阅读说明：本技术 一种基于改进型智能pi控制的逆变器中点电位平衡方法 (Inverter neutral point potential balancing method based on improved intelligent PI control ) 是由 辛瑞芝 李峰 卢佳南 徐悦 张洪波 刘鼎立 戴宁 孙建萍 于 2019-11-05 设计创作，主要内容包括：本发明涉及一种基于改进型智能PI控制的逆变器中点电位平衡方法,该控制方法主要针对传统PI控制参数调整难度较大,需要人为改动参数,不能实现参数自适应调整的缺点,在现有常规PI调节器控制基础上引入了单神经元自适应PI控制,可以实现PI参数自适应调整,具有速度快、响应快、系统稳定的特点。(The invention relates to an inverter neutral point potential balancing method based on improved intelligent PI control, which mainly aims at the defects that the traditional PI control parameter adjustment is difficult, parameters need to be changed manually and the parameter self-adaptive adjustment cannot be realized.)

1. An inverter midpoint potential balancing method based on improved intelligent PI control is characterized by comprising the following steps:

step 1, introducing a balance circuit on the basis of a traditional three-level inverter, and outputting a direct current side and a direct current side capacitor C₁、C₂Two switching tubes S are added in the middle_d1、S_d2And an inductance L₁；

Step 2, switching tubes Sd1 and Sd2 are sequentially switched on according to two complementary PWM signals, when Sd1 is switched on and Sd2 is switched off, a capacitor C1, a switching tube Sd1 and a balance inductor L1 on the direct current side form a charging loop, and energy is transferred from the capacitor C1 to the inductor L1; when the switching tube Sd1 is turned off and Sd2 is turned on, at this time, the dc side capacitor C2, the switching tube Sd2 and the balance inductor L1 form a discharge loop, and the energy transfer direction is from the inductor L1 to the capacitor C2; the voltages on the capacitors C1 and C2 form dynamic balance, and the midpoint potential fluctuates in a small range;

step 3, adopting a single neuron self-adaptive PI controller, adopting double closed-loop control for a balance circuit control strategy, and adopting a direct current side capacitor voltage U_C1And U_C2The deviation value of the current loop is used as the input of the voltage loop, the output of the voltage loop after passing through the PI link is used as the input of the current loop, and then the current loop and the current i flowing into the midpoint potential are used as the input of the current loop₃The deviation is taken and passes through a PI controller, the output of a current loop is compared with a triangular wave to obtain two complementary PWM signals for controlling the smoothingTwo switch tubes S in constant circuit_d1And S_d2To be turned on and off to reach U_C2Real-time tracking U_C1The effect of (1).

2. The method for controlling the neutral point potential balance of the T-type three-level inverter according to claim 1, wherein the connection mode of the balance circuit in the step 1 is as follows:

will switch the tube S_d1、S_d2In series, connecting S_d1One end of the capacitor is connected with the direct current side output and the capacitor C₁To convert S into_d2One end of the capacitor is connected with the direct current side output and the capacitor C₂Between, inductance L₁One end of is connected with the switch tube S_d1And S_d2Inductance L₁Another terminal of the capacitor is connected to the capacitor C₁And C₂In the meantime.

3. The method as claimed in claim 1 or 2, wherein the single-neuron adaptive PI controller in step 3, the capacitor voltage U, and the neutral point potential balance control method in the T-type three-level inverter are respectively configured as follows_C1、U_C2Converted into state quantity x required by neuron learning control after passing through a state converter₁、x₂，x₁＝e(k)-e(k+1)、x₂＝e(k)，w_i(k) To correspond to x_i(k) K is a scale factor of the neuron, and the value of K is greater than 0.

4. The method as claimed in claim 3, wherein the neurons generate the control signals by correlation search

5. The method as claimed in claim 4, wherein the adaptive controller for single neuron realizes adaptive and self-organizing functions by adjusting the weighting coefficients, and the adjustment of the weighting coefficients adopts supervised Hebb learning rule, which is related to the correlation function of the input, output and output deviation of neuron, i.e. the method for controlling the midpoint potential balance in T-type three-level inverter is characterized in that the adaptive controller for single neuron realizes adaptive and self-organizing functions by adjusting the weighting coefficients

Wherein z (k) is an output error signal, and z (k) is U_C1-U_C2η is learning Rate, η>0; c is a constant greater than zero.

6. The method as claimed in claim 5, wherein the learning algorithm is normalized and then the midpoint potential balance control method is applied

In the formula, η_P、η_IThe learning rates are integral and proportional respectively;

the comparative example (P) and the integral (I) respectively adopt different learning rates η_P、η_ISo that the respective weight coefficients can be respectively adjusted according to the needs, and the values of the weight coefficients are determined by simulation and experiments.

Technical Field

The invention belongs to the technical field of power electronic control, and particularly relates to an inverter neutral point potential balancing method based on improved intelligent PI control.

