CN102854465A - System and method for chaotic prediction of DFIG (doubly fed induction generator) running state based on phase-space reconstruction - Google Patents

Abstract

The invention discloses a system for chaotic prediction of a DFIG (doubly fed induction generator) running state based on phase-space reconstruction. The system is characterized by comprising a wind generation set, a test instrument and an upper computer with a chaotic prediction program. A working method of the system comprises the steps of acquiring a signal, filtering, correcting, calculating parameters, chaotically predicating, and predicating a wind turbine state. The system has the advantages as follows: (1) hardware devices are simple and practical; (2) data is automatically protected after power failure, chaotic real-time detection is continuous, and reliability is high; (3) measuring precision is high; (4) the real-time performance and the reliability of the system are high, requirements of monitoring a state of wind generation set, researching a transitional process, diagnosing and predicting a fault, and the like are met, and a high practical value is obtained.

Description

A chaotic prediction system and method for DFIG operating state based on phase space reconstruction

(1) Technical field:

The invention belongs to the technical field of wind power system prediction, in particular to a DFIG (full spelling in English—doubly-fed induction generator) operating state chaos prediction system and method based on phase space reconstruction.

(2) Background technology:

With the increasing capacity of wind turbines (mainly doubly-fed induction generators) connected to the power grid, due to the harsh operating conditions of wind turbines, such as large changes in external temperature difference and random changes in wind speed, etc. These uncertain external factors lead to a high failure rate of wind turbines, resulting in high post-operation and maintenance costs of wind farms, and the unique intermittent and random nature of wind power generation greatly increases the instability of its interconnected grid. As a complex coupled nonlinear system, the safe and reliable operation of doubly-fed induction generator sets directly affects the stable load distribution of its interconnected grid and the quality of grid power supply. Therefore, improving the operational reliability of doubly-fed induction generator sets plays an important role in ensuring the safe and high-quality operation of the power grid and improving the system economy. In order to realize the safe and stable operation of the doubly-fed induction generator set, judge the performance status and development trend of the unit components in time and improve its operating efficiency, the present invention adopts various methods and means to carry out online monitoring on the important components (gearbox, generator, etc.) of the unit Based on the analysis of the chaotic properties of the operating state of the doubly-fed induction generating set, a weighted first-order local prediction model for the operating state of the generating set is established by using the phase space reconstruction method to predict its operating state so as to detect faults in advance Symptoms, avoid and reduce serious equipment damage, determine a reasonable maintenance time and plan, so as to achieve the purpose of greatly reducing maintenance costs.

(3) Contents of the invention:

The object of the present invention is to provide a DFIG operating state chaos prediction system and method based on phase space reconstruction. Systems and methods for forecasting.

Technical solution of the present invention: a DFIG operating state chaos prediction system based on phase space reconstruction, characterized in that it includes a wind turbine, a tester and a host computer with a chaos prediction program; wherein, the tester collects the wind turbine The signal is bidirectionally connected with the host computer with chaos prediction program.

The tester is composed of a signal acquisition and conditioning circuit unit, an A/D conversion circuit unit, a single-chip microcomputer, a USB interface circuit unit, a data storage circuit unit, and a timing circuit unit; wherein, the signal acquisition and conditioning circuit unit collects wind power The signal of the unit, its output end connects the input end of A/D conversion circuit unit; Described single-chip microcomputer and A/D conversion circuit unit, USB interface circuit unit, data storage circuit unit and fixed circuit unit are bidirectionally connected respectively; The USB interface circuit unit is bidirectionally connected with the host computer with chaos prediction program.

The signal collection of the wind turbine is to collect the temperature signal of the driving side bearing of the gearbox of the wind turbine, the maximum temperature signal of the winding of the wind turbine, the average speed signal of the rotor of the wind turbine and the active power parameter signal of the wind turbine.

The signal acquisition and conditioning circuit unit is composed of a sensor, a data acquisition card, a resistor Rs, a filter circuit, a voltage follower, a conditioning circuit and a voltage regulator tube; its connection is a conventional connection; wherein the sensor uses an isolation template, and the input signal All are converted into 5V standard voltage signals; the data acquisition card is Advantech’s PCI-1711 12-bit multi-function data acquisition card, which has 16 single-ended analog inputs, 8 data signal channels, and an automatic channel/ The gain scanning circuit automatically controls the multi-channel strobe switch when sampling, and its connection is a conventional connection.

The A/D conversion circuit unit is composed of a conversion chip and a peripheral circuit; wherein the conversion chip adopts a CMOS process and has a three-state data output latch in the chip, and the input mode is a single channel, and the conversion time is 100 μs, and the power supply voltage It is a successive approximation type 8-bit conversion chip ADC0804 of +5V; the conversion chip ADC0804 includes pins DB0, pins DB1, pins DB2, pins DB3, pins DB4, pins DB5, pins DB6, pins DB7 , pin /WR, pin /RD, pin /CS, pin VIN(+), pin VIN(-), pin CLK-IN, pin CLK-R, and pin Vref/2, and According to pin DB0, pin DB1, pin DB2, pin DB3, pin DB4, pin DB5, pin DB6, pin DB7, pin/WR, pin/RD, pin/CS and MCU chip It is connected in a waiting delay mode; the peripheral circuit is composed of a capacitor C28, a resistor R32, two resistors R33, a capacitor C29, and a power supply VCC; the pin VIN (+) receives a signal conditioning circuit through a capacitor C28 and a resistor R33 The processed signal, the connection point of the capacitor C28 and the resistor R33 and the pin VIN (-) are connected to the common ground, and the differential voltage analog input mode is adopted; the pin CLK-R is grounded through another resistor R33 and the capacitor C29, and the tube The pin CLK-IN is connected to the connection point of the resistor R33 and the capacitor C29; the pin Vref/2 is connected to the power supply VCC through the resistor R32.

The single-chip microcomputer adopts Freescale's single-chip microcomputer MC9S12DP256.

The data storage circuit unit adopts the DS1225 chip of Dallas Company.

The USB interface circuit unit adopts the CH372 chip of Nanjing Qinheng Electronics.

The timing circuit unit adopts a PIC16F716 device with a watchdog.

A kind of working method of the chaotic prediction system that is used for wind power system is characterized in that it comprises the following steps:

(1) The timing circuit unit sets the collection interval timing time, and the signal collection and conditioning circuit unit collects the driving side bearing temperature of the wind turbine gearbox, the maximum temperature of the generator winding, the average rotor speed and the active power signal of the generator in real time;

(2) The signal collected in step (1) is filtered and self-calibrated by the signal acquisition and conditioning circuit unit and the A/D conversion circuit unit, and the driving side bearing temperature of the unit gearbox, the maximum temperature of the generator winding, the average rotor speed and Generator active power parameters are input into the data storage circuit unit;

(3) Through the USB interface circuit, the temperature of the driving side bearing of the wind turbine gearbox, the maximum temperature of the generator winding, the average speed of the rotor and the active power of the generator after processing in step (2) are transmitted to the host computer equipped with the chaos prediction program;

⑷Using the chaotic prediction method in the host computer for parameter calculation and processing.

The method for predicting chaos in the step (4) is to adopt a method based on phase space reconstruction to detect chaos, and is composed of the following steps:

①Use the CC method to determine the embedding dimension and time delay: when the wind turbine is running, for a certain state parameter time series x={ _xi },i=1,2,...,N, if the embedding dimension is m, the time delay is τ, then the reconstructed phase space is X={X _i }, and X _i is the phase point in the m-dimensional phase space:

X _i =[x _i ,x _i+τ ,…,x _i+(m-1)τ ] ^T ,i=1,2,…,M (1)

Then the associated integral of the embedded time series is

$\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Where M=N-(m-1)τ, d _ij =||x _i -x _j || _∞ is the ∞ norm; θ is the Heaviside function, and its expression is

$\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right.$

The correlation integral is a cumulative distribution function that expresses the probability that the distance between any two points in the phase space is less than r. Additionally define the test statistic for x={ _xi }:

S(m,N,r,τ)=C(m,N,r,τ)-C ^m (m,N,r,τ) (3)

S(m, N, r, τ) reflects the autocorrelation characteristics of the time series, the optimal time delay takes the first zero point of S(m, N, r, τ), and the point in the reconstructed phase space is closest to Evenly distributed, the reconstructed attractor orbit is completely expanded in the phase space;

② Use the largest Lyapunov exponent to identify the chaotic characteristics of DFIG: from the time delay τ and embedding dimension m calculated in step ①, apply the small data volume method to calculate the bearing temperature on the driving side of the wind turbine gearbox, the maximum temperature of the generator winding, and the average rotor speed and the maximum Lyapunov exponents of the four state parameter time series of generator active power, if the four maximum Lyapunov exponents are greater than 0, it means that there is DFIG chaotic characteristics, and the next step is predicted;

③Using the weighted first-order local method to predict DFIG:

Reconstruct the phase space of the delay time τ and dimension m obtained in step ①, and apply the weighted first-order local method to the wind turbine gearbox drive side bearing temperature, the maximum temperature of the generator winding, and the average rotor speed The time series of the four state parameters and the active power of the generator are respectively predicted;

Let the adjacent point of the central point (that is, the starting point of prediction) X _k be X _ki , the distance between two points is d _i , let d _m be the minimum value in d _i , and the weight of point X _ki is:

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}$

，得到预测式X_k+1=a+bX_k，构造下一中心点及其邻近点，重复③以进一步预测，直至预测到机组出现异常情况预测停止。Generally take l=1, then the weighted first-order local linear fitting is X _ki+1 =ae+bX _ki , e=[1,1,…,1] ^T . Solve according to the method of weighted least squares

, get the prediction formula X _k+1 =a+bX _k , construct the next center point and its adjacent points, repeat step 3 for further prediction, until the abnormal situation of the unit is predicted and the prediction stops.

Working principle of the present invention:

The doubly-fed induction generator system is a complex multi-dimensional nonlinear system. The change of its operating state is affected by many factors. It has great limitations to predict the operating state of the unit only relying on a certain factor. However, it is easy to obtain in actual production The time series of operating parameters can only reflect part of the information, so the method of reconstructing the phase space is considered for prediction. This method uses a single system output time series to construct a set of coordinate components that represent the dynamic characteristics of the original system, thereby approximately restoring the system. chaotic attractor. On this basis, use the reconstructed phase space (approximation of the actual system) to predict (mainly use the system's orbital development, that is, the chaos attractor to predict).

According to the Takens theorem, for a time series, when m≥2d+1 (m is the embedding dimension, d is the correlation dimension of the dynamical system), the attractor can be recovered in the m-dimensional reconstruction space. The phase trajectory in the reconstructed space remains diffeomorphic to the prime mover system.

The reconstruction method is as follows:

The CC method determines the embedding dimension and time delay: when the wind turbine is running, for a certain state parameter time series x={ _xi },i=1,2,...,N, if the embedding dimension is m, the time delay is τ , then the reconstructed phase space is X={X _i }, where X _i is the phase point in the m-dimensional phase space:

X _i =[x _i ,x _i+τ ,…,x _i+(m-1)τ ] ^T ,i=1,2,…,M (1)

Then the associated integral of the embedded time series is

$\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Where M=N-(m-1)τ, d _ij =||x _i -x _j || _∞ is the ∞ norm; θ is the Heaviside function, and its expression is

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}$

The correlation integral is a cumulative distribution function that expresses the probability that the distance between any two points in the phase space is less than r. Additionally define the test statistic for x={ _xi }:

S(m,N,r,τ)=C(m,N,r,τ)-C ^m (m,N,r,τ) (3)

S(m, N, r, τ) reflects the autocorrelation characteristics of the time series, the optimal time delay takes the first zero point of S(m, N, r, τ), and the point in the reconstructed phase space is closest to Uniformly distributed, the reconstructed attractor orbitals are fully expanded in phase space.

估算出最大Lyapunov指数。The largest Lyapunov exponent to identify the chaotic characteristics of DFIG: the present invention uses a small amount of data method with high reliability and a small amount of calculation to calculate the maximum Lyapunov exponent of the time series of wind turbine state parameters, and after reconstructing the phase space, find The nearest neighbors of each point, given by

Estimate the maximum Lyapunov exponent.

Chaos prediction: The present invention uses the weighted first-order local method prediction model to predict the running state of DFIG. It is an improvement on the original local method. Because of the introduction of weights, it has better adaptive ability and more High prediction accuracy.

Let the adjacent point of the center point X _k (that is, the starting point of prediction) be X _ki , i=1,2,…,q, the distance between two points is d _i , let d _m be the minimum value in d _i , point X _ki has a weight of

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}$

Generally take l=1, then the weighted 1st-order local linear fitting is

X _ki+1 =ae+bX _ki ，e=[1,1,…,1] ^T (4)

，得到预测式X_k+1=a+bX_k，然后构造下一中心点及其邻近点，进行下一步预测。Solve according to the method of weighted least squares

, get the prediction formula X _k+1 =a+bX _k , and then construct the next central point and its adjacent points for the next step of prediction.

The one-chip computer adopts Freescale's one-chip computer MC9S12DP256, and its enhanced capture timer has a more reliable counting function compared with other timers. The MC9S12DP256 microcontroller in this system uses a 16 MHz external crystal oscillator, and the bus clock frequency is configured as 8 MHz. In order to improve the measurement accuracy, the working clock of the timer module does not divide the frequency of the bus clock, that is, the counting frequency of the timer is also 8MHz, and the method of capturing the rising edge is adopted to allow input capture interrupt and timer overflow interrupt. After conditioning, the signal output by the speed sensor becomes a digital signal of speed frequency and transmits it to the input capture module of the single-chip microcomputer MC9S12DP256. By capturing the effective transition edge (such as the rising edge) of the measured signal, record the timing when the effective transition edge comes To calculate the difference between the value of the timer at the current capture time and the timer value at the previous capture time, this is the number of counts of the timer in one cycle of the signal under test, from which the two effective edges can be obtained Interval time and obtain the rotational speed accordingly.

 为选通端口，

 为可输出端口，为可写入端口，NC为不连接端口。The data storage circuit adopts the DS1225 chip of Dallas Company. A0-A12 are address input ports, DQ0-DQ7 are data input/output ports,

is the strobe port,

is an output port, It is a writable port, and NC is a non-connectable port.

The USB interface circuit adopts the CH372 chip of Nanjing Qinheng Electronics, which is composed of 8-bit bidirectional data bus D7~D0 that can be used as a passive parallel interface, read strobe input pin RD#, write strobe input pin WR#, chip select The input pin CS#, the interrupt output pin INT# and the address input pin A0 are composed; the 8-bit bidirectional data bus D7～D0, the read strobe input pin RD#, the write strobe input pin WR#, the chip The selection input pin CS#, the interrupt output pin INT# and the address input pin A0 are connected with the microcontroller.

The advantages of the present invention are: ① simple and practical hardware device; ② automatic protection of data after power failure, uninterrupted real-time detection of chaos, high reliability; ③ high measurement accuracy; ④ high real-time reliability of the system, which can meet the needs of wind turbines It has high practical value for the requirements of condition monitoring, transition process research, fault diagnosis and prediction, etc.

(4) Description of the drawings:

Fig. 1 is a schematic diagram of the overall structure of a chaos prediction system for doubly-fed induction generator set according to the present invention.

Fig. 2 is a schematic diagram of the circuit structure of a signal conditioning circuit unit in a chaos prediction system for a doubly-fed induction generator set according to the present invention.

Fig. 3 is a schematic diagram of the circuit structure of an A/D conversion interface circuit unit in a chaos prediction system for a doubly-fed induction generator set according to the present invention.

Fig. 4 is a schematic structural diagram of the data storage unit DS1225 chip used in the chaos prediction system of the doubly-fed induction generator set according to the present invention.

Fig. 5 is a schematic structural diagram of an interface circuit CH372 chip used in a chaos prediction system for a doubly-fed induction generator set according to the present invention.

Fig. 6 is a general flowchart of a working method of a chaos prediction system for a doubly-fed induction generator set according to the present invention.

(5) Specific implementation methods:

Embodiment: A DFIG operating state chaos prediction system based on phase space reconstruction (see Figure 1), characterized in that it includes a wind turbine, a tester, and a host computer with a chaos prediction program; wherein, the tester collects wind power The signal of the unit is bidirectionally connected with the host computer with chaos prediction program.

The tester (see Figure 1) is composed of a signal acquisition and conditioning circuit unit, an A/D conversion circuit unit, a single-chip microcomputer, a USB interface circuit unit, a data storage circuit unit and a timing circuit unit; wherein the signal acquisition and conditioning The unit collects the signal of the wind turbine, and its output end is connected to the input end of the A/D conversion circuit unit; the single-chip microcomputer is connected with the A/D conversion circuit unit, the USB interface circuit unit, the data storage circuit unit and the fixed circuit unit respectively in two directions ; The USB interface circuit unit is bidirectionally connected to the host computer with the chaos prediction program.

The signal collection of the wind turbine is to collect the temperature signal of the driving side bearing of the gearbox of the wind turbine, the maximum temperature signal of the winding of the wind turbine, the average speed signal of the rotor of the wind turbine and the active power parameter signal of the wind turbine.

The signal acquisition and conditioning circuit unit (see Figure 2) is composed of a sensor, a data acquisition card, a resistor Rs, a filter circuit, a voltage follower, a conditioning circuit and a voltage regulator tube; its connection is a conventional connection; wherein the sensor adopts an isolation The template converts all input signals into 5V standard voltage signals; the data acquisition card is Advantech's PCI-1711 12-bit multi-function data acquisition card, which has 16 single-ended analog inputs and 8 data signal channels, with There is an automatic channel/gain scanning circuit that automatically controls the multiplexing switch when sampling, and its connection is a conventional connection.

The A/D conversion circuit unit (see Figure 3) is composed of a conversion chip and peripheral circuits; wherein the conversion chip adopts CMOS technology, has a three-state data output latch in the chip, and the input mode is a single channel, and the conversion time is 100μs, the power supply voltage is +5V successive approximation 8-bit conversion chip ADC0804; the conversion chip ADC0804 includes pins DB0, pins DB1, pins DB2, pins DB3, pins DB4, pins DB5, pins DB6, pin DB7, pin /WR, pin /RD, pin /CS, pin VIN(+), pin VIN(-), pin CLK-IN, pin CLK-R and pin Vref/2, and by pin DB0, pin DB1, pin DB2, pin DB3, pin DB4, pin DB5, pin DB6, pin DB7, pin/WR, pin/RD, pin /CS is connected to the single-chip microcomputer chip in a waiting delay mode; the peripheral circuit is composed of capacitor C28, resistor R32, two resistors R33, capacitor C29, and power supply VCC; the pin VIN (+) is connected through capacitor C28 and a resistor R33 receives the signal processed by the signal conditioning circuit, the connection point of the capacitor C28 and the resistor R33 is connected to the pin VIN (-) and is connected to the common ground, and adopts a differential voltage analog input mode; the pin CLK-R passes through another resistor R33 and The capacitor C29 is grounded, and the pin CLK-IN is connected to the connection point of the resistor R33 and the capacitor C29; the pin Vref/2 is connected to the power supply VCC through the resistor R32.

The single-chip microcomputer adopts Freescale's single-chip microcomputer MC9S12DP256.

The data storage circuit unit adopts the DS1225 chip of Dallas Company.

The USB interface circuit unit adopts the CH372 chip of Nanjing Qinheng Electronics.

The timing circuit unit adopts a PIC16F716 device with a watchdog.

A working method (see Figure 6) for a chaotic prediction system of a wind power system, characterized in that it comprises the following steps:

(1) The timing circuit unit sets the collection interval timing time, and the signal collection and conditioning circuit unit collects the driving side bearing temperature of the wind turbine gearbox, the maximum temperature of the generator winding, the average rotor speed and the active power signal of the generator in real time;

(2) The signal collected in step (1) is filtered and self-calibrated by the signal acquisition and conditioning circuit unit and the A/D conversion circuit unit, and the driving side bearing temperature of the unit gearbox, the maximum temperature of the generator winding, the average rotor speed and Generator active power parameters are input into the data storage circuit unit;

(3) Through the USB interface circuit, the temperature of the driving side bearing of the wind turbine gearbox, the maximum temperature of the generator winding, the average speed of the rotor and the active power of the generator after processing in step (2) are transmitted to the host computer equipped with the chaos prediction program;

⑷Using the chaotic prediction method in the host computer for parameter calculation and processing.

The chaos prediction method in the step (4) is to use a method based on phase space reconstruction to detect chaos (see Figure 6), which consists of the following steps:

①Use the CC method to determine the embedding dimension and time delay: when the wind turbine is running, for a certain state parameter time series x={ _xi },i=1,2,...,N, if the embedding dimension is m, the time delay is τ, then the reconstructed phase space is X={X _i }, and X _i is the phase point in the m-dimensional phase space:

X _i =[x _i ,x _i+τ ,…,x _i+(m-1)τ ] ^T ,i=1,2,…,M (1)

Then the associated integral of the embedded time series is

$\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Where M=N-(m-1)τ, d _ij =||x _i -x _j || _∞ is the ∞ norm; θ is the Heaviside function, and its expression is

$\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right.$

The correlation integral is a cumulative distribution function that expresses the probability that the distance between any two points in the phase space is less than r. Additionally define the test statistic for x={ _xi }:

S(m,N,r,τ)=C(m,N,r,τ)-C ^m (m,N,r,τ) (3)

S(m, N, r, τ) reflects the autocorrelation characteristics of the time series, the optimal time delay takes the first zero point of S(m, N, r, τ), and the point in the reconstructed phase space is closest to Evenly distributed, the reconstructed attractor orbit is completely expanded in the phase space;

② Use the largest Lyapunov exponent to identify the chaotic characteristics of DFIG: from the time delay τ and embedding dimension m calculated in step ①, apply the small data volume method to calculate the bearing temperature on the driving side of the wind turbine gearbox, the maximum temperature of the generator winding, and the average rotor speed and the maximum Lyapunov exponents of the four state parameter time series of generator active power, if the four maximum Lyapunov exponents are greater than 0, it means that there is DFIG chaotic characteristics, and the next step is predicted;

③Using the weighted first-order local method to predict DFIG:

Reconstruct the phase space of the delay time τ and dimension m obtained in step ①, and apply the weighted first-order local method to the wind turbine gearbox drive side bearing temperature, the maximum temperature of the generator winding, and the average rotor speed The time series of the four state parameters and the active power of the generator are respectively predicted;

Let the adjacent point of the central point (that is, the starting point of prediction) X _k be X _ki , the distance between two points is d _i , let d _m be the minimum value in d _i , and the weight of point X _ki is:

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}$

，得到预测式X_k+1=a+bX_k，构造下一中心点及其邻近点，重复③以进一步预测，直至预测到机组出现异常情况预测停止。Generally take l=1, then the weighted first-order local linear fitting is X _ki+1 =ae+bX _ki , e=[1,1,…,1] _T . Solve according to the method of weighted least squares

, get the prediction formula X _k+1 =a+bX _k , construct the next center point and its adjacent points, repeat step 3 for further prediction, until the abnormal situation of the unit is predicted and the prediction stops.

Claims (10)

1. A DFIG operating state chaos prediction system based on phase space reconstruction is characterized in that it includes a wind turbine, a tester and a host computer with a chaos prediction program; The host computer of the chaos forecasting program is bidirectionally connected. 2. according to claim 1, a kind of DFIG operating state chaos prediction system based on phase space reconstruction is characterized in that said tester is composed of signal acquisition and conditioning circuit unit, A/D conversion circuit unit, single-chip microcomputer, USB interface A circuit unit, a data storage circuit unit and a timing circuit unit; wherein, the signal acquisition and conditioning unit collects the signal of the wind turbine, and its output end is connected to the input end of the A/D conversion circuit unit; the single-chip microcomputer and the A/D conversion The circuit unit, the USB interface circuit unit, the data storage circuit unit and the fixed circuit unit are bidirectionally connected; the USB interface circuit unit is bidirectionally connected to the host computer with the chaos prediction program. 3. according to claim 1, a kind of DFIG operating state chaos prediction system based on phase space reconstruction, it is characterized in that the signal of the said acquisition wind turbine is to gather the wind turbine gear box driving side bearing temperature signal, wind generator winding The maximum temperature signal, the average speed signal of the rotor of the wind turbine and the active power parameter signal of the wind turbine. 4. according to the said a kind of DFIG operating state chaos prediction system based on phase space reconstruction according to claim 2, it is characterized in that said A/D conversion circuit unit is made of conversion chip and peripheral circuit; Wherein said conversion chip adopts CMOS Technology, there is a three-state data output latch in the chip, the input mode is single channel, the conversion time is 100μs, and the power supply voltage is +5V successive approximation 8-bit conversion chip ADC0804; the conversion chip ADC0804 includes pin DB0, tube Pin DB1, Pin DB2, Pin DB3, Pin DB4, Pin DB5, Pin DB6, Pin DB7, Pin /WR, Pin /RD, Pin /CS, Pin VIN(+), Pin Pin VIN(-), pin CLK-IN, pin CLK-R and pin Vref/2, and according to pin DB0, pin DB1, pin DB2, pin DB3, pin DB4, pin DB5 , pins DB6, pins DB7, pins/WR, pins/RD, pins/CS are connected to the single-chip microcomputer chip in a waiting delay mode; the peripheral circuit is composed of a capacitor C28, a resistor R32, two resistors R33, The capacitor C29 and the power supply VCC are composed; the pin VIN (+) receives the signal processed by the signal conditioning circuit through the capacitor C28 and a resistor R33, and the connection point of the capacitor C28 and the resistor R33 is connected to the pin VIN (-) to be grounded together. The differential voltage analog input mode is adopted; the pin CLK-R is grounded through another resistor R33 and capacitor C29, and the pin CLK-IN is connected to the connection point of the resistor R33 and capacitor C29; the pin Vref/2 is connected to the connection point of the resistor R33 and capacitor C29; R32 is connected to the power supply VCC. 5. A kind of DFIG operating state chaos prediction system based on phase space reconstruction according to claim 2, characterized in that said single-chip microcomputer adopts Freescale's single-chip microcomputer MC9S12DP256. 6. A kind of DFIG operating state chaos prediction system based on phase space reconstruction according to claim 2, characterized in that said data storage circuit unit adopts the DS1225 chip of Dallas Company. 7. A DFIG operating state chaos prediction system based on phase space reconstruction according to claim 2, characterized in that the USB interface circuit unit adopts the CH372 chip of Nanjing Qinheng Electronics. 8. A kind of DFIG operating state chaos prediction system based on phase space reconstruction according to claim 2, characterized in that the timing circuit unit adopts a PIC16F716 device with a watchdog. 9. A working method for a chaotic prediction system of a wind power system, characterized in that it comprises the following steps: (1) The timing circuit unit sets the collection interval timing time, and the signal collection and conditioning circuit unit collects the driving side bearing temperature of the wind turbine gearbox, the maximum temperature of the generator winding, the average rotor speed and the active power signal of the generator in real time; (2) The signal collected in step (1) is filtered and self-calibrated by the signal acquisition and conditioning circuit unit and the A/D conversion circuit unit, and the driving side bearing temperature of the unit gearbox, the maximum temperature of the generator winding, the average rotor speed and Generator active power parameters are input into the data storage circuit unit; (3) Through the USB interface circuit, the temperature of the driving side bearing of the wind turbine gearbox, the maximum temperature of the generator winding, the average speed of the rotor and the active power of the generator after processing in step (2) are transmitted to the host computer equipped with the chaos prediction program; ⑷Using the chaotic prediction method in the host computer for parameter calculation and processing. 10. according to claim 9, a kind of working method for the chaos prediction system of wind power system, it is characterized in that the chaos prediction method in the said step (4) adopts the method detection chaos based on phase space reconstruction, is made of the following steps : ①Use the CC method to determine the embedding dimension and time delay: when the wind turbine is running, for a certain state parameter time series x={ _xi },i=1,2,...,N, if the embedding dimension is m, the time delay is τ, then the reconstructed phase space is X={X _i }, and X _i is the phase point in the m-dimensional phase space: ${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&CenterDot;</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}}<span\; class="notranslate">\; ]</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1,2</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; \&Center\; Dot;</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$ Then the associated integral of the embedded time series is $\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}\underset{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$ Where M=N-(m-1)τ, d _ij =||x _i -x _j || _∞ is the ∞ norm; θ is the Heaviside function, and its expression is $\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right.$ The correlation integral is a cumulative distribution function that expresses the probability that the distance between any two points in the phase space is less than r. Additionally define the test statistic for x={ _xi }: $\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; C</span>}}^{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&tau;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$ S(m, N, r, τ) reflects the autocorrelation characteristics of the time series, the optimal time delay takes the first zero point of S(m, N, r, τ), and the point in the reconstructed phase space is closest to Evenly distributed, the reconstructed attractor orbit is completely expanded in the phase space; ② Use the largest Lyapunov exponent to identify the chaotic characteristics of DFIG: from the time delay τ and embedding dimension m calculated in step ①, apply the small data volume method to calculate the bearing temperature on the driving side of the wind turbine gearbox, the maximum temperature of the generator winding, and the average rotor speed and the maximum Lyapunov exponents of the four state parameter time series of generator active power, if the four maximum Lyapunov exponents are greater than 0, it means that there is DFIG chaotic characteristics, and the next step is predicted; ③Using the weighted first-order local method to predict DFIG: Reconstruct the phase space of the delay time τ and dimension m obtained in step ①, and apply the weighted first-order local method to the wind turbine gearbox drive side bearing temperature, the maximum temperature of the generator winding, and the average rotor speed The time series of the four state parameters and the active power of the generator are respectively predicted; Let the adjacent point of the central point (that is, the starting point of prediction) X _k be X _ki , the distance between two points is d _i , let d _m be the minimum value in d _i , and the weight of point X _ki is: ${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; )</span>}}$ Generally take l=1, then the weighted first-order local linear fitting is X _ki+1 =ae+bX _ki , e=[1,1,…,1] ^T . Solve according to the method of weighted least squares , get the prediction formula X _k+1 =a+bX _k , construct the next center point and its adjacent points, repeat step 3 for further prediction, until the abnormal situation of the unit is predicted and the prediction stops.

Images