CN115561575A

CN115561575A - Method for distinguishing electrical abnormal state of offshore wind farm based on multi-dimensional matrix profile

Info

Publication number: CN115561575A
Application number: CN202211104143.8A
Authority: CN
Inventors: 宋杰; 周德生; 徐琴; 宋平; 陆启宇; 鲍伟; 黄华; 杨心刚; 熊祥鸿; 傅坚; 崔秋实; 彭柯; 石立贤; 郭恒
Original assignee: Chongqing University; State Grid Shanghai Electric Power Co Ltd; East China Power Test and Research Institute Co Ltd
Current assignee: Chongqing University; State Grid Shanghai Electric Power Co Ltd; East China Power Test and Research Institute Co Ltd
Priority date: 2022-09-09
Filing date: 2022-09-09
Publication date: 2023-01-03

Abstract

The invention discloses a method for distinguishing electrical abnormal states of an offshore wind plant based on a multi-dimensional matrix profile, which takes common fault states and disturbance phenomena of the offshore wind plant as invention objects, and verifies the effectiveness of the algorithm through theoretical analysis, simulation calculation and actual data correction. According to the method, the EEMD-based offshore wind power plant electrical abnormal state feature extraction strategy effectively decomposes the abnormal waveform into signals with respective time characteristic scales, and provides a basis for identifying the offshore wind power plant abnormal state. The invention reflects the time sequence segment change trend and the frequency difference change based on the similarity measurement mode among the multi-dimensional matrix contour subsequences of the improved DTW algorithm; the embedding of the heartbeat packet mechanism effectively reduces the calculation burden in the actual detection process, does not cause the condition of missing detection, and has wide engineering application value.

Description

Method for distinguishing electrical abnormal state of offshore wind farm based on multi-dimensional matrix profile

Technical Field

The invention relates to the technical field of electricity, in particular to an offshore wind farm electrical abnormal state identification method based on a multi-dimensional matrix profile.

Background

In recent years, the offshore wind power generation technology is developed rapidly in China due to the advantages of cleanness, reproducibility, rich wind power resources and the like. Compared with various limitations such as large noise influence, large occupied area and the like of land wind energy, the offshore wind energy has the potential of large-scale development due to abundant and stable wind resources.

However, the offshore wind field is far away from the coast, the offshore wind turbine has a severe environment and complex working conditions, the wind turbine is easy to break down and disturb, and the unit maintenance cost is higher in difficulty and cost than that of the onshore wind turbine. The economic loss caused by the shutdown is very large.

In addition, since the offshore wind farm is generally not limited by regions and the wind farm is generally large, this means that more wind turbines are provided, and also means that massive state data in the wind farm database needs to be monitored, so that the development of relevant technologies for identifying abnormal states of the offshore wind farm is urgently needed, the health condition of the offshore wind turbines is monitored, the operation and maintenance cost of the offshore wind farm is reduced, and the wind farm benefit is improved.

Disclosure of Invention

The invention aims to provide a method for distinguishing the electrical abnormal state of an offshore wind farm based on a multidimensional matrix profile, which takes the common fault state and disturbance phenomenon of the offshore wind farm as an invention object, verifies the effectiveness of the algorithm through theoretical analysis, simulation calculation and actual data verification, and solves the problem that the identification precision is insufficient due to rare data samples of the offshore wind farm state.

In order to achieve the above object, the present invention provides a method for identifying an electrical abnormal state of an offshore wind farm based on a multidimensional matrix profile, the method comprising:

acquiring data to be identified of an offshore wind farm;

decomposing the data to be identified to obtain an IMF component;

and (4) carrying out abnormal state identification in the constructed abnormal state feature sample library, and matching corresponding abnormal state features.

Optionally, the decomposing the data to be recognized specifically includes:

and decomposing the data to be identified by an EEMD algorithm by utilizing the characteristic that the white Gaussian noise offsets the noise in the multiple averaging processes.

Optionally, the identifying an abnormal state in the constructed abnormal state feature sample library specifically includes:

and performing abnormal state identification in the constructed abnormal state feature sample library through an MDMP algorithm, and determining the shape difference between the sequence to be identified and the abnormal sample feature library sequence by using DTW distance measurement, so that the method has the characteristic of window size self-adaption.

Optionally, the method further comprises: and constructing the abnormal state feature sample library.