Background

The three-level topological structure has the advantages of large output capacity, high output voltage, small current harmonic content and the like, so that the three-level topological structure is widely applied to the field of variable frequency speed regulation of high-voltage high-power alternating current motors.

In the current Pulse Width Modulation (PWM) method of the T-type three-level inverter, because the control model of the Space Vector Pulse Width Modulation (SVPWM) method is simpler, the method has the characteristics of small torque ripple, high voltage utilization rate and the like, and can also fully utilize the redundant state of the voltage vector to control the midpoint potential balance of the dc side of the inverter and reduce the switching loss of the inverter, so that the method is widely applied.

The neutral point potential balance problem is an inherent problem of a three-level inverter, and scholars at home and abroad propose a plurality of methods to solve the problem, and the common methods comprise the steps of constructing a virtual space vector, reasonably selecting an action sequence of a redundant vector and injecting a zero-sequence voltage component. However, the calculation is complex, the control system is difficult to realize, and the online realization is not facilitated.

Disclosure of Invention

Aiming at the defects of the traditional technology for solving neutral point potential imbalance based on PI control, the invention provides a neutral point potential balance control method of a T-type three-level inverter based on single neuron self-adaptive PI control. The control method mainly aims at the defects that the traditional PI control parameter adjustment is difficult, parameters need to be changed manually, and parameter self-adaptive adjustment cannot be realized, introduces single neuron self-adaptive PI control on the basis of the control of the conventional PI regulator, can realize PI parameter self-adaptive adjustment, and has the characteristics of high speed, quick response and stable system.

At present, the shift of the midpoint potential of the three-level inverter is caused by ① capacitor C₁And a capacitor C ₂② DC side DC③, as known from kirchhoff's current law KCL, if the sum of the midpoint current vectors on the direct current side is not zero, the potential of the midpoint will shift, thereby affecting the normal operation of the system.

As shown in fig. 1, it is a schematic diagram of a dc side structure of a T-type three-level inverter. U shape_dcIs a DC power supply, U₀Is a midpoint potential, C₁And C₂For a DC capacitor, the current i flowing into the midpoint is set₃The current i of two direct current capacitors can be obtained in the positive direction₁And i₂Comprises the following steps:

from kirchhoff's current law

And taking integrals from two sides simultaneously to obtain:

in the formula i₃Is the sum of the phase currents when the switch state is "O" state during a cycle.

Thus, it can be seen that: whether the midpoint potential of the T-type three-level inverter is stable or not is related to the current flowing into the midpoint potential in one period of the switching tube.

Three phases at the output end A, B, C of the T-type three-level inverter have three working states for each phase: p state, output voltage + U _dc2; o state, output voltage is 0; n state, output voltage of-U _dc2; the switching characteristics can be equivalent to:

wherein x is a, b, c.

The invention adopts the technical scheme that an inverter neutral point potential balancing method based on improved intelligent PI control is provided, and the method comprises the following steps:

step 1, a balancing circuit is introduced on the basis of the traditional T-type three-level inverter, namely a capacitor C at the output side of direct current and the direct current side₁、C₂Two switching tubes S are added in the middle_d1、S_d2And an inductance L₁。

Step 2, switching tube S_d1And S_d2Conducting in sequence according to two complementary PWM signals when S is_d1Is conducted and S_d2Capacitor C on the DC side when turned off₁Switch tube S_d1And a balance inductance L₁Form a charging loop, the energy of which is the secondary capacitor C₁Transfer to the inductance L₁The above step (1); when switching tube S_d1Is turned off and S_d2When conducting, the DC side capacitor C₂Switch tube S_d2And a balance inductance L₁Forming a discharge loop with energy transfer direction from the inductor L₁Transfer to a capacitor C₂The above. Energy transfer dynamic balance, at which time the capacitance C₁And C₂The upper voltage forms dynamic balance, and the potential of the neutral point fluctuates in a small range.

And 3, adopting a single-neuron self-adaptive PI controller, and adopting double closed-loop control by a balance circuit control strategy. DC side capacitor voltage U_C1And U_C2The deviation value of the current loop is used as the input of the voltage loop, the output of the voltage loop after passing through the PI link is used as the input of the current loop, and then the current loop and the current i flowing into the midpoint potential are used as the input of the current loop₃The deviation is taken and is compared with the triangular wave through a PI controller and the output of a current loop to obtain two complementary PWM signals which are used for controlling two switching tubes S in a balance circuit_d1And S_d2To finally reach U_C2Real-time tracking U_C1The effect of (1).

The invention has the beneficial effects that:

because the traditional PI has the problem that parameters are not easy to set, the invention improves the traditional PI of an outer ring, introduces a balance circuit and a single neuron self-adaptive PI controller, and improves the stability of the system. The invention adopts a single neuron-based adaptive PI control strategy, can adjust the parameters of the controller on line in real time, has the advantages of certain anti-interference capability, adaptive capability and the like, is easy to realize and has good application prospect.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are specific embodiments of the invention, and that other drawings within the scope of the present application can be obtained by those skilled in the art without inventive effort.

