage switch sequence set is designed;

[0008] Step 4: As described in claim S2, based on the newly generated switching sequence set, construct a cost function based on the sum of squares of the error between the output voltage prediction value and the reference value based on the voltage gradient, and take the partial derivative of the vector action time to obtain an analytical expression for the optimal action time of the sequence;

[0009] Step 5: To minimize the voltage tracking error, calculate the sum of the squares of the voltage errors under the action of each vector in the sequence set, and select the switching sequence that minimizes the cost function as the optimal switching sequence;

[0010] Step 6: Output the optimal switching sequence and corresponding action time, compare with the carrier and send out a PWM signal;

[0011] In the step 1, a time domain mathematical model is constructed according to Kirchhoff's voltage and current laws, and is discretized using the forward Euler method.

[0012] In the step 2, the clamping phase and clamping state are calculated according to the large current clamping method; the maximum and minimum reference voltages at the next moment are selected, and the absolute values ​​of the corresponding inverter output currents are compared; if the output current corresponding to the maximum reference voltage is less than the output current corresponding to the minimum reference voltage, the clamping target is the phase where the minimum reference voltage is located, and when the voltage is greater than 0, the upper bridge arm is clamped, and when it is less than 0, the lower bridge arm is clamped; if the output current corresponding to the maximum reference voltage is greater than the output current corresponding to the minimum reference voltage, the clamping target is the phase where the maximum reference voltage is located, and when the voltage is greater than 0, the upper bridge arm is clamped, and when it is less than 0, the lower bridge arm is clamped;

[0013] In step three, a new five-segment symmetrically arranged switching sequence is designed based on the current clamping bridge arm condition; after the clamping phase is determined by the large current clamping method, the preselected vector set changes, and the vector set only contains one zero vector and three effective voltage vectors. To reduce the number of switching device operations, the switching sequence is redistributed. Based on the sequence characteristic of only operating one switch at a time, two effective voltage vectors and one zero vector are selected and distributed symmetrically within the five time segments of a cycle;

[0014] In step 4, a cost function for the sum of squared output voltage errors is constructed based on the newly generated switching sequence set; an expression for the predicted voltage final value during the sequence action cycle is obtained based on the voltage gradient and action time of each vector in the sequence; a cost function for the sum of squared errors between the predicted voltage and the reference voltage is further constructed, and an analytical expression for the optimal action time of each switching vector within the cycle is solved using the Lagrangian conditional extreme value method;

[0015] The present invention has the advantages of:

[0016] 1) The present invention takes into account the power loss problem of the device, obtains a new pre-selected voltage vector set under the high current clamping method, and rearranges the switching sequence with the purpose of reducing the number of switching device operations, thereby reducing the switching loss of the device and effectively lowering the device junction temperature;

[0017] 2) The present invention constructs a value function of the sum of squares of the output voltage error based on the voltage gradient change, and uses the Lagrange condition extreme value method to solve the vector action time, effectively improving the voltage tracking accuracy and reducing the output voltage ripple; BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is the topology diagram of the three-phase two-level voltage source inverter

[0019] Figure 2 Preselected voltage space vector diagram based on large current clamping

[0020] Figure 3 The output voltage trajectory diagram of the optimal switching sequence based on loss optimization

[0021] Figure 4 Comparison diagram of switching sequence action time information and carrier

[0022] Figure 5 Maximum current clamping control block diagram

[0023] Figure 6 Optimal switching sequence model prediction voltage control flow chart based on loss optimization DETAILED DESCRIPTION

[0024] The specific embodiments of the present invention will be described in detail with reference to preferred examples and accompanying drawings. It should be noted that the examples provided are only used to better understand the technical solutions of the present invention and do not constitute any limitation on the scope of protection of the present invention.

[0025] This invention provides a model-predictive voltage control method for reducing power loss in voltage-source inverter devices. This method uses high-current clamping to reduce the number of device switching operations under high current conditions. By rearranging the switching sequence of the clamped voltage preselected vector into five symmetrical segments, it improves the high losses associated with conventional model-predictive control of optimal switching sequences. The overall solution not only achieves excellent control results, but also reduces device losses and achieves a more balanced loss distribution.

[0026] The specific implementation steps of the present invention are as follows:

[0027] Step 1: According to Kirchhoff’s voltage and current laws, the mathematical model of the three-phase two-level inverter topology with LC filtering in the αβ coordinate system is:

[0028]

[0029] The discrete mathematical model obtained by using the forward Euler discretization method is:

[0030]

[0031] Among them, αβ represents the parameters in the two-phase stationary coordinate system, i Lαβ Indicates the output current of the inverter side, i oαβ Represents the output current after filtering, e αβ Indicates the output port voltage of the inverter, u oαβ Represents the filter capacitor voltage, L0 represents the line inductance, R0 represents the line resistance, and C represents the filter capacitor.

[0032] Step 2: Determine the clamping phase and the corresponding bridge arm switch state according to the large current clamping method.

[0033] First, the reference voltage at time k+1 is calculated by Lagrange extrapolation method. The calculation formula is:

[0034]

[0035] Since the dynamic change of the inverter output current is relatively slow, the i L (k) replaces i L (k+1).

[0036] Select the maximum and minimum reference voltages, the expression is:

[0037]

[0038] Depend on Figure 5 , by comparing the reference voltage extreme point V in real time max and V min and its corresponding inverter output current I _Vmax and I _Vmin The absolute value of I _Vmax Less than the absolute value I _Vmin When V min The phase where it is located is the phase that needs clamping, otherwise V max The phase in question is the one that needs clamping. The upper or lower bridge arm is automatically selected for clamping based on the voltage polarity of the phase to be clamped: the upper bridge arm is clamped for positive voltage and the lower bridge arm is clamped for negative voltage.