Optionally, the constructing the abnormal state feature sample library specifically includes:

acquiring an abnormal state time sequence of an offshore wind farm;

removing noise of the abnormal state time series;

decomposing the denoised abnormal state time sequence into IMF components, reconstructing the IMF components into a multi-dimensional time sequence according to the IMF components, and constructing an abnormal state characteristic sample library.

Optionally, decomposing the denoised abnormal state time series into IMF components by an EEMD algorithm, specifically including: in the IMF decomposition process, gaussian white noise with the same amplitude is added for multiple times, and the statistical characteristic that the Gaussian white noise has uniform frequency distribution is utilized to offset the noise in the multiple averaging process.

Optionally, the method further comprises: and determining the wind field detection state according to the heartbeat packet mechanism.

Optionally, the heartbeat packet mechanism sends a specific signal to the server at regular intervals, the server replies a response signal after receiving the specific signal, and the server determines the communication state of the current client according to the response signal; the heartbeat packet mechanism is provided with a trigger mechanism, and the trigger mechanism is as follows:

wherein, t ₁ For the time of last transmission of the heartbeat packet signal, t ₂ For the current time of sending the heartbeat packet signal, v _b And v _m Respectively a wind speed fluctuation reference value and a wind speed maximum value, U _b And f _b Respectively as voltage and frequency fluctuation reference values;

if the output of s is 1, it indicates that the heartbeat packet signal received by the computer terminal is normal, and at this time, the abnormal state identification system can be in a low-frequency detection state; if the s signal received by the computer terminal is 0, it is proved that a fault may occur in the future, the identification strategy is required to be started, after a period of high-frequency detection, no abnormality is displayed, and if the heartbeat packet signal is 1, the system enters a low-frequency detection state.

Optionally, the method further comprises:

and (3) judging the distance according to the most similar subsequence identified by the multi-dimensional matrix contour algorithm, wherein the identified mechanism is as follows:

wherein the content of the first and second substances,

average of the most similar subsequence distances, { d ₁ ,d ₂ ,…,d _N The average of the most similar subsequence distances at the N true anomalous state events. It can be seen that only when the output is "1", the current object to be identified is the abnormal type matched with the abnormal feature library; otherwise, indicating that the current offshore wind farm is operating normally.

Optionally, the method further comprises:

determining whether the matched abnormal state features meet requirements;

and if the matched abnormal state features meet the requirements, determining that the identified abnormal state is correct.

Optionally, the method further comprises: and if the judged abnormal state characteristics do not meet the requirements, determining that the identified abnormal state is wrong and the data to be identified is not abnormal.

Compared with the prior art, the invention has the following beneficial effects:

the invention provides a method for distinguishing an electrical abnormal state of an offshore wind farm based on a multi-dimensional matrix profile, which takes common fault states and disturbance phenomena of the offshore wind farm as invention objects, verifies the effectiveness of the algorithm through theoretical analysis, simulation calculation and actual data correction and solves the problem of insufficient identification precision caused by rare data samples of the state of the offshore wind farm.

Drawings

FIG. 1 is a flow chart illustrating an EEMD signal decomposition of a wind turbine voltage sequence data of an offshore wind farm provided in one or more embodiments of the present invention;

FIG. 2 is a flow chart of an MDMP based offshore wind farm electrical anomaly status identification provided in one or more embodiments of the present invention;

FIG. 3 is an offshore wind farm model provided in one or more embodiments of the invention;

FIG. 4 is a graph illustrating results of various signal processing methods under a single-phase ground fault waveform provided in one or more embodiments of the invention;

FIG. 5 is an exploded view of a malfunctioning EEMD provided in one or more embodiments of the present invention;

FIG. 6 is a library of electrical anomaly status features for an offshore wind farm provided in one or more embodiments of the present invention;

FIG. 7 illustrates an anomaly status identification for a certain wind turbine of an offshore wind farm based on Matlab simulation, as provided in one or more embodiments of the present invention;

fig. 8 illustrates identification of an electrical anomaly state of a certain wind turbine in an offshore wind farm according to one or more embodiments of the present invention.

Detailed Description

The invention will be further described by means of specific examples in conjunction with the accompanying drawings, which are provided for illustration only and are not intended to limit the scope of the invention.

The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this disclosure and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It is to be understood that although the terms "first," "second," "third," etc. may be used herein to describe structures/information, these structures/information should not be limited by these terms. These terms are only used to distinguish one type of structure/information from another. For example, a first structure/information can also be referred to as a second structure/information, and similarly, a second structure/information can also be referred to as a first structure/information without departing from the scope of the present invention.

In addition, in the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.

The invention provides a method for distinguishing electrical abnormal states of an offshore wind farm based on a multi-dimensional matrix profile, which takes common fault states and disturbance phenomena of the offshore wind farm as an invention object, verifies the effectiveness of the algorithm through theoretical analysis, simulation calculation and actual data correction, and solves the problem that the identification precision is insufficient due to rare data samples of the offshore wind farm state.

The invention takes EEMD as a feature extraction method, and aims to solve the problem that the accuracy of feature extraction is influenced by signal confusion under different modes in order to solve the mode confusion phenomenon frequently occurring in the process of decomposing Intrinsic Mode Functions (IMFs) of EMD. The algorithm adds Gaussian white noise with the same amplitude for multiple times in the IMF decomposition process, and utilizes the statistical characteristic that the Gaussian white noise has uniform frequency distribution to offset the noise in the multiple averaging process. Due to the badness of the working environment of the offshore wind farm and the fluctuation of the wind speed, the EEMD can improve the feature extraction precision, effectively reduce the signal noise and separate the IMF component with the representativeness of the abnormal state for the characteristic extraction of the abnormal state.

The EEMD decomposes time series data under the abnormal state of the offshore wind farm, such as voltage, current, etc., into a single component signal containing only a single instantaneous frequency, i.e., consisting of a finite number of IMF components and a residual component.

Given a certain wind power plant certain wind turbine voltage sequence data X (t) at a certain offshore wind farm, an algorithm flow chart is shown in figure 1, and the specific decomposition steps are as follows:

1) Setting the overall average times F;

2) Adding Gaussian white noise n with standard normal distribution into original data X (t) _i (t) obtaining a new original time series data x for the noise signal added in the ith experiment _i (t), as shown in equation (1):

x _i (t)＝X(t)+n _i (t) (1)

3) Determining x using EMD algorithm _i (t) fitting all local maximum and minimum points by cubic spline interpolation to obtain the ith mean envelope function X _mean(i) (t) of (d). X is to be _i (t) culling the mean envelope function X _mean(i) (t) obtaining an intermediate signal h _i (t) calculated as in equation (2):

h _i (t)＝x _i (t)-x _mean(i) (t) (2)

4) Judgment h _i (t) whether the IMF condition is satisfied: (1) h is _i (t) the maximum values are equal in number or different by 1 from the minimum values; (2) h is _i The average of the envelope defined by the local maxima and the envelope defined by the local minima in (t) is zero. If the condition is not met, repeating the steps 3) to 4); if so, h _i (t) is IMF to C _i (t) amount, and data x _i (t) separating the IMF components to obtain a residual signal m _i (t) is represented by the formula (3):

m _i (t)＝x _i (t)-C _i (t) (3)

5) M is to be _i (t) repeating steps 2) to 4) to obtain all IMF components and a residual r (t), as shown in formula (4):

6) In order to eliminate white noise added in the decomposition process, the IMF is subjected to mean processing to obtain a final IMF component C _i (t) is represented by the formula (5):

the EEMD algorithm is applied to the offshore wind power plant abnormal state signal decomposition, the white noise energy is added for multiple times, the frequency component in the abnormal state is completely reserved, the decomposed IMF reflects the characteristics of different time scales in the abnormal state, and therefore a complete abnormal state characteristic sample library is obtained, and a practical basis is provided for realizing abnormal state identification. Because real offshore wind farm abnormal data are rare, the method utilizes Matlab/Simulink to carry out offshore wind farm abnormal simulation to obtain various abnormal states of the fan, and extracts the characteristics of EEMD after decomposing the abnormal states respectively based on the abnormal states and combines the characteristics to construct an offshore wind farm abnormal state characteristic sample library.

Referring to fig. 1, the present invention provides an abnormal state identification technique based on a Dynamic Time Warping (DTW) multi-dimensional matrix profile (MDMP) algorithm. The method improves the multi-dimensional matrix contour algorithm, improves DTW distance measurement on the basis of the original Euclidean distance, effectively solves the problem that the time difference of identifying sequence trend is neglected in Euclidean distance calculation, and realizes the self-adaption of window size in the algorithm.

The input of the algorithm is a reconstructed multidimensional time sequence T after EEMD decomposes abnormal states, and the output is a multidimensional matrix contour and contour indexes M, M _list . The multi-dimensional matrix outline M represents the distance between the multi-dimensional subsequence in T and the multi-dimensional subsequence in L, and is arranged in ascending order, which means that the similarity degree is from high to low; multidimensional matrix contour index M _list The position of the most similar subsequence matched by T in L is characterized.

Specifically, firstly, initializing an index matrix storing the most similar subsequence in the multi-dimensional matrix outline; then, calculate the distance between the query sequence query and the subsequence at present, and store in the distance matrix D, when calculating the distance between two sequences, the invention abandons the original Euclidean distance, adopt DTW algorithm, the algorithm can utilize dynamic regulation to catch the mapping relation on the trend of two time sequences effectively, not merely solved the adaptivity problem of the sliding window effectively, and reflect trend and shape similarity of dynamic variation amplitude of time sequence, has remedied the time difference of the Euclidean distance to the trend of the identification sequence and drawn insufficiently, its principle is briefly as follows:

given that the solution of two time series x (i), i =1,2, …, m and x (j), j =1,2, …, n, DTW becomes an optimization problem, under the three constraints of boundary, continuity and monotonicity, the normalized path must be from d (1,1) to d (m, n), d is the euclidean distance calculation metric, and then the optimal normalized path cumulative distance r (i, j) is:

DTW＝min{r(m,n)} (2)

therefore, a distance matrix D between the current window and the query is obtained according to the DTW algorithm. Then, D is arranged in a row in an ascending order, and the indexes D arranged in the ascending order are obtained _list . Finally, accumulating and averaging the D arranged according to the ascending order of the columns to obtain a multi-dimensional matrix contour M and an index M _list And obtaining the most similar multi-dimensional subsequence and the position of the most similar multi-dimensional subsequence through the distance minimum value and index judgment, thereby realizing the matching of similar subsequences. The flow of the DTW distance-based multi-dimensional matrix profile algorithm is shown in fig. 2.

In addition, it is noted that, as shown in fig. 2, the algorithm always obtains the minimum distance and index of the most similar subsequence, and the matched result is not the most similar subsequence in the true sense. In order to reduce the error rate of abnormal identification, the invention applies a threshold value judgment algorithm, and carries out logic judgment of '0' and '1' on the most similar subsequence identified by the multi-dimensional matrix contour algorithm through a threshold value, wherein the judgment standard is shown as a formula (3):

in the formula (I), the compound is shown in the specification,

the average value of the most similar subsequence distance is epsilon, a set threshold value is epsilon, the size of the threshold value is determined by data analysis, and the data analysis process is as follows: assuming that the total number of times of identifying abnormal states is V, where the number of times of the abnormal states actually existing is N, calculating an average value of the distance between the most similar subsequences in the N abnormal identified state events, and recording the average value as D, then D = { D = ₁ ,d ₂ ,…,d _n Its threshold is set to: ε = maxD, the minimum of the distance mean of the most similar subsequences of the N true anomalous states. Therefore, it can be seen that the current working condition is the abnormal state only if the subsequence segment with j output of "1", and the accurate matching with the abnormal state type in the sample library is successfully realized, so that the abnormal state type is identified; otherwise, the current observation data is indicated to have no abnormal occurrence.

Because the time of the abnormal state of the wind power plant is far shorter than the time of the normal state, the wind power plant is in the normal state in most of time under the actual condition, and if the computer or related equipment is always in the high-frequency inspection state, a lot of useless computing power is consumed. Therefore, a heartbeat packet mechanism is designed, and calculation power can be distributed according to the state of the current offshore wind farm, so that the calculation cost of a low-probability useless identification area is reduced.

Generally, the heartbeat packet mechanism is to send a specific signal to the server at regular intervals, and if the server receives the specific signal, the server replies a response signal, which mainly serves to make the server know the current communication state of the client. For the design of the heartbeat packet mechanism, the actual condition of the offshore wind farm is considered, compared with a standard heartbeat packet signal, the state of the wind farm is not determined by whether the heartbeat packet signal is received by a computer terminal, but whether the received heartbeat packet signal is abnormal or not is determined by the computer terminal, therefore, the invention sets a trigger mechanism of the heartbeat packet signal, and the abnormal condition trigger condition is shown in a formula (4).

Sudden wind speed or large wind speed: there are large fluctuations in wind speed or in the case of high wind speed for a long time.

Voltage amplitude abrupt change: the voltage amplitude fluctuates obviously;

voltage frequency sudden change: the frequency of the voltage suddenly increases or decreases;

sudden phase current amplitude change: the amplitude of the phase current fluctuates obviously;

weather early warning: the weather forecast shows that severe weather exists in the future for a short time.

Wherein, t ₁ For the time of last transmission of the heartbeat packet signal, t ₂ For the current time of sending the heartbeat packet signal, v _b And v _m Respectively expressed as a reference value of fluctuation of wind speed and a maximum value of wind speed, U _b And f _b Respectively, as voltage and frequency ripple reference values. If the output of s is 1, it indicates that the heartbeat packet signal received by the computer terminal is normal, and at this time, the abnormal state identification system can be in a low-frequency detection state; if the s signal received by the computer terminal is 0, it is proved that a fault may occur in the future, the provided identification strategy needs to be started, after a period of high-frequency detection, no abnormity is displayed, and if the heartbeat packet signal received is 1, the system enters a low-frequency detection state.

The invention adopts simulation software Matlab/Simulink to establish a simulation model of an offshore wind power plant in Shanghai city with the frequency of 50Hz, the wind power plant is composed of 16 double-fed asynchronous machines with the power of 1.5MW, the output voltage of the tail end of a fan is 575V, and the voltage is increased to 25kV by a transformer and transmitted by a 10km submarine medium-voltage submarine cable. After the offshore booster station rises to 120kV, the offshore booster station is transmitted by a 15km high-voltage submarine cable and then is connected with an onshore power grid in a grid mode, as shown in figure 3. In order to obtain the electrical abnormal state data of the offshore wind farm, a large number of simulation experiments are carried out, and common electrical faults of the offshore wind farm under different transition resistances (1-10 omega) at L1-L4 positions are simulated, wherein the common electrical faults include single-phase grounding, two-phase grounding, three-phase grounding and interphase short circuit; the perturbation is set at the L5 position: transient pulses, transient oscillations, voltage dips and current surges, as shown in table 1, total abnormal operating conditions at 52 different parameters. The duration of the fault and disturbance is set to 0.1s (5 periods, the working time of the relay protection device is generally 4-6 periods of the fault), the voltage is used as a collection object, the sampling frequency is 5 kHz, and 100 points are sampled in each period.

The characteristic frequency represented by abnormal waveform time-frequency is extracted by adopting an EEMD algorithm, and a result obtained after a single-phase earth fault voltage signal at the tail end of the fan is subjected to Fourier transform, EMD and EEMD is shown in FIG. 4, wherein the voltage waveform of the fault at the tail end of the fan is shown as a red curve. Due to the selection of the Fourier transform sliding window, the adaptability to a multi-scale signal process or a mutation process is poor, and representative fault frequency characteristics cannot be obtained by observing a decomposed spectrogram. The contrast mode decomposition methods EMD and EEMD, both decompose the original signal into IMF components represented by high and low frequencies. Compared with the EMD, the EEMD can capture more frequency components and refine fault features, so that the EEMD can better decompose abnormal waveforms into signals with respective time characteristic scale representations, the modal aliasing problem of the EMD is solved, and the superiority of the EEMD in extracting electrical abnormal state features of the offshore wind farm is shown.

The electrical abnormal state features extracted by the EEMD algorithm are shown by taking the fan tail end fault and onshore grid-connected point disturbance as examples, and are shown in FIG. 5. Fig. 5 is an EEMD exploded view of a single-phase ground fault and a three-phase ground fault at the tail end of an offshore wind turbine, wherein the decomposed IMFs 1 to 7 are composed of high-frequency and low-frequency components, and are arranged in the order from high frequency to low frequency, and the IMF components contain characteristic information of voltage waveforms in various frequency bands under an electrical abnormal state. And then, taking the IMF decomposed by the abnormal signal EEMD as the fault characteristic in the abnormal state, and reconstructing the fault characteristic into a multidimensional time sequence T1-T7. For this reason, a sample library of electrical abnormal state features of the offshore wind farm is created as shown in fig. 6.

TABLE 1 offshore wind farm Electrical anomaly status types

And performing state identification on the measured data by utilizing an electrical abnormal state feature library established based on the simulation data. Fig. 7 shows a partial characteristic sample library in an abnormal state, including common fault and disturbance types of an offshore wind farm, a red curve in the sample library shows measured abnormal data, and the MDMP-based identification method realizes matching of an identification object with an IMF characteristic component in the characteristic sample library.

As shown in fig. 7, the MDMP identification method is used to identify the attributes of the optimal dimensionality subsequence pair, to successfully match the high and low frequency IMF characteristic components of the fault type 1 in the tested object and the sample library, to determine the fault type of the current tested data, and to implement the electrical abnormal state identification of the offshore wind farm.

In addition, in the practical operation and maintenance of the offshore wind farm, the abnormal state identification is influenced by severe environmental factors such as extreme weather, so that the high-frequency component of the IMF part decomposed by the EEMD is interfered by noise and the state characteristic cannot be accurately represented. In order to verify the effectiveness of the abnormal state identification strategy based on the MDMP in various scenes, the method analyzes the fault position, the fault type, the sizes of different transition resistances and the abnormal state high-frequency signal under the influence of lightning stroke, and realizes the identification of the electrical abnormal state of the offshore wind field based on the IMF combination result selected by the MDMP, as shown in table 2.

TABLE 2 IMF combination results selected based on MDMP under various scenarios

As can be seen from table 2, under the high-frequency noise interference of the electrical abnormal state of the offshore wind farm, for various fault conditions, the combinations of the IMF components selected by the MDMP are different, and the selected IMF components consider both the high-frequency components and the low-frequency components, which indicates that the MDMP identification method comprehensively considers the high, medium and low-frequency components of the IMF, realizes the abnormal identification, and thus fully proves the self-adaptive capability and robustness of the abnormal identification of the provided method.

To further prove the identification accuracy and effectiveness of the method of the present invention, the results are shown in table 3, compared with two identification strategies based on VMD and LS-SVM and WT and Logistic regression.

Results of comparing the precision of the method presented in Table 3 with that of other methods

Because the number of samples used in the modeling process is small, the two algorithms for comparison do not obtain satisfactory identification precision; on the contrary, the identification method provided by the invention does not need a large number of samples for training the identification model, not only is the time consumed short, but also the excellent identification performance is achieved, and the effectiveness and superiority of the method are shown.

The identification strategy provided by the invention can be used for monitoring the large-time-scale steady-state data based on SCADA (supervisory control and data acquisition) and the real-time data of the electric energy quality with higher resolution precision in the monitoring of the actual electrical abnormal state of the offshore wind power. The SCADA monitoring data is taken as an example to verify the engineering applicability of the MDMP identification method, and the SCADA data of a wind power plant at sea of Shanghai city in the actual abnormal operation state of a fan is selected. This data was sampled at a fan capacity of 13MW with an average wind speed of 11.2m/s ten minutes at a sampling frequency of 30s for a total of 20160 data points.

The result of identifying the real abnormal sample data of the offshore wind field is shown in fig. 8, the MDMP identification method is used for realizing the matching of the high-frequency characteristic T1, the low-frequency characteristic T4 and the low-frequency characteristic T7, and the red curve part in the graph shows that the identification object is successfully matched with the characteristic sample library, which shows the effectiveness of the method for processing the real data.

In addition, the MDMP algorithm provided by the invention utilizes DTW as the similarity measurement among the multidimensional time sequences, and compared with the traditional Euclidean distance, the DTW not only can reflect the trend of the dynamic change amplitude of the time sequences and the similarity degree of the shapes, but also makes up the problem of window adaptivity in the algorithm. The results of MDMP calculation of the distance between the multidimensional subsequences are shown in Table 4, where the blue part represents the distance value of the closest subsequence in different dimensions.

TABLE 4 distance calculation of multidimensional subsequences based on MDMP Algorithm

In order to verify the effectiveness of the heartbeat package mechanism designed by the invention, the invention simulates the application of the provided identification strategy to the actual abnormal state detection process through the real SCADA data, and sets v _b And v _m Set to 2m/s and 25m/s, U, respectively _b And f _b Are respectively set to be 0.05U _{Forehead (forehead)} And 0.05f _{Forehead (forehead)} . Table 5 compares the anomaly identification results of the embedded heartbeat packet mechanism and the non-embedded heartbeat packet mechanism, and it can be seen that the detection frequency and time of the identification strategy applying the heartbeat packet mechanism are significantly reduced, specifically reduced by 75.89%, compared with the detection process of the non-embedded heartbeat packet mechanism, and in addition, although the identification frequency is significantly reduced, the identification rate of the application heartbeat packet mechanism is still 100%, which indicates that there is no problem of detection missing due to an abnormal state after the heartbeat packet mechanism is embedded, and the effectiveness of the anomaly identification center heartbeat packet mechanism is verified.

TABLE 5 anomaly identification results for embedded and non-embedded heartbeat packet mechanisms

The invention provides a method for distinguishing the electrical abnormal state of an offshore wind farm based on a multi-dimensional matrix profile, which solves the problem that the identification precision is insufficient due to rare data samples of the state of the offshore wind farm. The method takes common fault states and disturbance phenomena of the offshore wind field as an object of the invention, and the effectiveness of the algorithm is verified through theoretical analysis, simulation calculation and actual data correction. The EEMD-based offshore wind farm electrical abnormal state feature extraction strategy effectively decomposes abnormal waveforms into signals with respective time characteristic scales, and provides a basis for identifying the offshore wind farm abnormal state. The time sequence segment change trend and the frequency difference change are reflected by a multi-dimensional matrix contour subsequence similarity measurement mode based on an improved DTW algorithm; the embedding of the heartbeat packet mechanism effectively reduces the calculation burden in the actual detection process, does not cause the condition of missing detection, and has wide engineering application value.

While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.

Claims

1. The method for distinguishing the electrical abnormal state of the offshore wind farm based on the multi-dimensional matrix profile is characterized by comprising the following steps of:

acquiring data to be identified of an offshore wind farm;

decomposing the data to be identified to obtain an IMF component;

2. The method for distinguishing the electrical abnormal state of the offshore wind farm based on the multi-dimensional matrix profile according to claim 1, wherein the step of decomposing the data to be distinguished comprises the following specific steps:

and decomposing the data to be identified by using an EEMD algorithm by utilizing the characteristic that white Gaussian noise offsets the noise in the process of multiple mean values.

3. The method for identifying the electrical abnormal state of the offshore wind farm based on the multidimensional matrix profile as claimed in claim 1, wherein the identifying the abnormal state in the constructed abnormal state feature sample library specifically comprises:

4. The method for offshore wind farm electrical anomaly status discrimination based on multi-dimensional matrix profiles according to claim 1, characterized in that said method further comprises:

and constructing the abnormal state feature sample library.

5. The method for distinguishing the electrical abnormal state of the offshore wind farm based on the multi-dimensional matrix profile according to claim 4, wherein the constructing the abnormal state feature sample library specifically comprises:

acquiring an abnormal state time sequence of an offshore wind farm;

removing noise of the abnormal state time series;

6. The method for offshore wind farm electrical anomaly status discrimination based on multi-dimensional matrix profiles according to claim 1, characterized in that said method further comprises:

and determining the wind field detection state according to the heartbeat packet mechanism.

7. The method for distinguishing the electrical abnormal state of the offshore wind farm based on the multidimensional matrix profile as claimed in claim 7, wherein the heartbeat packet mechanism sends a specific signal to the server at regular intervals, the server replies a response signal after receiving the specific signal, and the server determines the current communication state of the client according to the response signal; the heartbeat packet mechanism is provided with a trigger mechanism, and the trigger mechanism is as follows:

if the output of s is 1, it indicates that the heartbeat packet signal received by the computer end is normal, and at this time, the abnormal state identification system can be in a low-frequency detection state; if the s signal received by the computer terminal is 0, it is proved that a fault may occur in the future, the identification strategy is required to be started, after a period of high-frequency detection, no abnormality is displayed, and if the heartbeat packet signal is 1, the system enters a low-frequency detection state.

8. The method for offshore wind farm electrical anomaly status discrimination based on multi-dimensional matrix profiles according to claim 1, characterized in that said method further comprises:

wherein the content of the first and second substances,

average of the distance of the most similar subsequences, { d } ₁ ,d ₂ ,…,d _N The mean of the most similar subsequence distances under the N true abnormal state events. It can be seen that there is only a transmissionWhen the number of the objects is '1', the current object to be identified is an abnormal type matched with the abnormal feature library; otherwise, indicating that the current offshore wind farm is operating normally.

9. The method for offshore wind farm electrical anomaly status discrimination based on multi-dimensional matrix profiles according to claim 1, characterized in that said method further comprises:

determining whether the matched abnormal state features meet requirements;

10. The method for identifying electrical anomaly status of an offshore wind farm based on a multi-dimensional matrix profile according to claim 9, further comprising:

and if the judged abnormal state characteristics do not meet the requirements, determining that the identified abnormal state is wrong and the data to be identified is not abnormal.