CN111856185A - Embedded system and method for wind turbine power quality monitoring and improvement - Google Patents

Abstract

The invention relates to an embedded system for monitoring and improving the electric energy quality of a wind turbine generator, which comprises a main control module, a signal acquisition module, a frequency detection module, a power control module, a communication module, an upper computer, a storage module and a wind speed sensor, wherein one end of the signal acquisition module, one end of the frequency detection module and one end of the power control module are respectively connected with the existing wind turbine generator, the other end of the signal acquisition module, one end of the frequency detection module and one end of the power control module are respectively connected with the main control module, the storage module and the wind speed sensor are respectively connected with the main control module, and the. The invention also relates to a method for monitoring and improving the electric energy quality of the wind turbine generator, which is embedded in the embedded system and used for monitoring and improving the electric energy quality of the wind turbine generator. Compared with the prior art, the method has the advantages of high reliability, good real-time performance, realization of visualization of power quality monitoring, strong practicability and the like.

Description

Embedded system and method for wind turbine power quality monitoring and improvement

technical field

The invention relates to the technical field of wind power generation, in particular to an embedded system and method for monitoring and improving the power quality of wind turbines.

Background technique

As a kind of clean energy, wind energy has a very broad application prospect. However, due to the extremely poor stability of wind energy and the rapid and random changes in speed and direction, how to effectively detect the voltage flicker, power instability, Power quality problems such as harmonic pollution, and how to quickly and accurately control the wind power system to adjust and improve power quality are the primary problems to be solved in wind power research. Therefore, in recent years, research on power quality monitoring of wind turbines has become very popular, such as using PLC technology. , sensor technology, microcontroller technology, etc.

Using the existing technology, such as 16-bit single-chip microcomputer, to design the monitoring, improvement and visualization of the power quality of wind turbines has the disadvantages of limited data operation and analysis capabilities, and low security and real-time performance of the system. With the continuous improvement of data volume calculation of modern power equipment, including real-time requirements, these data processing devices have been unable to meet the power requirements in calculation, and cannot guarantee the reliability of power quality monitoring of wind turbines, resulting in poor overall data processing effect of the power system. It affects the data analysis and the final monitoring effect, and cannot effectively adjust and improve the power quality, so as to cause impact and fluctuations on the power grid, resulting in inestimable losses. The research on the power quality of wind turbines in the existing technology is also ongoing For example, Chinese patent CN110865259A discloses a method and device for evaluating the power quality of a wind farm. Although the method and device in this patent can evaluate the power quality, it cannot adjust and improve the power quality.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an embedded system for monitoring and improving the power quality of wind turbines with high reliability, realizing visualization of power quality monitoring, strong practicability and good real-time performance in order to overcome the above-mentioned defects of the prior art. and method.

The object of the present invention can be realized through the following technical solutions:

An embedded system for monitoring and improving the power quality of wind turbines, the system is connected with existing wind turbines, the embedded system includes: a main control module, a signal acquisition module, a frequency detection module, a power control module, a communication module module, host computer, storage module and wind speed sensor; one end of the signal acquisition module, the frequency detection module and the power control module are respectively connected with the existing wind turbine, and the other end is respectively connected with the main control module; the storage module and The wind speed sensors are respectively connected with the main control module; the main control module communicates with the upper computer through the communication module.

Preferably, the main control module is an FPGA chip.

Preferably, the signal acquisition module includes a voltage sensor, a current sensor, a signal conditioning circuit, and a voltage and current sampling circuit; one end of the voltage sensor and the current sensor are respectively connected with the existing wind turbine, and the other ends are respectively connected with the signal conditioning circuit The output end of the signal conditioning circuit is connected to the input end of the voltage and current sampling circuit; the output end of the voltage and current sampling circuit is connected to the main control module.

Preferably, the frequency detection module includes an isolation transformer, a filter shaping circuit, a Schmitt trigger, a double D trigger and an inverter connected in sequence; the primary side of the isolation transformer is connected to the existing wind turbine; The output end of the inverter is connected with the main control module.

Preferably, the power control module includes a grid-side and rotor-side PWM converter, an SVPWM inverter control unit, an IGBT drive unit, and an FPGA main control unit that are connected in sequence; the grid-side and rotor-side PWM converters The input end is respectively connected with the existing power grid and the generator of the existing wind turbine; the output end of the IGBT drive unit is connected with the main control module.