[0039] Step 3: Preselect the space vector set after high current clamping. Figure 2As shown, it only contains three valid vectors and one zero vector, and the vector corresponding to each clamping state after clamping is only within two sectors. In order to reduce the number of switching device operations, two valid voltage vectors and one zero vector are selected from the pre-selected set according to the sequence characteristics of only operating one switch at a time, and are distributed within the five time periods of one cycle. Taking the Q1 upper bridge arm clamping as an example, the Q1 upper bridge arm is continuously turned on, the switch state is 1, and the pre-selected voltage vectors are U1, U2, U6, and U7. There are two vector action sequences: U1→U2→U7→U2→U1 and U1→U6→U7→U6→U1. The corresponding vector action sequences under different clamping conditions are summarized in the following table:

[0040] Table 1 Summary of the clamping corresponding vector action sequence

[0041]

[0042] Step 4: Based on the voltage gradient corresponding to each voltage vector in the newly generated five-segment switching sequence, construct a value function of the sum of squared output voltage errors and take its partial derivative to obtain the optimal action time analytical expression for each voltage vector within a control cycle, as follows:

[0043] The voltage gradient corresponding to the voltage vector of the nth segment (n=1, 2, ..., 5) of the switching sequence in one control cycle can be expressed as:

[0044]

[0045] In the newly generated symmetrical five-segment switching sequence, the action time and voltage gradient of the voltage vectors in the first and fifth segments are the same, and the action time and voltage gradient of the voltage vectors in the second and fourth segments are the same. Based on the action mode of the switching sequence, a voltage prediction model based on voltage gradient is constructed:

[0046]

[0047] Among them, t1, t2, and t3 correspond to the first, second, and third voltage vector action times respectively.

[0048] Predicted voltage u at time k+1 oαβ With reference voltage The expression for the sum of squared errors is as follows:

[0049]

[0050] where t 1i , t 2i is the action time of the first and second voltage vectors in the i-th candidate vector sequence.

[0051] In order to minimize the sum of squares of voltage tracking errors, the expression for the sum of squares of voltage tracking errors needs to satisfy:

[0052]

[0053] Solving the above formula, we can get the optimal action time t opt1 , t opt2 , t opt3 The expression is as follows:

[0054]

[0055] Step 5: Based on the order and duration of the new switching sequence after the five-segment arrangement, calculate the sum of the squares of the voltage errors under the action of each vector in the sequence. The value function expression is:

[0056]

[0057] Where u oan It represents the predicted voltage value after the nth switching state is applied, which can be obtained by recursion through the following formula:

[0058]

[0059] Perform rolling optimization on the value function.

[0060] Step 6: Based on the optimal switching sequence and its corresponding switching action time obtained by rolling optimization in step 5, compare them with the carrier and issue a PWM signal.

[0061] The above is a specific embodiment of the present invention and its advantages, but the scope of protection of the present invention is not limited thereto. Those skilled in the art may make changes and modifications to the above embodiment without departing from the technical spirit and principles described in the present invention, and such changes and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An improved optimal switching sequence model predictive control strategy for voltage source inverter loss optimization, characterized in that: Based on the model predictive control of the optimal switching sequence of a conventional two-level topology, a large current clamping method is introduced to obtain a new five-stage switching sequence set, which achieves low output voltage ripple while reducing device switching losses. The method includes the following steps: S1: Based on the current clamping phase and clamping device results, a new symmetrical five-stage switching sequence set is designed; S2: Based on the new symmetric five-segment switching sequence set, a value function based on the sum of squares of the error between the output voltage prediction value and the reference value based on the voltage gradient is constructed, and the partial derivative of the vector action time is taken to obtain the analytical expression of the optimal action time of the sequence.

2. The method according to claim 1, characterized in that In S1, four candidate voltage vectors are obtained based on the selected clamping phase and clamping power device results. According to the principle that only one phase is in action when the sequence switch state is switched, two valid vectors and one zero vector are selected from the candidate voltage vectors to generate a new symmetrical five-segment switching sequence set.

3. The method according to claim 1, characterized in that In S2, based on the voltage gradient corresponding to each voltage vector in the new five-segment switching sequence, a value function of the sum of squared output voltage errors is constructed, and its partial derivative is taken to obtain the optimal action time analytical expression of each voltage vector within a control cycle, as follows: The voltage gradient corresponding to the voltage vector for the nth segment of the switching sequence in one control cycle, where n = 1, 2, ..., 5, can be expressed as: Among them, αβ represents the parameters in the two-phase stationary coordinate system, u oαβ Represents the filter capacitor voltage, i Lαβ Indicates the output current of the inverter side, i oαβ It represents the output current after filtering, and C represents the filter capacitor; In the newly generated symmetrical five-segment switching sequence, the action time and voltage gradient of the voltage vectors in the first and fifth segments are the same, and the action time and voltage gradient of the voltage vectors in the second and fourth segments are the same. Based on the action mode of the switching sequence, a voltage prediction model based on voltage gradient is constructed: Among them, t1, t2, and t3 correspond to the first, second, and third voltage vector action times respectively; Predicted voltage u at time k+1 oαβ With reference voltage The expression for the sum of squared errors is as follows: where t 1i , t 2i is the action time of the first and second voltage vectors in the i-th candidate vector sequence; In order to minimize the sum of squares of voltage tracking errors, the expression for the sum of squares of voltage tracking errors needs to satisfy: Solving the above formula, we can get the optimal action time t of the sequence opt1 , t opt2 , t opt3 The expression is as follows: According to the analytical expression of the optimal action time, the optimal action time of each candidate vector switching sequence can be further calculated.

Images