High-Voltage DC Transmission System and Control Method Based on Fuzzy Neural Wavelet Algorithm

Abstract

The present invention relates to a high-voltage direct current (HVDC) transmission system based on a fuzzy neural wavelet algorithm, which includes a source-side synchronous motor, a source-side transformer, a source-side converter, a DC transmission line, a grid-side converter, a grid-side transformer, a grid-side synchronous motor, and a converter control circuit. The present invention also relates to a control method for a high-voltage direct current transmission system based on a fuzzy neural wavelet algorithm. This method mainly includes a fuzzy neural wavelet control method and an adaptive adjustment mechanism, and is mainly applied to a fuzzy neural wavelet controller to control the operation of the converter in combination with a converter control circuit composed of a main controller and an inner-loop controller. The present invention improves the efficiency of the neural network algorithm, enables the algorithm to more effectively identify and control non-linear systems, and ensures good convergence, fast learning ability, and strong approximation ability.

Description

Technical Field

[0001] The present invention belongs to the field of control of high - voltage direct - current (HVDC) transmission, and relates to a HVDC transmission system and a control method based on a fuzzy neural wavelet algorithm. Background Art

[0002] According to China's resource endowment and distribution, new - energy resources are mainly concentrated in the "Three - North" inland areas of Northeast, Northwest, and North China. In addition, there is a considerable amount of offshore wind - power resources in the eastern coastal areas that can be developed and utilized. For the large - scale centralized development of new - energy bases adjacent to the AC main grid in the above - mentioned areas, large - scale cross - provincial and cross - regional consumption of new - energy power is achieved through the mode of AC collection and HVDC external transmission. For long - distance and large - capacity power transmission, the high - voltage direct - current (HVDC) transmission system, due to its functions of multi - power - supply and multi - receiving - point, has been widely used in transmission projects.

[0003] However, it becomes challenging to transmit a large amount of power through the high - voltage alternating - current system and operate safely and stably over long distances to meet the growing power demand. Electromechanical low - frequency oscillation (LFO) is an important factor affecting the stable operation of the power system. The fundamental reasons for the generation of LFO are the imbalance between power demand and supply, serious faults in transmission lines, or faults in generator sets, etc.

[0004] Due to the operating characteristics of power - electronic device - based equipment, the HVDC transmission system is highly dependent on control systems and control strategies. To improve the stability of DC transmission lines, the present invention proposes an auxiliary damping controller based on fuzzy neural wavelet control (NFWC). In recent years, fuzzy neural wavelet controllers have been widely used in the control of non - linear dynamic devices. Fuzzy neural control is the combination of neural networks and fuzzy logic, aiming to combine the advantages of both and suppress their respective disadvantages. The neural network introduces its learning and computing characteristics into fuzzy logic. Therefore, the introduction of the neural network makes up for the disadvantages of fuzzy logic, and the two complement each other, giving them common advantages. However, the complexity of neural - network problems requires a large number of neurons, which reduces the efficiency of the neural network. Combining wavelets and neural networks to construct wavelet neural networks to solve these problems. Fuzzy neural wavelet networks can more effectively identify and control non - linear systems because they have good convergence, fast learning ability, and strong approximation ability, making them an ideal choice for control systems.

[0005] The present invention conducts simulation tests based on the single - machine infinite - bus (SMIB), compares the efficiency and performance of this controller with those of traditional lead - lag control (LLC) and fuzzy neural Takagi - Sugeno Kang control (NFTSKC), and calculates the quantization results of different performance indicators. Summary of the Invention

[0006] The object of the present invention is to solve the deficiencies of the prior art and provide a high-voltage direct current (HVDC) transmission system and control method based on a fuzzy neural wavelet algorithm. The aim is to propose an auxiliary damping controller using fuzzy neural wavelet control (NFMC) to suppress low-frequency oscillations in a long-distance and high-capacity HVDC power transmission system, maintain the stable operation of the power system, and ensure uninterrupted power transmission during disturbances.

[0007] The present invention realizes the solution of its technical problems through the following technical solutions:

[0008] The HVDC transmission system based on the fuzzy neural wavelet algorithm includes: a source-side synchronous motor, a source-side transformer, a source-side converter, a DC transmission line, a grid-side converter, a grid-side transformer, a grid-side synchronous motor, and a converter control circuit. The source-side converter is controlled by the converter control circuit. The converter control circuit includes a first inverting adder, a fuzzy neural wavelet controller, a main controller, an inner-loop controller of the source-side converter, an inner-loop controller of the grid-side converter, and a pulse width modulator. The positive and negative input terminals of the first inverting adder respectively input the rotor speed of the source-side synchronous motor and the set rotor speed threshold. The first inverting adder outputs a signal to the fuzzy neural wavelet controller. The fuzzy neural wavelet controller outputs a signal to the main controller. The main controller outputs a signal to the inner-loop controller of the source-side converter and the inner-loop controller of the grid-side converter. The inner-loop controllers of the source-side converter and the grid-side converter respectively output signals to the pulse width modulator in the control circuit. The pulse width modulator of the source-side converter control circuit outputs a control signal to the source-side converter, and the pulse width modulator of the grid-side converter control circuit outputs a control signal to the grid-side converter.

[0009] Further, the converter control circuit is a double-loop control unit. The converter control circuit includes a fuzzy neural wavelet controller, a main control circuit, and a current inner-loop control circuit.

[0010] Further, the fuzzy neural wavelet controller includes two branches. One branch includes a fuzzification converter for performing fuzzy control on the speed difference of the synchronous motor, a logical reasoning converter, and a defuzzification converter. The other branch includes an adaptive regulator for performing adaptive adjustment based on the speed difference of the synchronous motor and the fuzzy control result. The output terminal of the defuzzification converter is connected to the input terminal of the main controller.

[0011] Further, the main control circuit controls the operation of the controller circuit according to the set start / stop controller, and adds the input signal of the fuzzy neural wavelet controller, the set current reference signal, and the actually measured current value of the DC transmission line to obtain the current inner-loop control signals of the source-side converter and the grid-side converter.

[0012] Further, the DC current converter of the current inner loop control circuit calculates the reference value of the DC current according to the acquired DC voltage and the current control signal output by the main controller. After the current reference value and the DC current are compared by a third inverting adder, a DC deviation is obtained.

[0013] The inner loop controller includes three branches. The first branch is the current control circuit: the measured value of the DC transmission line current and the current compensation value are input to the two positive input terminals of the second inverting adder, the reference value of the DC transmission line current is input to its negative input terminal, and the output terminal is the deviation of the DC transmission line current. After passing through the first PI controller, a first angle value is output.

[0014] The second branch is the extinction angle control circuit: after the current compensation value and the current deviation value pass through the first dynamic saturation regulator, a combined value of the current compensation value and the current deviation is obtained; the combined value of the current compensation value and the current deviation is divided by the current compensation value and then multiplied by the extinction angle compensation value to obtain the minimum value of the extinction angle; the measured value of the extinction angle is input to the positive input terminal of the fourth inverting adder, and the reference value of the extinction angle and the minimum value of the extinction angle are respectively input to its two negative input terminals. The output terminal is the deviation of the extinction angle. After passing through the second PI controller, a second angle value is output.

[0015] The third branch is the DC voltage control circuit: after the current compensation value and the current deviation value pass through the second dynamic saturation regulator, a combined value of the current compensation and the current deviation is obtained. The combined value of the current compensation and the current deviation is divided by the current compensation value and then multiplied by the compensation value of the DC voltage to obtain the minimum value of the DC voltage. The measured value of the DC voltage is input to the positive input terminal of the fifth inverting adder, and the reference value of the DC voltage and the minimum value of the DC voltage are respectively input to its two negative input terminals. The output terminal is the deviation of the DC voltage. After passing through the third PI controller, a third angle value is output.

[0016] The angle values of the three branches pass through a minimum function selector to output the minimum value of the angle, and pass through a pulse width modulator to output the control signal of the converter.

[0017] The present invention also provides a high-voltage DC transmission control method based on a hybrid fuzzy neural wavelet algorithm. This method is mainly divided into a fuzzy neural wavelet control method and an adaptive adjustment mechanism, and is mainly applied to a fuzzy neural wavelet controller. The converter control circuit composed of the main controller and the inner loop controller controls the operation of the converter, and includes the following steps:

[0018] S1. Establish a multi-input single-output fuzzy neural wavelet network, and based on the rotor speed data of the synchronous motor, use the backpropagation algorithm and optimization technology to train the fuzzy neural wavelet network to obtain a fuzzy neural network prediction model for controlling the converter based on the rotor speed of the synchronous motor.

[0019] S2. Based on the fuzzy neural network prediction model in step S1, construct a fuzzy neural wavelet controller;

[0020] S3. Based on the fuzzy neural wavelet controller constructed in step S2, combine it with the current inner loop controller to generate an overall control strategy to control the converter in the HVDC transmission system.

[0021] Furthermore, the fuzzy neural wavelet controller established in S2 provides an architecture for n inputs and m rules. The fuzzy neural wavelet controller has a seven-layer structure, including an input layer, an output layer, and five hidden layers in the structure;

[0022] The first layer is the input values, that is, the deviation values between the rotor speed of the synchronous motor and the threshold, a total of n values;

[0023] In the second layer, the input values are fuzzified using the membership functions of their respective input variables. The Gaussian membership function is:

[0024]

[0025] In the formula, η nm is the Gaussian membership function, r nm represents the mean value, ζ nm represents the variance of the function, n is the nth input, and m is the mth calculation condition of the system;

[0026] In the third layer, the fuzzy system processes the rules according to the input values fuzzified in the second layer. The output of the third layer is:

[0027]

[0028] In the formula: μ n is the fuzzified value;

[0029] The rule calculated at time m is:

[0030]

[0031] Among them, x 1 , x 2 , … x n are input variables, C 1m , C 2m , C 3m … C nm are Gaussian membership functions;

[0032] The fourth layer is the subsequent part. Use a wavelet neural network instead of a polynomial or S-shaped function to sum the output values of the third layer and multiply by w n . The output of the fourth layer is as follows:

[0033]

[0034] where y n is the membership function of the wavelet neural network, w n is the output weight vector, ψ is the Mexican-Hat wavelet function, and is given as

[0035]

[0036] z is the independent variable of the Mexican-Hat wavelet function,

[0037]

[0038] In the formula, x n is the input of the wavelet, p nm is the dilation parameter of the wavelet, q nm is the translation parameter of the wavelet,

[0039] The 5th and 6th layers are the defuzzification layers, which convert the fuzzy output values back to numerical values again;

[0040] The 7th layer is the output layer, as follows:

[0041]

[0042] Furthermore, the step S1 uses the backpropagation algorithm and optimization techniques to train the fuzzy neural wavelet network, and the basic expression for training is:

[0043]

[0044] where E is the cost function value for updating the controller parameters, e = ρ ref - ρ; ρ ref is the reference value of the synchronous motor rotor speed, and ρ is the actual measured value of the synchronous motor rotor speed.

[0045] Furthermore, in order to complete the training of the fuzzy neural wavelet network more accurately and quickly, based on the basic expression for training and combined with the final training result, the cost function for updating the controller parameters is proposed as:

[0046]

[0047] Furthermore, the optimization technique is the gradient descent method, and the information is provided by the gradient vector, and its formula is as follows:

[0048] b nm(k+1) = b nm(k) - αg k

[0049] where α is the iteration step size; b nm is the update parameter vector sum; g k is the vector obtained by taking the derivative of the controller parameters;

[0050] b nm = [r nm , ζ nm , w m , p nm , q nm ;

[0051]

[0052] In the formula, r nm represents the average value; ζ nm is the variance of the function; w m is the output weight vector; p nm is the dilation parameter of the wavelet, q nm is the translation parameter of the wavelet,

[0053] Substitute the input x of the wavelet in the fuzzy neural wavelet network m , the variance ζ of the function nm , the fuzzification value μ m , the membership function y of the wavelet neural network m , the rotor speed deviation e of the synchronous motor, the Mexican - Hat wavelet function ψ m , and the independent variable of the Mexican - Hat wavelet function into the expressions of b nm and g k to obtain the expression of the gradient descent method. However, it is necessary to separately perform iterative solution and derivation for each parameter in the b nm and g k vectors. The iterative formulas for each parameter are as follows:

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] Among them,

[0060]

[0061] The advantages and beneficial effects of the present invention are:

[0062] To improve the stability of high - voltage direct - current (HVDC) transmission systems, the present invention proposes a HVDC transmission system and control method based on hybrid fuzzy neural wavelet: This system is based on the traditional HVDC transmission system and incorporates an auxiliary damping controller with a hybrid fuzzy neural wavelet algorithm based on adaptive control. It combines the advantages of neural networks, fuzzy logic, and wavelet algorithms: The neural network introduces its learning and computing characteristics into fuzzy logic and combines wavelets with the neural network, improving the efficiency of neural network problems, enabling the algorithm to more effectively identify and control non - linear systems, and ensuring good convergence, fast learning ability, and strong approximation ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is the schematic diagram of the converter control of the present invention;

[0064] Figure 2 is the inner - loop control diagram of the converter of the present invention (the inner - loop control diagrams of the source - side converter and the load - side converter are the same);

[0065] Figure 3 is the working flow chart of the converter control circuit of the present invention;

[0066] Figure 4 is the simulation test system diagram for verifying the effectiveness and superiority of the present invention;

[0067] Figure 5 is the fuzzy neural wavelet control structure diagram of the present invention;

[0068] Figure 6 shows three responses of the system of the present invention under heavy - load conditions: (a) speed deviation, (b) AC line power flow, (c) DC transmission line power flow;

[0069] Figure 7 shows three response diagrams after setting a fault on the nominal load: (a) speed deviation, (b) AC line power flow, (c) DC transmission line power flow.

[0070] Description of reference numerals in the drawings: 1 is a source-side synchronous motor, 2 is a source-side transformer, 3 is a source-side converter, 4 is a DC transmission line, 4-1 is a measuring device for the voltage and current of the source-side DC transmission line, 4-2 is a measuring device for the voltage of the grid-side DC transmission line, 5 is a grid-side converter, 6 is a grid-side transformer, 7 is a grid-side synchronous motor, 8 is a fuzzy neural wavelet controller, 9 is a first inverting adder, 10 is a fuzzification converter, 11 is a logical reasoning converter, 12 is an adaptive regulator, 13 is a defuzzification converter, 14 is a set reference current value, 15 is a main controller, 16 is a start / stop controller, 17 is an inner-loop controller of the source-side converter, 17-1 is a first filter, 17-2 is a second inverting adder, 17-3 is a first PI controller, 17-4 is a DC current converter, 17-5 is a third inverting adder, 17-6 is a first dynamic saturation regulator, 17-7 is a first multiplication / division operator, 17-8 is a fourth inverting adder, 17-9 is a second PI controller, 17-10 is a second dynamic saturation regulator, 17-11 is a second multiplication / division operator, 17-12 is a fifth inverting adder, 17-13 is a third PI controller, 17-14 is a minimum function selector, 17-15 is a pulse width modulator, 18 is an inner-loop controller of the grid-side converter, 19 is a synchronous motor group in SMIB, 20 is a transformer in SMIB, 21 is a load in SMIB, 22 is a DC transmission system in SMIB, 23 is an AC transmission system in SMIB, 24 is an infinite bus in SMIB. Detailed implementation manners

[0071] The present invention will be further described in detail below through specific embodiments. The following embodiments are only descriptive and not restrictive, and the protection scope of the present invention cannot be limited thereby.

[0072] Chinese interpretations of English terms involved in the present invention:

[0073] LCC-HVDC (Line Commutated Converter High Voltage Direct CurrentSystem): High Voltage Direct Current Transmission System Based on Line Commutated Converter

[0074] LFO (Low-Frequency Oscillation): Low-Frequency Oscillation

[0075] NFWC (Neuro-Fuzzy Wavelet control): Fuzzy Neural Wavelet Control

[0076] PI (Proportional Integra): Proportional Integral

[0077] PWM (Pulse Width Modulation): Pulse Width Modulation.

[0078] SMIB (single Machine Infinite Bus) System: Single Machine Infinite Bus System

[0079] LLC (Lead-lag control): Lead-lag control

[0080] NFTSKC (Neuro-Fuzzy Takagi Sugeno Kang Control): Neuro-Fuzzy Takagi Sugeno Kang Control

[0081] The high-voltage DC transmission system based on the fuzzy neural wavelet algorithm is innovative in that it includes a source-side synchronous motor, a source-side transformer, a source-side converter, a DC transmission line, a grid-side converter, a grid-side transformer, a grid-side synchronous motor, and a converter control circuit. The source-side converter is controlled by the converter control circuit. The converter control circuit includes a first inverting adder, a fuzzy neural wavelet controller, a main controller, an inner-loop controller of the source-side converter, an inner-loop controller of the grid-side converter, and a pulse width modulator. The positive and negative input terminals of the first inverting adder respectively input the rotor speed of the source-side synchronous motor and the set rotor speed threshold. The first inverting adder outputs a signal to the fuzzy neural wavelet controller. The fuzzy neural wavelet controller outputs a signal to the main controller. The main controller outputs a signal to the inner-loop controller of the source-side converter and the inner-loop controller of the grid-side converter. The inner-loop controllers of the source-side converter and the grid-side converter respectively output signals to the pulse width modulator in the control circuit. The pulse width modulator of the source-side converter control circuit outputs a control signal to the source-side converter, and the pulse width modulator of the grid-side converter control circuit outputs a control signal to the grid-side converter.

[0082] The converter control circuit is a double-loop control unit. The converter control circuit includes a fuzzy neural wavelet controller, a main control circuit, and a current inner-loop control circuit.

[0083] The fuzzy neural wavelet controller includes two branches. One branch includes a fuzzification converter for performing fuzzy control on the speed difference of the synchronous motor, a logical reasoning converter, and a defuzzification converter. The other branch includes an adaptive regulator for performing adaptive adjustment based on the speed difference of the synchronous motor and the fuzzy control result. The output terminal of the defuzzification converter is connected to the input terminal of the main controller.

[0084] The main control circuit controls the operation of the start / stop controller circuit according to the settings, and adds the input signal of the fuzzy neural wavelet controller, the set current reference signal, and the actually measured current value of the DC transmission line to obtain the current inner-loop control signals of the source-side converter and the grid-side converter.

[0085] The DC current converter of the current inner-loop control circuit converts the obtained DC voltage and the current control signal output by the main controller into a reference value of the DC current. After the reference value of the DC current and the DC current are compared by a third anti-summer, a DC deviation is obtained;

[0086] The inner-loop controller includes three branches. The first is the current control circuit: the measured value of the DC transmission line current and the current compensation value are input to the two positive input terminals of the second anti-summer, the reference value of the DC transmission line current is input to its negative input terminal, and the output terminal is the deviation of the DC transmission line current. After passing through the first PI controller, a first angle value is output;

[0087] The second is the arc extinction angle control circuit: after the current compensation value and the current deviation value pass through the first dynamic saturation regulator, a combined value of the current compensation value and the current deviation is obtained; after the combined value of the current compensation value and the current deviation is divided by the current compensation value and then multiplied by the arc extinction angle compensation value, a minimum value of the arc extinction angle is obtained; the measured value of the arc extinction angle is input to the positive input terminal of the fourth anti-summer, and the reference value of the arc extinction angle and the minimum value of the arc extinction angle are respectively input to its two negative input terminals, and the output terminal is the deviation of the arc extinction angle. After passing through the second PI controller, a second angle value is output;

[0088] The third branch is the DC voltage control circuit: after the current compensation value and the current deviation value pass through the second dynamic saturation regulator, a combined value of the current compensation and the current deviation is obtained. After the combined value of the current compensation and the current deviation is divided by the current compensation value and then multiplied by the compensation value of the DC voltage, a minimum value of the DC voltage is obtained. The measured value of the DC voltage is input to the positive input terminal of the fifth anti-summer, and the reference value of the DC voltage and the minimum value of the DC voltage are respectively input to its two negative input terminals, and the output terminal is the deviation of the DC voltage. After passing through the third PI controller, a third angle value is output;

[0089] The angle values of the three branches pass through the minimum function selector to output the minimum value of the angle, and pass through the pulse width modulator to output the control signal of the converter.

[0090] The present invention also provides a high-voltage DC transmission control method based on a hybrid fuzzy neural wavelet algorithm. Its innovation lies in that: this method is mainly divided into a fuzzy neural wavelet control method and an adaptive adjustment mechanism, and is mainly applied to the fuzzy neural wavelet controller, and combines the main controller and the inner-loop controller to form a converter control circuit to control the operation of the converter. It includes the following steps:

[0091] S1. Establish a multi-input single-output fuzzy neural wavelet network. Based on the rotor speed data of the synchronous motor, use the backpropagation algorithm and optimization techniques to train the fuzzy neural wavelet network, and obtain a fuzzy neural network prediction model for controlling the converter based on the rotor speed of the synchronous motor.

[0092] S2. Based on the fuzzy neural network prediction model in step S1, construct a fuzzy neural wavelet controller that can operate accurately and stably.

[0093] S3. Based on the fuzzy neural wavelet controller constructed in step S2, combine it with the current inner-loop controller to generate an overall control strategy to control the converter in the high-voltage DC transmission system.

[0094] The fuzzy neural wavelet controller established in S2 provides an architecture for n inputs and m rules. The fuzzy neural wavelet controller has a seven-layer structure, including an input layer, an output layer, and five hidden layers in the structure.

[0095] The first layer is the input values, that is, the deviation values between the rotor speed of the synchronous motor and the threshold, a total of n values.

[0096] In the second layer, the input values are fuzzified using the membership functions of their respective input variables. The Gaussian membership function is:

[0097]

[0098] where η nm is the Gaussian membership function, r nm represents the mean value, ζ nm represents the variance of the function, n is the nth input, and m is the mth calculation condition of the system.

[0099] In the third layer, the fuzzy system processes the rules according to the input values fuzzified in the second layer. The output of the third layer is:

[0100]

[0101] where: μ n is the fuzzified value;

[0102] The rule calculated at time m is:

[0103]

[0104] where x 1 , x 2 , … x n are input variables, C 1m , C 2m , C 3m … Cnm is the Gaussian membership function;

[0105] The fourth layer is the subsequent part. It uses a wavelet neural network to replace the polynomial or S-shaped function, sums the output values of the third layer, and multiplies by w n , and the output of the fourth layer is as follows:

[0106]

[0107] where y n is the membership function of the wavelet neural network, w n is the output weight vector, ψ is the Mexican-Hat wavelet function, and is given as

[0108]

[0109] z is the independent variable of the Mexican-Hat wavelet function. The meaning represented by Z in the present invention is as follows:

[0110]

[0111] In the formula, x n is the input of the wavelet, p nm is the dilation parameter of the wavelet, q nm is the translation parameter of the wavelet.

[0112] The 5th and 6th layers are the defuzzification layers, which convert the fuzzy output values back into numerical values again. The 7th layer is the output layer, as shown below:

[0113]

[0114] The step S1 uses the backpropagation algorithm and optimization technology to train the fuzzy neural wavelet network. The basic expression for training is:

[0115]

[0116] where E is the cost function value for updating the controller parameters, e = ρ ref - ρ; ρ ref is the reference value of the synchronous motor rotor speed, and ρ is the actual measured value of the synchronous motor rotor speed.

[0117] S12. In order to complete the training of the fuzzy neural wavelet network more accurately and quickly, based on the function expression in S11 and combined with the final training result, a new cost function for updating the controller parameters is proposed as:

[0118]

[0119] The optimization technique is the gradient descent method, also known as the steepest descent method. The information is provided by the gradient vector, i.e., the generalized update law, and its formula is as follows:

[0120] b nm(k+1) = b nm(k) -αg k

[0121] where α is the iteration step size; b nm is the updated parameter vector sum; g k is the vector obtained by taking the derivative of the controller parameters;

[0122] b nm = [r nm , ζ nm , w m , p nm , q nm ;

[0123]

[0124] In the formula, r nm represents the average value; ζ nm is the variance of the function; w m is the output weight vector; p nm is the dilation parameter of the wavelet, and q nm is the translation parameter of the wavelet.

[0125] In practical applications, substituting the physical quantities in the fuzzy neural wavelet network into the expressions of b nm and g k , the expression of the gradient descent method can be obtained. However, it is necessary to iteratively solve and take the derivative of each parameter in the b nm and g k vectors respectively. The iterative formulas for each parameter are as follows:

[0126]

[0127]

[0128]

[0129]

[0130]

[0131] Among them,

[0132]

[0133] The following further illustrates the structure, working process, and working principle of the present invention with an embodiment of the present invention:

[0134] The present invention conducts relevant simulations based on the Single Machine Infinite Bus (SMIB) system. The power of the synchronous motor 19 on the source side is 2100 MVA, and the output voltage is 13.8 kV. It is connected to the source-side transformer 20, which converts the voltage value of 13.8 kV to 500 kV. The source-side transformer 20 and a 250 MW load 21 are connected to the same bus. There are three lines connecting the source-side transformer 20 and the infinite bus 24. Two of them are 300 km long AC transmission lines 23 connected in a double-circuit manner, namely AC line L1 and AC line L2. The LCC-HVDC transmission line 22 is connected in parallel with the AC transmission line 23. The power capacity of the HVDC transmission line 11 is 1000 MW, and the rated voltage and current are 500 kV and 2 kA respectively. The length of the line is also 300 km, and it adopts the fuzzy neural wavelet control + current inner loop control method. The HVDC transmission line starts at 0.02 s. By placing the system in two different fault and load environments, the effectiveness of the proposed fuzzy neural wavelet control method is verified, and the performance results of the fuzzy neural wavelet control are compared with those of the conventional lead-lag control (LLC) and the fuzzy neural Takagi Sugeno Kang control (NFTSKC).

[0135] The operation of the HVDC system is as Figure 3 shown in the flowchart. After comparing the rotor speed of the synchronous motor with the set reference value of the rotor speed, it is output to the fuzzy neural wavelet controller. Then, after converting the speed deviation through the fuzzy neural wavelet control algorithm, the current value required for the main control is obtained. After comparing it with the reference current value inside the main controller, the current control signals for the source-side converter and the grid-side converter are obtained respectively. After passing through the current inner loop control, the control signals required for the converter are obtained, and then the operation of the converter is controlled.

[0136] Example 1: Comparison results under a large fault condition at rated load

[0137] The test system used for the experiment is as Figure 4 shown; by applying a three-phase fault on line L2, the performance of the proposed control scheme under nominal load is verified. This fault lasts for 5 cycles continuously and is removed after 0.1 s. The rated load power of the generator is set to Pe = 0.75 p.u.

[0138] The simulation results are shown in Figure 6. It can be observed that in the SMIB system without the HVDC line, the oscillation will continue for a quite long time. The introduction of the HVDC line 22 makes the system in a stable state. For the sake of comparison, three different auxiliary damping controllers are installed in the control of the DC power transmission system to stabilize the system more quickly. The traditional LLC controller takes the longest time to suppress the oscillation; the NFTSKC controller suppresses the oscillation in a shorter time and has better performance than the LLC. Comparing the oscillation end time, the NFTSKC controller performs 84% better than the LLC controller. The NFWC controller has the best effect in suppressing the oscillation and can suppress the oscillation in the shortest time. The time for the NFWC controller to suppress the oscillation is 86% better than that of the LLC. Figure 6(b) is the graph of the power of the AC power line L2 varying with time. Figure 6(c) is the graph of the power of the HVDC line varying with time.

[0139] Example 2: Comparison results under heavy load and large fault conditions

[0140] The robustness and effectiveness of the wavelet-based fuzzy neural wavelet controller are verified under heavy load. Under heavy load, the power of the generator is set to Pe = 1.0 p.u. At line L2, a three-phase self-clearing fault is applied at 0.1 s for 5 consecutive cycles. Under the condition of large load, when there is no HVDC line 22 in the system, it will become unstable after the fault occurs. When the HVDC line 22 is installed in the SMIB system, the operation of the system becomes stable, but the oscillation time is still very long. Figure 7(a) is the graph of the rotor speed deviation varying with time. Different auxiliary damping controllers consume different times in suppressing the oscillation. Among them, the auxiliary damping controller LLC has the worst effect; the LLC has the worst damping oscillation suppression effect; compared with the LLC, the NFWC controller has better performance and can suppress the oscillation in a shorter time. From the start of the system oscillation to its re-stabilization, the time consumed by the NFTSKC is 75.6% less than that of the LLC, and the time consumed by the NFWC is 81.5% less than that of the LLC. Figure 7(b) is the graph of the power of the AC power line L2 varying with time. Figure 7(c) is the graph of the power of the HVDC line varying with time.

[0141] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate means for implementing the functions specified in one or more flows and / or one or more blocks in the flow Figure 1 one or more flows and / or blocks Figure 1 means for implementing the functions specified in one or more blocks.

[0143] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implement the functions specified in one or more flows and / or one or more blocks in the flow Figure 1 one or more flows and / or blocks Figure 1 means for implementing the functions specified in one or more blocks.

[0144] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more flows and / or one or more blocks in the flow Figure 1 one or more flows and / or blocks Figure 1 means for implementing the functions specified in one or more blocks.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the specific implementation manners of the present invention can still be modified or equivalently replaced, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. High-voltage DC transmission system based on fuzzy neural wavelet algorithm, Characterized in that: It includes a source-side synchronous motor, a source-side transformer, a source-side converter, a DC transmission line, a grid-side converter, a grid-side transformer, a grid-side synchronous motor, and a converter control circuit. The source-side converter is controlled by the converter control circuit; the converter control circuit includes a first inverse adder, a fuzzy neural wavelet controller, a main controller, an inner-loop controller of the source-side converter, an inner-loop controller of the grid-side converter, and a pulse width modulator; the positive and negative input terminals of the first inverse adder respectively input the rotor speed of the source-side synchronous motor and the set rotor speed threshold, and the first inverse adder outputs a signal to the fuzzy neural wavelet controller; the fuzzy neural wavelet controller outputs a signal to the main controller, the main controller outputs a signal to the inner-loop controller of the source-side converter and the inner-loop controller of the grid-side converter, and the inner-loop controllers of the source-side converter and the grid-side converter respectively output signals to the pulse width modulator in the control circuit; the pulse width modulator of the source-side converter control circuit outputs a control signal to the source-side converter, and the pulse width modulator of the grid-side converter control circuit outputs a control signal to the grid-side converter; The converter control circuit is a double-closed-loop control unit; the converter control circuit includes a fuzzy neural wavelet controller, a main control circuit, and a current inner-loop control circuit; The fuzzy neural wavelet controller includes two branches; one branch includes a fuzzification converter for performing fuzzy control on the synchronous motor speed difference, a logical reasoning converter, and a defuzzification converter; the other branch includes an adaptive regulator for performing adaptive adjustment according to the synchronous motor speed difference and the fuzzy control result; the output terminal of the defuzzification converter is connected to the input terminal of the main controller.

2. The high-voltage DC transmission system based on fuzzy neural wavelet algorithm according to claim 1, Characterized in that: The main control circuit controls the operation of the start / stop controller control circuit according to the setting, and adds the input signal of the fuzzy neural wavelet controller, the set current reference signal, and the actually measured current value of the DC transmission line to obtain the current inner-loop control signals of the source-side converter and the grid-side converter.

3. The high-voltage DC transmission system based on fuzzy neural wavelet algorithm according to claim 1, Characterized in that: The DC current converter of the current inner-loop control circuit converts the obtained DC voltage and the current control signal output by the main controller into a reference value of the DC current, and the reference value of the DC current and the DC current are compared by a third inverse adder to obtain a DC deviation; The inner-loop controller includes three branches. The first is the current control circuit: the two positive input terminals of the second inverse adder input the measured value of the DC transmission line current and the current compensation value, its negative input terminal inputs the reference value of the DC transmission line current, and the output terminal is the deviation of the DC transmission line current. After passing through the first PI controller, it outputs a first angle value; The second one is the arc extinction angle control circuit: After the current compensation value and the current deviation value pass through the first dynamic saturation regulator, a combined value of the current compensation value and the current deviation is obtained; the combined value of the current compensation value and the current deviation is divided by the current compensation value and then multiplied by the arc extinction angle compensation value to obtain the minimum value of the arc extinction angle; the measured value of the arc extinction angle is input to the positive input terminal of the fourth inverting adder, and the reference value of the arc extinction angle and the minimum value of the arc extinction angle are respectively input to its two negative input terminals, and the output terminal is the deviation of the arc extinction angle. After passing through the second PI controller, the second angle value is output; The third branch is the DC voltage control circuit: After the current compensation value and the current deviation value pass through the second dynamic saturation regulator, a combined value of the current compensation and the current deviation is obtained. The combined value of the current compensation and the current deviation is divided by the current compensation value and then multiplied by the compensation value of the DC voltage to obtain the minimum value of the DC voltage. The measured value of the DC voltage is input to the positive input terminal of the fifth inverting adder, and the reference value of the DC voltage and the minimum value of the DC voltage are respectively input to its two negative input terminals. The output terminal is the deviation of the DC voltage. After passing through the third PI controller, the third angle value is output; The angle values of the three branches pass through the minimum function selector to output the minimum value of the angle, and pass through the pulse width modulator to output the control signal of the converter.

4. A high-voltage DC transmission control method based on a hybrid fuzzy neural wavelet algorithm, characterized in that: This method is divided into a fuzzy neural wavelet control method and an adaptive adjustment mechanism, which are applied to a fuzzy neural wavelet controller and combined with a main controller and an inner-loop controller to form a converter control circuit to control the operation of the converter. It includes the following steps: S1. Establish a multi-input single-output fuzzy neural wavelet network. Based on the rotor speed data of the synchronous motor, use the backpropagation algorithm and optimization technology to train the fuzzy neural wavelet network to obtain a fuzzy neural network prediction model for controlling the converter based on the rotor speed of the synchronous motor; S2. Based on the fuzzy neural network prediction model in step S1, construct a fuzzy neural wavelet controller; S3. Based on the fuzzy neural wavelet controller constructed in step S2, combine it with the current inner-loop controller to generate an overall control strategy to control the converter in the high-voltage DC transmission system; The fuzzy neural wavelet controller established in S2 provides a framework for n inputs and m rules. The fuzzy neural wavelet controller has a seven-layer structure, including an input layer, an output layer, and five hidden layers in the structure; The first layer is the input value, that is, the deviation value between the rotor speed of the synchronous motor and the threshold, with a total of n values; In the second layer, the input values are fuzzified using the membership functions of their respective input variables. The Gaussian membership function is: where η nm is the Gaussian membership function, r nm represents the mean value, ζ nm represents the variance of the function, n is the nth input, and m is the mth calculation condition of the system; In the third layer, the fuzzy system processes the rules according to the input values fuzzified in the second layer. The output of the third layer is: where: μ n is the fuzzification value; The rule calculated at time m is: Among them, x 1 , x 2 , … x n are input variables, and C 1m , C 2m , C 3m … C nm are Gaussian membership functions; The fourth layer is the subsequent part. Using a wavelet neural network to replace the polynomial or sigmoid function, the output values of the third layer are summed and multiplied by w n , and the output of the fourth layer is as follows: where y n is the membership function of the wavelet neural network, w n is the output weight vector, ψ is the Mexican-Hat wavelet function, and is given as z is the independent variable of the Mexican-Hat wavelet function, where x n is the input of the wavelet, p nm is the dilation parameter of the wavelet, q nm is the translation parameter of the wavelet, The fifth and sixth layers are the defuzzification layers, which convert the fuzzy output values back into numerical values again; The seventh layer is the output layer, as follows:

5. The HVDC transmission control method based on the hybrid fuzzy neural wavelet algorithm according to claim 4, characterized in that: in step S1, the backpropagation algorithm and optimization technology are used to train the fuzzy neural wavelet network, and the basic expression for training is: where E is the cost function value for updating the controller parameters, e = ρ ref - ρ; ρ ref is the reference value of the synchronous motor rotor speed, and ρ is the actual measured value of the synchronous motor rotor speed.

6. The HVDC transmission control method based on the hybrid fuzzy neural wavelet algorithm according to claim 5, characterized in that: in order to complete the training of the fuzzy neural wavelet network more accurately and quickly, based on the basic expression for training and combined with the final result of training, a cost function for updating the controller parameters is proposed as: 。 7. The HVDC transmission control method based on the hybrid fuzzy neural wavelet algorithm according to claim 4, characterized in that: the optimization technology is the gradient descent method, and the information is provided by the gradient vector, and its formula is as follows: b nm(k+1) = b nm(k) - αg k where α is the iteration step size; b nm is the sum of updated parameter vectors; g k is the vector obtained by taking the derivative of the controller parameters; b nm = [r nm , ζ nm , w m , p nm , q nm ; where r nm represents the average value; ζ nm is the variance of the function; w m is the output weight vector; p nm is the dilation parameter of the wavelet, q nm is the translation parameter of the wavelet; The input x of the wavelet in the fuzzy neural wavelet network m , the variance ζ of the function nm , the fuzzification value μ m , the membership function y of the wavelet neural network m , the rotor speed deviation e of the synchronous motor, the Mexican-Hat wavelet function ψ m , substitute the independent variable z of the Mexican-Hat wavelet function into b nm and g k in the expression, and obtain the expression of the gradient descent method. However, it is necessary to separately iterate and solve the derivatives of each parameter in the b nm and g k vectors. The iteration formulas for each parameter are as follows: wherein,

Images