Fuzzy PI control method for BP neural network optimized totem pole PFC circuit

Abstract

The invention provides a fuzzy PI control method of a totem-pole PFC circuit optimized by a BP neural network. The method comprises the following steps: constructing the totem-pole PFC circuit; and constructing a voltage and current sampling circuit. A voltage loop and current loop double closed loop controller is constructed, the voltage loop controller adopts a fuzzy PI controller optimized by a BP neural network, and the integral rate of the fuzzy PI controller is optimized through the BP neural network. In a current loop controller, output of a voltage loop controller is multiplied by output of a phase-locked loop, obtained reference current iref is subtracted from inductive current iL, a duty ratio is obtained through a current loop PI controller, a duty ratio signal is sent into a PWM generator to generate a PWM signal, the PWM signal is used for controlling on and off of two GaN switching tubes, input current tracks an input voltage phase, and the phase-locked loop is switched off. Therefore, the purpose of power factor correction is achieved. According to the invention, the stability and response speed of the totem-pole bridgeless PFC circuit are improved.

Description

Technical Field

[0001] The present invention relates to a fuzzy PI control method for a totem-pole PFC circuit optimized by a BP neural network, belonging to the field of power factor correction circuits. Background Art

[0002] A totem-pole bridgeless power factor correction (PFC) circuit controlled by an average current mode and using gallium nitride switching devices operating in continuous conduction mode (CCM) can compare a reference voltage V ref with an actual output bus voltage V bus . The difference is processed by a voltage controller and the output is denoted as A. The AC input voltage V ac is divided by its maximum value to obtain B. A and B are multiplied by a multiplier to obtain a reference value i ref of the input current. The actual value of the inductor current is compared with the reference current i ref . The difference is limited between 0 and 1 after passing through a current controller to obtain a duty cycle D. A pulse width modulation (PWM) controller calculates according to the duty cycle D, and finally generates a PWM signal, thereby controlling the on and off of power devices.

[0003] In recent years, people have enhanced their awareness of environmental protection and advocated a sustainable lifestyle, which has led to a desire to reduce electronic waste. This requires the same switching power supply to be able to supply multiple different electrical devices such as mobile phones, laptops, and electric vehicles with a wide range of voltage and power differences.

[0004] With a large difference in voltage range and power range, the PFC circuit, which is the front stage of a two-stage AC / DC converter, is required to have better stability and response speed. The control of the totem-pole PFC circuit is through a double closed-loop control of a current loop and a voltage loop. Through the double closed-loop control, the input current follows the phase of the input voltage to achieve the purpose of power factor correction.

[0005] For voltage loop control, the currently commonly used method is a PI controller. However, the parameter design of the PI controller is usually fixed, which makes it difficult to flexibly adjust the control effect according to the current different states. In this way, when the load changes greatly, its stability and response speed are relatively poor, and it is not very suitable for switching power supplies with a large difference in voltage range and power range.

[0006] The current loop of the present invention adopts the average current mode, and the voltage loop adopts a fuzzy PI controller optimized by a BP neural network. Aiming at the above problems, the fuzzy PI controller flexibly adjusts the parameters of the PI controller according to different situations, and further uses a BP neural network to optimize the integral rate. Through this controller, the stability can be enhanced and the response speed can be improved when facing load changes, effectively improving the problems existing in the above PI controller. Summary of the Invention

[0007] The object of the present invention is to improve the problems of poor stability and slow response speed of the totem-pole bridgeless PFC circuit for load changes in a wide output voltage range and a wide output power range.

[0008] To improve the above problems, the present invention proposes a fuzzy PI control method for a totem-pole PFC circuit optimized by a BP neural network. This control method fuzzifies the error e of the DC bus voltage and the load power error ep through fuzzy control, performs fuzzy reasoning according to fuzzy rules, and finally defuzzifies to realize the proportional coefficient K of the voltage loop p and the integral coefficient K of the voltage loop i for real-time adjustment, and optimizes the integral rate of the fuzzy PI controller through a BP neural network to realize the real-time adjustment of the integral rate i rate Through this control method, the control stability and response speed of the totem-pole PFC circuit can be significantly improved.

[0009] The technical solution of the present invention:

[0010] A control method for a gallium nitride totem-pole PFC circuit with a wide output range includes the following steps:

[0011] S1: Construct a totem-pole PFC circuit, including a 220V AC power supply, an inductor, two GaN switching tubes, two MOSFET transistors, and an output capacitor.

[0012] S2: Construct a sampling circuit for voltage and current, including an inductor current sampling circuit, an input voltage sampling circuit, an output bus voltage sampling circuit, a load current sampling circuit, and a load voltage sampling circuit.

[0013] S3: Construct a double closed-loop controller for the voltage loop and the current loop. The voltage loop controller adopts a fuzzy PI controller optimized by a BP neural network, and optimizes the integral rate of the fuzzy PI controller through a BP neural network.

[0014] S4: In the current loop controller, multiply the output of the voltage loop controller by the output of the phase-locked loop to obtain the reference current i ref and the inductor current i LTake the difference, and then obtain the duty cycle through the current loop PI controller. Send the duty cycle signal into the PWM generator to generate a PWM signal, which is used to control the on and off of two GaN switching tubes, so that the input current tracks the phase of the input voltage, thereby achieving the purpose of power factor correction.

[0015] Preferably, the S3 specifically includes:

[0016] S31: Establish double closed-loop control of the voltage loop and the current loop;

[0017] S32: The voltage loop sampling fuzzy PI controller includes: According to the expected value V of the DC bus voltage ref and the actual value V of the DC bus voltage bus to compare and obtain the DC bus voltage error e. According to the expected value V of the load voltage dcref and the actual value V of the load voltage L as well as the actual value i of the load current o to obtain the expected value P of the output power ref and the error ep of the actual value P out Take e and ep as the inputs of the fuzzy PI controller, fuzzify the two, and determine the fuzzy output quantity; Design the fuzzy set, universe of discourse, membership function according to the fuzzy control quantity, select the fuzzy inference method and the defuzzification method; Take the expected value P of the output power ref and the actual value P out and the voltage error e as the inputs of the BP neural network. Each hidden layer neuron performs weighted summation on the inputs of the BP neural network and adds a bias, applies an activation function to the result to introduce nonlinearity, and obtains the hidden layer output. Take the hidden layer output as the input, repeat the process of weighted summation and adding bias, and apply the activation function again to obtain the final result. Optimize the voltage loop integral rate i of the fuzzy PI controller through the BP neural network rate .

[0018] S33: Use the voltage loop proportional coefficient K p and the voltage loop integral coefficient K i obtained by the fuzzy PI controller, and the voltage loop integral rate i rate obtained by the BP neural network to adjust the voltage loop.

[0019] Preferably, the fuzzy state universe of discourse of the input and output quantities in the S32 is divided into 7 levels. The fuzzy set of the fuzzy PI controller includes {NB, NM, NS, ZO, PS, PM, PB}, where NB represents negative large, NM represents negative medium, NS represents negative small, ZO represents zero, PS represents positive small, PM represents positive medium, and PB represents positive large. Different fuzzy sets represent different error degrees.

[0020] Preferably, the membership function adopts the triangular membership function trimf and the trapezoidal membership function trapmf.

[0021] Preferably, the fuzzy inference method adopts the Mamdani algorithm.

[0022] Preferably, the defuzzification method is the centroid method.

[0023] Preferably, the BP neural network is divided into an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is 3, corresponding to the error of the DC bus voltage, the actual value of the output power, and the expected value of the output power respectively; the number of hidden layers is two, and each layer has 12 nodes; the number of nodes in the output layer is 1, corresponding to the integration rate.

[0024] The beneficial effects of the present invention: The present invention has good stability and response speed for the change of load power. When the load changes, it effectively reduces the change amount of the DC bus voltage and speeds up the speed of the DC bus voltage returning to the stable state. Description of the Drawings

[0025] Figure 1 It is a schematic structural diagram of the totem-pole bridgeless PFC circuit of the present invention and a schematic control flow diagram of the control system.

[0026] Figure 2 It is a schematic specific calculation flow diagram of the fuzzy PI control method for the totem-pole PFC circuit optimized by the BP neural network of the present invention.

[0027] Figure 3 It is a schematic diagram of the BP neural network structure of the present invention.

[0028] Figure 4 It is a simulation diagram comparing the change of the DC bus voltage between the fuzzy PI control method and the ordinary PI control method for the totem-pole PFC circuit optimized by the BP neural network of the present invention when the load changes greatly.

[0029] Figure 5 It is a THD value diagram of the circuit of the fuzzy PI control method for the totem-pole PFC circuit optimized by the BP neural network of the present invention.

[0030] In the figure: 11 GaN switch; 12 Si MOSFET switch; 13 4 mH inductor; 14 200 uF capacitor; 15 220 V input AC power supply; 21 voltage loop fuzzy PI controller; 22 current loop fuzzy PI controller; 23 subtractor one; 24 subtractor two; 25 multiplier one; 26 PWM generator; 27 power calculation module; 28 phase-locked loop; 29 BP neural network; 41 subtractor three; 42 scaling factor; 43 quantization factor A; 44 three-input multiplication module; 45 quantization factor B; 46 amplitude limiting; 47 fuzzy module; 48 quantization factor C; 49 output one of the fuzzy module; 50 output two of the fuzzy module; 51 proportionality factor A; 52 proportionality factor B; 53 multiplier two; 54 multiplier three; 55 original proportionality coefficient; 56 original integral coefficient; 57 adder one; 58 adder two; 71 input layer; 72 first hidden layer; 73 second hidden layer; 74 output layer; 81 output voltage waveform of the double closed-loop control; 82 output voltage waveform of the fuzzy PI control optimized by the BP neural network. Detailed implementation manners

[0031] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the present invention will be further described below through embodiments in conjunction with the accompanying drawings.

[0032] In a specific application example, in the totem pole PFC circuit with the above control method, the parameter settings of each functional module and electronic component are as follows:

[0033] AC input voltage: 190 - 240 VAC, power frequency 50 Hz;

[0034] Output voltage: 400 VDC;

[0035] Output power level: 500 W;

[0036] Inductor: 4 mH;

[0037] Electrolytic capacitor: 200 μF;

[0038] Two GaN switches: enhancement type, 22 A @ 25 °C, 650 V, switching frequency 100 kHz;

[0039] Two power Si MOSFETs: 18 A @ 25 °C, 650 V, switching frequency is the power frequency.

[0040] As Figure 1 shown, it is the totem pole PFC circuit of the present invention. The circuit is composed of two GaN switches 11, two Si MOSFET switches 12, a 4 mH inductor 13, a 200 uF capacitor 14, and a 220 V input AC power supply 15.

[0041] The control module consists of a voltage-loop fuzzy PI controller 21 and a current-loop fuzzy PI controller 22.

[0042] Sample the actual value V of the DC bus voltage bus , and the expected value V of the DC bus voltage ref . Through subtractor 23, calculate the DC bus voltage error value e. After being adjusted by the voltage loop 21, the output of the voltage loop 21 is multiplied by the output of the phase-locked loop 28 through multiplier 25 as the reference current i ref , the reference current i ref and the inductor current i L . Through subtractor 24, calculate the difference. After being adjusted by the current loop 22, the output of the current loop 22 is used as the duty cycle, and the duty cycle is used to control the GaN switch through the PWM generator 26.

[0043] The fuzzy controller needs to calculate the expected value P and the actual value P of the output power by the power calculation module 27 according to the expected value V of the load voltage dcref , the actual value V of the load voltage L , and the actual value i of the load current o . Then subtract the two to get the power error ep. As ref shown, the voltage error e is obtained by subtractor 41. The voltage error e is first multiplied by the scaling factor 42, then passed through the quantization factor K out 43, and then passed through the quantization function y = x e1 processing. The quantization function is implemented by the three-input multiplier module 44, and then passed through the quantization factor K 3 45. Finally, through the limiter 46, it is sent to the fuzzy module 47 for fuzzification; the power error ep passes through the quantization factor K e2 48, and then through the limiter 46, it is sent to the fuzzy module 47 for fuzzification. ep

[0044] The outputs F out1 49, F out2 50 of the fuzzy module 47 are respectively multiplied by the proportionality factor K up 51, the proportionality factor K ui 52 to obtain the proportional adjustment coefficient dK p and the integral adjustment coefficient dK i . The proportional adjustment coefficient dK p is multiplied by the original proportionality coefficient K p0 55 through multiplier 53 to obtain the proportionality coefficient change ΔK p i , and the integral adjustment coefficient dK i0 is multiplied by the original integral coefficient K i 56 through multiplier 54 to obtain the integral coefficient change ΔK p . The proportionality coefficient change ΔK​​p After adder 1 57 and the original proportionality coefficient K p0 are added to 55 to obtain the actual proportionality coefficient K p , the integral adjustment coefficient dK i After adder 2 58 and the original integral coefficient K i0 are added to 56 to obtain the actual integral coefficient K i .

[0045] The above process is expressed by the following formula:

[0046] K p = K p0 + ΔK p = K p0 + K p0 ·dK p = K p0 + K p0 ·K up ·F out1

[0047] K i = K i0 + ΔK i = K i0 + K i0 ·dK i = K i0 + K i0 ·K ui ·F out2

[0048] According to e and ep, and the relationship between F out1 、F out2 the following fuzzy rule table is established:

[0049] F out1 's fuzzy rule table is:

[0050]

[0051] F out2 's fuzzy rule table is:

[0052]

[0053] Use the BP neural network 29 to dynamically adjust the integral rate of the integrator of the fuzzy PI controller, so as to achieve the purpose of further accelerating the speed of restoring stability and reducing the overshoot by flexibly adjusting the integral rate, that is, improving the response speed and enhancing the stability.

[0054] The structure of the BP neural network used in the present invention is as Figure 3As shown. The BP neural network is divided into an input layer 71, a first hidden layer 72, a second hidden layer 73, and an output layer 74. The number of nodes in the input layer is 3, corresponding to the error e of the DC bus voltage, the actual value P of the output power out , and the expected value P ref of the output power; the number of hidden layers is two, with 12 nodes in each layer; the number of nodes in the output layer is 1, corresponding to the integration rate i rate .

[0055] An input vector X is composed of the three inputs of the BP neural network. The input of each neuron in the first hidden layer is the weighted sum of the weights and the inputs:

[0056] z (1) =W (1) X + b (1)

[0057] Among them, W (1) is the weight matrix from the input layer to the first hidden layer, and b (1) is the bias of the first hidden layer. Then, it is processed through the activation function to obtain the output of the first hidden layer:

[0058] a (1) =σ (1) (z (1) )

[0059] Among them, σ (1) (x) represents the activation function of the first hidden layer.

[0060] The input of each neuron in the second hidden layer is the weighted sum of the weights and the output of the first hidden layer:

[0061] z (2) =W (2) a (1) + b (2)

[0062] Among them, W (2) is the weight matrix from the first hidden layer to the second hidden layer, and b (2) is the bias of the second hidden layer. Then, it is processed through the activation function to obtain the output of the second hidden layer:

[0063] a (2) =σ (2) (z (2) )

[0064] Among them, σ (2) (x) represents the activation function of the second hidden layer.

[0065] The input of each neuron in the output layer is the weighted sum of the weights and the output of the second hidden layer:

[0066] z (3) =W(3) a (2) +b (3)

[0067] wherein, W (3) is the weight matrix from the second hidden layer to the output layer, and b (3) is the bias of the output layer. Then, through the activation function, the output result of the output layer is obtained, that is, the final result integration rate i of the neural network rate :

[0068] i rate =σ (3) (z (3) )

[0069] wherein, σ (3) (x) represents the activation function of the output layer

[0070] The activation function used in the first hidden layer 72 of the BP neural network is the S-shaped logarithmic function logsig, the activation function used in the second hidden layer 73 is the hyperbolic tangent S-shaped function tansig, and the activation function used in the output layer is the linear function purelin

[0071] As Figure 4 shown, it is the waveform diagram of the output voltage when the load changes by 50%, which includes the output voltage waveform 81 under double closed-loop control and the output voltage waveform 82 under the fuzzy PI control of the totem pole PFC circuit optimized by the BP neural network. The minimum voltage under double closed-loop control is 337.4V, the difference from the stable state is 34.7V, the adjustment time is 0.213s, and the voltage change amplitude is 54.9V; the minimum voltage under the fuzzy PI control of the totem pole PFC circuit optimized by the BP neural network is 360.0V, the difference from the stable state is 12.1V, the adjustment time is 0.149s, and the voltage change amplitude is 28.4V

[0072] According to the above comparison results, it can be seen that the fuzzy PI control of the totem pole PFC circuit optimized by the BP neural network proposed by the present invention has better stability and response speed, the output voltage change is smaller and more stable, and the time to return to stability is shorter

[0073] Through the action of the voltage loop and the current loop, the present invention enables the input current to follow the phase of the input voltage well, so as to achieve the purpose of power factor correction. As Figure 5 shown, it is the simulation diagram of the input current harmonics obtained by fast Fourier transform

[0074] The calculation formula of the total harmonic distortion THD is

[0075]

[0076] wherein, In is the amplitude of the nth harmonic component in the input current.

[0077] From Figure 5 it can be seen that the total harmonic distortion THD of the input current is 3.29%.

[0078] The relationship formula between the power factor and THD is:

[0079]

[0080] where cosα 1 is the displacement factor.

[0081] According to the above calculation formula, it can be obtained that the power factor is greater than 0.99.

[0082] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.

Claims

1. A fuzzy PI control method for a totem pole PFC circuit optimized by a BP neural network, characterized in that: The following steps are involved: S1: Build a totem pole PFC circuit, including a 220V AC power supply, an inductor, two GaN switches, two MOSFET transistors, and an output capacitor; S2: construct voltage and current sampling circuits, including an inductor current sampling circuit, an input voltage sampling circuit, an output bus voltage sampling circuit, a load current sampling circuit, and a load voltage sampling circuit; S3: Construct a dual closed-loop controller of voltage loop and current loop, where the voltage loop controller adopts a fuzzy PI controller optimized by BP neural network, and the integral rate of the fuzzy PI controller is optimized by BP neural network; S4: In the current loop controller, the output of the voltage loop controller is multiplied by the output of the phase-locked loop to obtain the reference current i ref With the inductor current i L The duty cycle is obtained by making a difference, and then the duty cycle signal is sent to the PWM generator to generate a PWM signal, which is used to control the opening and closing of the two GaN switch tubes so that the input current tracks the input voltage phase, thereby achieving the purpose of power factor correction.

2. The fuzzy PI control method of a totem pole PFC circuit optimized by a BP neural network according to claim 1, characterized in that: The S3 specifically includes: S31: Establish dual closed-loop control of voltage loop and current loop; S32: voltage loop sampling fuzzy PI controller, including: according to the expected value of DC bus voltage V ref And the actual value of DC bus voltage V bus By comparison, the DC bus voltage error e is obtained. According to the expected load voltage V dcref and the actual value of the load voltage V L And the actual value of the load current i o Get the expected value of output power P ref and the actual value P out The error ep is taken as the input of the fuzzy PI controller, and the two are fuzzified to determine the fuzzy output; according to the fuzzy control quantity, the fuzzy set, domain, and membership function are designed, and the fuzzy reasoning method and defuzzification method are selected; the expected value of the output power P ref , actual value P out , the voltage error e is used as the input of the BP neural network, each hidden layer neuron performs weighted summation on the input of the BP neural network and adds a bias, applies the activation function to the result, introduces nonlinearity, obtains the hidden layer output, uses the hidden layer output as the input, repeats the weighted summation and biasing process, applies the activation function again, obtains the final result, and optimizes the voltage loop integral rate i of the fuzzy PI controller through the BP neural network rate ; S33: Voltage loop proportional coefficient K obtained by fuzzy PI controller p , voltage loop integration coefficient K i , the voltage loop integration rate i obtained by BP neural network rate Adjust the voltage loop.

3. The fuzzy PI control method of a totem pole PFC circuit optimized by a BP neural network according to claim 1, characterized in that: The fuzzy state domain of the input and output quantities in the S32 is divided into 7 levels, and the fuzzy sets of the fuzzy PI controller include {NB, NM, NS, ZO, PS, PM, PB}, where NB represents large negative, NM represents medium negative, NS represents small negative, ZO represents zero, PS represents small positive, PM represents medium positive, and PB represents large positive; different fuzzy sets represent different degrees of error.

4. The fuzzy PI control method of a totem pole PFC circuit optimized by a BP neural network according to claim 1, characterized in that: The membership function in S32 adopts a triangle membership function trimf and a trapezoidal membership function trapmf.

5. The fuzzy PI control method of a totem pole PFC circuit optimized by a BP neural network according to claim 1, characterized in that: The fuzzy reasoning method in S32 adopts the Mamdani algorithm.

6. The fuzzy PI control method of a totem pole PFC circuit optimized by a BP neural network according to claim 1, characterized in that: The defuzzification method in S32 is the centroid method.

7. The fuzzy PI control method of a totem pole PFC circuit optimized by a BP neural network according to claim 1, characterized in that: The BP neural network in S32 is divided into an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is 3, corresponding to the error of the DC bus voltage, the actual value of the output power, and the expected value of the output power respectively; the number of hidden layers is two, each with 12 nodes; the number of nodes in the output layer is 1, corresponding to the integration rate.

Images