A method for monitoring and improving the power quality of a wind turbine for the above-mentioned embedded system, comprising:

Step 1: According to the wind speed and wind direction samples uploaded by the main control module, use the wind speed prediction sub-method to obtain the predicted wind speed;

Step 2: Obtain the voltage and current signals on the rotor and stator sides of the existing wind turbine, and obtain the frequency through the frequency detection module;

Step 3: Use the harmonic distortion analysis sub-method and the transient disturbance monitoring sub-method to analyze the power quality output by the existing wind turbines at the current moment;

Step 4: Obtain the rotational speed data of the existing wind turbine generator rotor;

Step 5: Calculate the power of the generators in the current existing wind turbine;

Step 6: Determine whether the current power of the generator meets the load demand, if so, go to Step 8, otherwise, go to Step 7;

Step 7: use the power control sub-method for power control, and then perform step 8;

Step 8: Obtain wind speed forecast data;

Step 9: Determine whether the absorbed power of the unit needs to be adjusted, if so, adjust the absorbed power of the unit, and then execute step 10, otherwise, directly execute step 10;

Step 10: End the current round of calculation, re-acquire the wind speed data at the current moment, and determine whether the difference between the wind speed and the wind speed forecast data at the previous moment is greater than the preset error, if so, return to step 1, and then perform the next round of calculation, otherwise, return Step 2, perform the next round of calculation.

Preferably, the wind speed prediction sub-method includes:

Step 1-1: Acquire the voltage and current of the stator side output terminal of the existing wind turbine generator and the voltage and current signals of the rotor side collected by the signal acquisition module, and simultaneously acquire the wind speed sample data collected by the wind speed sensor;

Step 1-2: Normalize the data to obtain a sample training set;

The normalization process is specifically:

Among them, v _i is the original value of the wind speed sample data; v _max and v _min are the maximum and minimum values in the wind speed sample data, respectively; v _i ' is the normalized output value of the wind speed;

Step 1-3: Initialize the improved PSO algorithm;

Step 1-4: Calculate the initial fitness, and the calculation method of the fitness is:

Among them, N is the sample size; v _i ^* is the predicted value of wind speed;

Steps 1-5: Perform iterative optimization to update particle velocity and position. The specific methods are:

和

分别为D维上的第i个粒子进行k+1次迭代的位置和速度；z为收缩因子；φ为总加速因子；g_en为总迭代次数；K为收缩系数；Among them, d=1,2,3,...,D; i=1,2,3,...,N; k is the current iteration number; v _id is the current speed; c ₁ and c ₂ are acceleration factors, and c ₁ and c ₂ are both greater than zero; r ₁ and r ₂ are random functions, and the value range is [0, 1];

and

are the position and velocity of the ith particle in the D dimension for k+1 iterations; z is the shrinkage factor; φ is the total acceleration factor; _gen is the total number of iterations; K is the shrinkage coefficient;

Steps 1-6: Calculate fitness, update individual and global best fitness and optimal solution;

Step 1-7: Determine whether the fitness meets the requirements or whether the number of iterations reaches the maximum number of iterations, if so, execute steps 1-8, otherwise, return to steps 1-5;

Steps 1-8: Use the global optimal solution as the parameter of the least squares support vector machine LSSVM, train the LSSVM, and obtain wind speed prediction data.

Preferably, the harmonic distortion analysis sub-method includes:

Step 2-1: Use the FIR filter to de-noise the voltage and current signals on the rotor and stator sides;

Step 2-2: Input data in reverse order, and output the data obtained by using the windowed FFT method in natural order;

Step 2-3: Calculate the total harmonic distortion rate, the calculation method is:

Among them, U _n is the effective value of the nth harmonic voltage; In is the effective value of the _nth harmonic current; THD _u is the total harmonic distortion rate of voltage; THD _i is the total harmonic distortion rate of current; U ₁ is the rms value of the fundamental wave voltage; I ₁ is the rms value of the fundamental wave current;

Step 2-4: Determine whether the total harmonic distortion rate exceeds the preset value, if so, cut the existing wind turbines out of the grid, otherwise, end the current cycle.

Preferably, the transient disturbance monitoring sub-method includes:

Step 3-1: Determine whether the voltage and current signals on the rotor and stator sides obtained in Step 2 exceed the preset values, if so, cut the fan out of the grid to end the current cycle, otherwise, go to Step 3-2;

Step 3-2: Use the FIR filter to de-noise the voltage and current signal data on the stator side;

Step 3-3: Use the Hilbert-Huang transform detection algorithm HHT to analyze the data obtained in step 3-2. The HHT algorithm includes two parts, the EMD and the Hilbert spectrum analysis method. First, the given signal is decomposed into several natural modes using the EMD method. state function IMF, and then perform Hilbert transform on each IMF, obtain the characteristic parameters, and obtain the Hilbert spectrum of the corresponding original signal. The specific steps are:

Decompose the empirical mode to get the intrinsic mode function group:

Among them, s _i (t) is the natural modal component in the original signal; d(t) is the DC component in the original signal; r(t) is the residual function data sequence obtained by decomposition;

Then perform Hilbert transform on all IMFs to obtain the analytical expression of each IMF transformed:

Among them, a _i is the IMF amplitude, and the above formula is the Hilbert spectrum, which can be written as:

Integrate the above formula in the time domain to obtain the corresponding Hilbert marginal spectrum:

where the instantaneous frequency is:

Step 3-4: Determine whether transient disturbances have occurred, including voltage swell disturbances, voltage sag disturbances, and frequency fluctuation disturbances. If so, go to step 3-5, otherwise, go back to step 3-1 and execute the next cycle;

Step 3-5: determine whether the transient disturbance is a periodic disturbance, if so, execute step 3-6, otherwise, return to step 3-1 to execute the next cycle;

Step 3-6: Calculate the amount of disturbance;

Step 3-7: Determine whether the disturbance amount exceeds the grid-connected standard, if so, cut the wind turbine out of the grid, otherwise, return to step 3-1 to execute the next cycle.

Preferably, the power control sub-method includes:

Step 4-1: Use the power control model to control the power of the unit;

Described power control model comprises PID sub-model and coordinate transformation sub-model based on neural network;

The described PID sub-model based on the neural network is specifically the incremental PID control algorithm that selects the neural network as the training model:

u(k)=u(k-1)+Δu(k)

Δu(t)=k _p (e(k)-e(k-1))+k _i e(k)+k _d (e(k)-2e(k-1)+e(k-2))

Among them, u(k-1) is the state at the previous moment; u(k) is the state at the current moment; k _p , k _i and k _d are the proportional coefficient, integral coefficient and differential coefficient, respectively;

The steps for training a neural network are as follows:

S1: Determine the number of nodes N and Q of the input layer and the hidden layer, and give the initial value of the weighting coefficient of each layer, select the learning rate η and the inertia coefficient α, and set k=1;

S2: Calculate the current time error:

e(k)=rin(k)-yout(k)

Among them, rin(k) is the expected quantity; yout(k) is the output quantity;

S3: Calculate the input and output of neurons in each layer of the neural network model, and select the final output of the neural network as k _p , _ki and k _d ;

S4: Calculate u(k);

S5: Perform neural network learning, modulate weighting coefficients, and realize self-adaptive adjustment of PID control parameters;

S5: Set k=k+1, return to S1 and repeat;

The 2s/2r transformation matrix of the coordinate transformation sub-model is specifically:

Among them, θ is the angle between the α axis and the d axis;

The inverse matrix of the 2s/2r transformation matrix is specifically:

Step 4-2: Recalculate the active power and reactive power, and determine whether the preset threshold is satisfied, if so, end the current cycle, otherwise, return to step 4-1.

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

1. The reliability of power quality monitoring is high, and at the same time, the power quality is effectively adjusted and improved: the embedded system for monitoring and improving the power quality of wind turbines in the present invention uses FPGA as the processor, and has the ability of high-performance parallel algorithms. The reliability and real-time performance of power quality monitoring are improved; at the same time, the power quality is effectively improved by monitoring transient disturbances, harmonic distortion analysis and power control, reducing the impact and fluctuation on the power grid.

2. Realize the visualization of power quality monitoring: The embedded system used for monitoring and improving the power quality of wind turbines in the present invention is provided with a host computer, and the data related to power quality monitoring is visualized in the host computer to realize the visualization of power quality monitoring. .

3. Strong practicability: the embedded system and method for monitoring and improving the power quality of wind turbines in the present invention can effectively detect power quality problems such as voltage fluctuation, power instability, and harmonic pollution caused by unstable wind speed. And it can quickly and accurately control the wind power system to adjust and improve the power quality, reducing economic losses and casualties caused by power quality problems.

4. Good real-time performance: The embedded system used for monitoring and improving the power quality of wind turbines in the present invention takes the FPGA chip as the main control module, and uses its advantages of high parallel computing capability, fast speed and good real-time performance to realize the monitoring of wind turbines. Monitoring and improvement of power quality.

Description of drawings

1 is a schematic structural diagram of an embedded system in the present invention;

FIG. 2 is a schematic flowchart of the power quality monitoring and improvement method in the present invention;

3 is a schematic flowchart of a sub-method of wind speed prediction in the present invention;

4 is a schematic flowchart of the harmonic distortion analysis sub-method in the present invention;

5 is a schematic flowchart of a sub-method for monitoring transient disturbance in the present invention;

6 is a schematic flowchart of a power control sub-method in the present invention;

FIG. 7 is a control block diagram of a PID sub-model based on a neural network in the present invention.

The numbers in the figure show:

1. Existing wind turbines, 2. Main control module, 3. Signal acquisition module, 4. Frequency detection module, 5. Power control module, 6. Communication module, 7. Host computer, 8. Storage module, 9. Wind speed sensor , 301, voltage sensor, 302, current sensor, 303, signal conditioning circuit, 304, voltage and current sampling circuit, 401, isolation transformer, 402, filter shaping circuit, 403, Schmitt trigger, 404, double D trigger, 405, inverter, 501, grid side and rotor side PWM converter, 502, SVPWM inverter control unit, 503, IGBT drive unit.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

An embedded system for monitoring and improving the power quality of wind turbines, its structure is shown in Figure 1, including: a main control module 2, a signal acquisition module 3, a frequency detection module 4, a power control module 5, a communication module 6, The upper computer 7, the storage module 8 and the wind speed sensor 9. One end of the signal acquisition module 3, the frequency detection module 4 and the power control module 5 are respectively connected with the existing wind turbine 1, the other end is respectively connected with the main control module 2, the storage module 8 and the wind speed sensor 9 are respectively connected with the main control module 2, The main control module 2 communicates with the upper computer 7 through the communication module 6 .

The main modules are described in detail below:

1. Main control module 2

In the hardware design, the main control module is the circuit that constitutes the minimum system operation, which mainly includes the configuration unit, the power supply unit, and the external expansion memory unit. The chip of the main control module 2 selects the FPGA of the SPARTAN-6 series xc6slx25-3csg324 of Xilinx Company, the configuration method adopts the JTAG mode, and the power introduction pins are divided into the phase-locked loop power supply, the core power supply, and the I/O port power supply. Add filter capacitors to ensure stable power supply. Due to the lack of space in the FPGA, a Flash and an SDRAM are expanded as expansion chips.

2. Signal acquisition module 3

The signal acquisition module 3 includes a voltage sensor 301, a current sensor 302, a signal conditioning circuit 303, and a voltage and current sampling circuit 304. One end of the voltage sensor 301 and the current sensor 302 is respectively connected to the existing wind turbine 1, and the other end is respectively connected to the signal conditioning circuit 303. The output end of the signal conditioning circuit 303 is connected to the input end of the voltage and current sampling circuit 304 , and the output end of the voltage and current sampling circuit 304 is connected to the main control module 2 .

The voltage and current sampling circuit 304 selects two ADS7864 chips from TI Company, and the sampling includes the voltage and current signals on the stator side and the voltage and current signals on the rotor side.

3. Frequency detection module 4

The frequency detection module 4 includes an isolation transformer 401, a filter shaping circuit 402, a Schmitt trigger 403, a double D trigger 404 and an inverter 405 connected in sequence. The primary side of the isolation transformer is connected to the existing wind turbine 1, the inverter The output end of 405 is connected with the main control module 2 .

4. Power control module 5

It includes grid-side and rotor-side PWM converters 501, SVPWM inverter control unit 502 and IGBT drive unit 503, FPGA main control unit, and the input ends of grid-side and rotor-side PWM converters 501 are respectively connected with the existing grid and The generator of the existing wind turbine 1 is connected, and the output end of the IGBT drive unit 503 is connected to the main control module 2 .

5. Communication module 6

The network communication between the FPGA and the host computer is realized by the LAN91C111 chip of SMSC Company, and the Ethernet communication protocol standard IEEE802.3 is used to complete the data exchange.

6. Host computer 7

The data involved in this embodiment can be displayed in the upper computer 7, and the upper computer 7 realizes the visualization of power quality monitoring.

Seven, wind speed sensor 9

The wind cup type photoelectric sensor is used to detect the wind speed, and the model is STM2.

This embodiment also relates to a method for monitoring and improving the power quality of a generator set, the process of which is shown in Figure 2, including:

Step 1: According to the wind speed and wind direction samples uploaded by the main control module 2, use the wind speed prediction sub-method to obtain the predicted wind speed;

Step 2: Obtain the voltage and current signals on the rotor and stator sides of the existing wind turbine 1, and obtain the frequency through the frequency detection module 4 at the same time;

Step 3: Use the harmonic distortion analysis sub-method and the transient disturbance monitoring sub-method to analyze the power quality output by the existing wind turbine 1 at the current moment;

Step 4: Obtain the rotational speed data of the generator rotor of the existing wind turbine 1;

Step 5: Calculate the power of the generator in the current existing wind turbine 1;

Step 6: Determine whether the current power of the generator meets the load demand, if so, go to Step 8, otherwise, go to Step 7;

Step 7: use the power control sub-method for power control, and then perform step 8;

Step 8: Obtain wind speed forecast data;

Step 9: Determine whether the absorbed power of the unit needs to be adjusted, if so, adjust the absorbed power of the unit, and then execute step 10, otherwise, directly execute step 10;

Step 10: End the current round of calculation, re-acquire the wind speed data at the current moment, and determine whether the difference between the wind speed and the wind speed forecast data at the previous moment is greater than the preset error, if so, return to step 1, and then perform the next round of calculation, otherwise, return Step 2, perform the next round of calculation.

Each sub-method is described in detail below:

1. Wind speed prediction sub-method

The flow of the wind speed prediction sub-method is shown in Figure 3, including:

Step 1-1: Obtain the voltage and current at the output terminal of the generator stator side of the existing wind turbine 1 and the voltage and current signals at the rotor side collected by the signal collection module 3, and at the same time obtain the wind speed sample data collected by the wind speed sensor 9;

Step 1-2: Normalize the data to obtain a sample training set;

The normalization process is specifically:

Among them, v _i is the original value of the wind speed sample data; v _max and v _min are the maximum and minimum values in the wind speed sample data, respectively; v _i ' is the normalized output value of the wind speed;

Step 1-3: Initialize the improved PSO algorithm;

Step 1-4: Calculate the initial fitness, and the calculation method of the fitness is:

Among them, N is the sample size; v _i ^* is the predicted value of wind speed;

Steps 1-5: Perform iterative optimization to update particle velocity and position. The specific methods are:

和

分别为D维上的第i个粒子进行k+1次迭代的位置和速度；z为收缩因子；φ为总加速因子；g_en为总迭代次数；K为收缩系数；Among them, d=1,2,3,...,D; i=1,2,3,...,N; k is the current iteration number; v _id is the current speed; c ₁ and c ₂ are acceleration factors, and c ₁ and c ₂ are both greater than zero; r ₁ and r ₂ are random functions, and the value range is [0, 1];

and

are the position and velocity of the ith particle in the D dimension for k+1 iterations; z is the shrinkage factor; φ is the total acceleration factor; _gen is the total number of iterations; K is the shrinkage coefficient;

The shrinkage factor z is introduced into the PSO algorithm. The change of the Zhousuo factor can not only ensure the convergence of the PSO algorithm, but also not be affected by the speed boundary, which can speed up the global search speed of the population and enhance the local search ability of particles;

Steps 1-6: Calculate fitness, update individual and global best fitness and optimal solution;

Step 1-7: Determine whether the fitness meets the requirements or whether the number of iterations reaches the maximum number of iterations, if so, execute steps 1-8, otherwise, return to steps 1-5;

Steps 1-8: Use the global optimal solution as the parameter of the least squares support vector machine LSSVM, train the LSSVM, and obtain wind speed prediction data.

2. Harmonic Distortion Analysis Sub-method

The flowchart of the harmonic distortion analysis sub-method is shown in Figure 4, including:

Step 2-1: The sampled voltage or current signal contains high-frequency interference noise or high-frequency harmonic components. Therefore, a finite-length digital FIR filter is designed through the ROM look-up table method to complete the signal denoising. Since the sampling of the signal cannot be infinite Long sampling, so sampling is a finite length signal for analysis;

Step 2-2: Input data in reverse order, and output the data obtained by using the windowed FFT method in natural order;

The method of using Fourier window truncation is to truncate the time domain signal with finite length and turn it into a finite length discrete sequence, so that it can be analyzed by discrete Fourier transform. The selection of window function also effectively reduces the frequency spectrum. When the leakage effect occurs, the harmonic components of the signal can be obtained through the spectrum of the signal. After the signal is de-noised, the data is taken out in a bit reverse order and sent to the windowed FFT module for operation, and the output order is the natural order;

Step 2-3: Calculate the total harmonic distortion rate, the calculation method is:

Among them, U _n is the effective value of the nth harmonic voltage; In is the effective value of the _nth harmonic current; THD _u is the total harmonic distortion rate of voltage; THD _i is the total harmonic distortion rate of current; U ₁ and I ₁ are the rms value of the fundamental wave voltage and the rms value of the fundamental wave current, respectively;

Step 2-4: Determine whether the total harmonic distortion rate exceeds the preset value, if so, disconnect the existing wind turbine 1 from the grid, otherwise, end the current cycle.

3. Transient disturbance monitoring sub-method

The flow of the transient disturbance monitoring sub-method is shown in Figure 5, including:

Step 3-1: Determine whether the voltage and current signals on the rotor and stator sides obtained in Step 2 exceed the preset values, if so, cut the fan out of the grid to end the current cycle, otherwise, go to Step 3-2;

Step 3-2: Use the FIR filter to de-noise the voltage and current signal data on the stator side;

Step 3-3: Use the Hilbert-Huang transform detection algorithm HHT to analyze the data obtained in step 3-2. The HHT algorithm includes two parts, the EMD and the Hilbert spectrum analysis method. First, the given signal is decomposed into several natural modes using the EMD method. state function IMF, and then perform Hilbert transform on each IMF, obtain the characteristic parameters, and obtain the Hilbert spectrum of the corresponding original signal. The specific steps are:

Decompose the empirical mode to get the intrinsic mode function group:

Among them, s _i (t) is the natural modal component in the original signal; d(t) is the DC component in the original signal; r(t) is the residual function data sequence obtained by decomposition;

Then perform Hilbert transform on all IMFs to obtain the analytical expression of each IMF transformed:

Among them, a _i is the IMF amplitude, and the above formula is the Hilbert spectrum, which can be written as:

Integrate the above formula in the time domain to obtain the corresponding Hilbert marginal spectrum:

where the instantaneous frequency is:

Step 3-4: Determine whether transient disturbances have occurred, including voltage swell disturbances, voltage sag disturbances and frequency fluctuation disturbances, if so, go to step 3-5, otherwise, go back to step 3-1, and execute the following one cycle;

Step 3-5: determine whether the transient disturbance is a periodic disturbance, if so, execute step 3-6, otherwise, return to step 3-1 to execute the next cycle;

Step 3-6: Calculate the amount of disturbance;

Step 3-7: Determine whether the disturbance amount exceeds the grid-connected standard, if so, cut the wind turbine out of the grid, otherwise, return to step 3-1 to execute the next cycle.

Fourth, the power control sub-method

The flow of the power control sub-method is shown in Figure 6, including:

Step 4-1: Use the power control model to control the power of the unit;

The power control model includes a neural network-based PID sub-model and a coordinate transformation sub-model. The control block diagram of the neural network-based PID sub-model is shown in Figure 7.

The PID sub-model based on neural network is an incremental PID control algorithm that selects neural network as the training model:

u(k)=u(k-1)+Δu(k)

Δu(t)=k _p (e(k)-e(k-1))+k _i e(k)+k _d (e(k)-2e(k-1)+e(k-2))

Among them, u(k-1) is the state at the previous moment; u(k) is the state at the current moment; k _p , k _i and k _d are the proportional coefficient, integral coefficient and differential coefficient, respectively;

The steps for training a neural network are as follows:

S1: Determine the number of nodes N and Q of the input layer and the hidden layer, and give the initial value of the weighting coefficient of each layer, select the learning rate η and the inertia coefficient α, and set k=1;

S2: Calculate the current time error:

e(k)=rin(k)-yout(k)

Among them, rin(k) is the expected quantity; yout(k) is the output quantity;

S3: Calculate the input and output of neurons in each layer of the neural network model, and select the final output of the neural network as k _p , _ki and k _d ;

S4: Calculate u(k);

S5: Perform neural network learning, modulate weighting coefficients, and realize self-adaptive adjustment of PID control parameters;

S5: Set k=k+1, return to S1 and repeat;

The 2s/2r transformation matrix of the coordinate transformation sub-model is specifically:

Among them, θ is the angle between the α axis and the d axis;

The inverse matrix of the 2s/2r transformation matrix is specifically:

Step 4-2: Recalculate the active power and reactive power, and determine whether the preset threshold is satisfied, if so, end the current cycle, otherwise, return to step 4-1.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person skilled in the art can easily think of various equivalents within the technical scope disclosed by the present invention. Modifications or substitutions should be included within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims (10)

1. an embedded system for monitoring and improving the power quality of wind turbines, the system is connected with existing wind turbines (1), it is characterized in that, described embedded system comprises: main control module (2), signal an acquisition module (3), a frequency detection module (4), a power control module (5), a communication module (6), a host computer (7), a storage module (8) and a wind speed sensor (9); the signal acquisition module (3) One end of the frequency detection module (4) and the power control module (5) are respectively connected with the existing wind turbine (1), and the other end is respectively connected with the main control module (2); the storage module (8) and the wind speed sensor (9) are respectively connected with the main control module (2); the main control module (2) communicates with the upper computer (7) through the communication module (6). 2 . The embedded system for monitoring and improving the power quality of wind turbines according to claim 1 , wherein the main control module ( 2 ) is an FPGA chip. 3 . 3. The embedded system for monitoring and improving the power quality of wind turbines according to claim 1, wherein the signal acquisition module (3) comprises a voltage sensor (301), a current sensor (302) , a signal conditioning circuit (303) and a voltage and current sampling circuit (304); one end of the voltage sensor (301) and the current sensor (302) are respectively connected with the existing wind turbine (1), and the other ends are respectively connected with the signal conditioning circuit The input end of (303) is connected; the output end of the signal conditioning circuit (303) is connected with the input end of the voltage and current sampling circuit (304); the output end of the voltage and current sampling circuit (304) is connected to the main control module (2) Connected. 4. An embedded system for monitoring and improving the power quality of wind turbines according to claim 1, wherein the frequency detection module (4) comprises an isolation transformer (401) connected in sequence, a filter shaping a circuit (402), a Schmitt trigger (403), a double D trigger (404) and an inverter (405); the primary side of the isolation transformer is connected to the existing wind turbine (1); the The output end of the inverter (405) is connected with the main control module (2). 5. An embedded system for monitoring and improving the power quality of wind turbines according to claim 1, wherein the power control module (5) comprises grid-side and rotor-side PWM converters connected in sequence (501), an SVPWM inverter control unit (502) and an IGBT drive unit (503); the input ends of the grid-side and rotor-side PWM converters (501) are respectively connected to the existing grid and the existing wind turbines (1 ) is connected to the generator; the output end of the IGBT drive unit (503) is connected to the main control module (2). 6. A method for monitoring and improving the power quality of wind turbines for the embedded system according to claim 1, characterized in that, comprising: Step 1: According to the wind speed and wind direction samples uploaded by the main control module (2), the wind speed prediction sub-method is used to obtain the predicted wind speed; Step 2: acquiring the voltage and current signals on the rotor and stator sides of the existing wind turbine (1), and simultaneously acquiring the frequency through the frequency detection module (4); Step 3: using the harmonic distortion analysis sub-method and the transient disturbance quantity monitoring sub-method to analyze the power quality output by the existing wind turbine (1) at the current moment; Step 4: Obtain the rotational speed data of the generator rotor of the existing wind turbine (1); Step 5: Calculate the power of the generator in the current existing wind turbine (1); Step 6: Determine whether the current power of the generator meets the load demand, if so, go to Step 8, otherwise, go to Step 7; Step 7: use the power control sub-method for power control, and then perform step 8; Step 8: Obtain wind speed forecast data; Step 9: Determine whether the absorbed power of the unit needs to be adjusted, if so, adjust the absorbed power of the unit, and then execute step 10, otherwise, directly execute step 10; Step 10: End the current round of calculation, re-acquire the wind speed data at the current moment, and determine whether the difference between the wind speed and the wind speed forecast data at the previous moment is greater than the preset error, if so, return to step 1, and then perform the next round of calculation, otherwise, return Step 2, perform the next round of calculation. 7. The method for monitoring and improving the power quality of wind turbines according to claim 6, wherein the wind speed prediction sub-method comprises: Step 1-1: Acquire the voltage and current at the output terminal of the generator stator side and the voltage and current signals at the rotor side of the existing wind turbine generator set (1) collected by the signal acquisition module (3), and simultaneously acquire the wind speed samples collected by the wind speed sensor (9) data; Step 1-2: Normalize the data to obtain a sample training set; The normalization process is specifically:

Among them, v _i is the original value of the wind speed sample data; v _max and v _min are the maximum and minimum values in the wind speed sample data, respectively; v _i ' is the normalized output value of the wind speed; Step 1-3: Initialize the improved PSO algorithm; Step 1-4: Calculate the initial fitness, and the calculation method of the fitness is:

Among them, N is the sample size; v _i ^* is the predicted value of wind speed; Steps 1-5: Perform iterative optimization to update particle velocity and position. The specific methods are:

Among them, d=1,2,3,...,D; i=1,2,3,...,N; k is the current iteration number; v _id is the current speed; c ₁ and c ₂ are acceleration factors, and c ₁ and c ₂ are both greater than zero; r ₁ and r ₂ are random functions, and the value range is [0, 1];

and

are the position and velocity of the ith particle in the D dimension for k+1 iterations; z is the shrinkage factor; φ is the total acceleration factor; _gen is the total number of iterations; K is the shrinkage coefficient; Steps 1-6: Calculate fitness, update individual and global best fitness and optimal solution; Step 1-7: Determine whether the fitness meets the requirements or whether the number of iterations reaches the maximum number of iterations, if so, execute steps 1-8, otherwise, return to steps 1-5; Steps 1-8: Use the global optimal solution as the parameter of the least squares support vector machine LSSVM, train the LSSVM, and obtain wind speed prediction data. 8. The method for monitoring and improving the power quality of wind turbines according to claim 6, wherein the harmonic distortion analysis sub-method comprises: Step 2-1: Use the FIR filter to de-noise the voltage and current signals on the rotor and stator sides; Step 2-2: Input data in reverse order, and output the data obtained by using the windowed FFT method in natural order; Step 2-3: Calculate the total harmonic distortion rate, the calculation method is:

Among them, U _n is the effective value of the nth harmonic voltage; In is the effective value of the _nth harmonic current; THD _u is the total harmonic distortion rate of voltage; THD _i is the total harmonic distortion rate of current; U ₁ is the rms value of the fundamental wave voltage; I ₁ is the rms value of the fundamental wave current; Step 2-4: Determine whether the total harmonic distortion rate exceeds the preset value, if so, disconnect the existing wind turbine (1) from the grid, and then end the current cycle, otherwise, directly end the current cycle. 9 . The method for monitoring and improving the power quality of a wind turbine according to claim 6 , wherein the transient disturbance monitoring sub-method comprises: 10 . Step 3-1: Determine whether the voltage and current signals on the rotor and stator sides obtained in Step 2 exceed the preset values, if so, cut the fan out of the grid to end the current cycle, otherwise, go to Step 3-2; Step 3-2: Use the FIR filter to de-noise the voltage and current signal data on the stator side; Step 3-3: Use the Hilbert-Huang transform detection algorithm HHT to analyze the data obtained in step 3-2. The HHT algorithm includes two parts, the EMD and the Hilbert spectrum analysis method. First, the given signal is decomposed into several natural modes using the EMD method. state function IMF, and then perform Hilbert transform on each IMF, obtain the characteristic parameters, and obtain the Hilbert spectrum of the corresponding original signal. The specific steps are: Decompose the empirical mode to get the intrinsic mode function group:

Among them, s _i (t) is the natural modal component in the original signal; d(t) is the DC component in the original signal; r(t) is the residual function data sequence obtained by decomposition; Then perform Hilbert transform on all IMFs to obtain the analytical expression of each IMF transformed:

Among them, a _i is the IMF amplitude, and the above formula is the Hilbert spectrum, which can be written as:

Integrate the above formula in the time domain to obtain the corresponding Hilbert marginal spectrum:

where the instantaneous frequency is:

Step 3-4: Determine whether transient disturbances have occurred, including voltage swell disturbances, voltage sag disturbances, and frequency fluctuation disturbances. If so, go to step 3-5, otherwise, go back to step 3-1 and execute the next cycle; Step 3-5: determine whether the transient disturbance is a periodic disturbance, if so, execute step 3-6, otherwise, return to step 3-1 to execute the next cycle; Step 3-6: Calculate the amount of disturbance; Step 3-7: Determine whether the disturbance amount exceeds the grid-connected standard, if so, cut the wind turbine out of the grid, otherwise, return to step 3-1 to execute the next cycle. 10. The method for monitoring and improving the power quality of wind turbines according to claim 6, wherein the power control sub-method comprises: Step 4-1: Use the power control model to control the power of the unit; Described power control model comprises PID sub-model and coordinate transformation sub-model based on neural network; The described PID sub-model based on the neural network is specifically the incremental PID control algorithm that selects the neural network as the training model: u(k)=u(k-1)+Δu(k) Δu(t)=k _p (e(k)-e(k-1))+k _i e(k)+k _d (e(k)-2e(k-1)+e(k-2)) Among them, u(k-1) is the state at the previous moment; u(k) is the state at the current moment; k _p , k _i and k _d are the proportional coefficient, integral coefficient and differential coefficient, respectively; The steps for training a neural network are as follows: S1: Determine the number of nodes N and Q of the input layer and the hidden layer, and give the initial value of the weighting coefficient of each layer, select the learning rate η and the inertia coefficient α, and set k=1; S2: Calculate the current time error: e(k)=rin(k)-yout(k) Among them, rin(k) is the expected quantity; yout(k) is the output quantity; S3: Calculate the input and output of neurons in each layer of the neural network model, and select the final output of the neural network as k _p , _ki and k _d ; S4: Calculate u(k); S5: Perform neural network learning, modulate weighting coefficients, and realize self-adaptive adjustment of PID control parameters; S5: Set k=k+1, return to S1 and repeat; The 2s/2r transformation matrix of the coordinate transformation sub-model is specifically:

Among them, θ is the angle between the α axis and the d axis; The inverse matrix of the 2s/2r transformation matrix is specifically:

Step 4-2: Recalculate the active power and reactive power, and determine whether the preset threshold is satisfied, if so, end the current cycle, otherwise, return to step 4-1.

Images