FIG. 1 is a schematic diagram of a DC side structure of a T-type three-level inverter;

FIG. 2 is a schematic of a topology in which a T-type three-level inverter incorporates a balancing circuit;

FIG. 3 is a schematic diagram of a single neuron adaptive PI controller principle.

Detailed Description

The present invention will be described in detail below with reference to the drawings and examples, including but not limited to the following examples.

The invention relates to an inverter midpoint potential balancing method (short method) based on improved intelligent PI control, which comprises the following steps:

step 1, a balancing circuit is introduced on the basis of the traditional T-type three-level inverter, namely a capacitor C at the output side of direct current and the direct current side₁、C₂Two switching tubes S are added in the middle_d1、S_d2And an inductance L₁。

As shown in fig. 2, it is a topological diagram of the T-type three-level inverter introduced with the balancing circuit. Will switch the tube S_d1、S_d2In series, connecting S_d1One end of the capacitor is connected with the direct current side output and the capacitor C₁To convert S into_d2One end of the capacitor is connected with the direct current side output and the capacitor C₂Between, inductance L₁One end of is connected with the switch tube S_d1And S_d2Inductance L₁Another terminal of the capacitor is connected to the capacitor C₁And C₂In the meantime.

Step 2, switching tube S_d1And S_d2Conducting in sequence according to two complementary PWM signals when S is_d1Is conducted and S_d2Capacitor C on the DC side when turned off₁Switch tube S_d1And a balance inductance L₁Form a charging loop, the energy of which is the secondary capacitor C₁Transfer to the inductance L₁The above step (1); when switching tube S_d1Is turned off and S_d2When conducting, the DC side capacitor C₂Switch tube S_d2And a balance inductance L₁Forming a discharge loop with energy transfer direction from the inductor L₁Transfer to a capacitor C₂The above. Energy transfer dynamic balance, at which time the capacitance C₁And C₂The upper voltage forms dynamic balance, and the potential of the neutral point fluctuates in a small range.

And 3, adopting a single-neuron self-adaptive PI controller, and adopting double closed-loop control by a balance circuit control strategy. DC side capacitor voltage U_C1And U_C2The deviation value of the current loop is used as the input of the voltage loop, the output of the voltage loop after passing through the PI link is used as the input of the current loop, and then the current loop and the current i flowing into the midpoint potential are used as the input of the current loop₃The deviation is taken and is compared with the triangular wave through a PI controller and the output of a current loop to obtain two complementary PWM signals which are used for controlling two switching tubes S in a balance circuit_d1And S_d2To finally reach U_C2Real-time tracking U_C1The effect of (1).

Because the traditional PI has the problem that the parameters are not easy to set, the invention improves the traditional PI of the outer ring, introduces the single-neuron self-adaptive PI controller and improves the stability of the system. Fig. 3 is a schematic diagram of a principle of a single neuron adaptive PI controller.

Capacitor voltage U_C1、U_C2Converted into state quantity x required by neuron learning control after passing through a state converter₁、x₂Where x₁＝e(k)-e(k+1)、x₂＝e(k)，w_i(k) To correspond to x_i(k) K is the ratio of neurons, and K is the weighting factor (i ═ 1,2)Coefficient, the value of which is greater than 0.

Neurons generate control signals by associative search, i.e.

The single neuron adaptive controller realizes the adaptive and self-organizing functions by adjusting the weighting coefficient, and the adjustment of the weighting coefficient adopts a supervised Hebb learning rule which is related to the correlation function of the input, the output and the output deviation of the neuron, namely

Wherein z (k) is an output error signal, and z (k) is U_C1-U_C2η is learning Rate, η>0; c is a constant greater than 0.

In order to ensure the convergence and robustness of the above-mentioned single neuron self-adaptive PI control learning algorithm formula (6), the above-mentioned learning algorithm is normalized and processed

In the formula, η_P、η_IRespectively, integral and proportional learning rates.

The comparative example (P) and the integral (I) may respectively adopt different learning rates η_P、η_ISo that the respective weight coefficients can be respectively adjusted according to the needs, and the values of the weight coefficients are determined by simulation and experiments.

From the above analysis, the control of the single neuron adaptive control strategy is mainly focused on the voltage outer loop, and the current inner loop still adopts the traditional PI control. The control structure is simple in design and easy to realize, and meanwhile, the stability and the engineering practicability of the system are improved.

Finally, it is to be noted that: the above examples are only specific embodiments of the present invention, and are not intended to limit the technical solutions of the present invention, and the scope of the present invention is not limited thereto. Those skilled in the art will understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein.