Distributed photovoltaic power station intelligent operation early warning method

Abstract

The invention discloses a distributed photovoltaic power station intelligent operation early warning method, and relates to the distributed photovoltaic technology field, the method collects the real-time output power of each photovoltaic power station, sets a self-adaptive power abnormity threshold value for each photovoltaic power station, and when the output power is less than the threshold value, the self-adaptive power abnormity threshold value is set for each photovoltaic power station. And if yes, preliminarily judging that the photovoltaic power station has a fault. And calculating the similarity of the characteristic attributes of the surrounding normal photovoltaic power stations and the fault photovoltaic power station, and adding the normal photovoltaic power stations with high similarity to the alternative comparison photovoltaic power stations. Adding a distance factor into the alternative comparison photovoltaic power stations for screening, selecting an optimal photovoltaic power station by utilizing a Pareto optimal solution method, adding the optimal photovoltaic power station into an actual comparison photovoltaic power station, calculating a correlation coefficient of output power variation of the actual comparison photovoltaic power station and the fault photovoltaic power station, and obtaining by utilizing the correlation coefficient that if the photovoltaic power station is caused by cloud layer shielding, no early warning is performed, and if the photovoltaic power station is caused by cloud layer shielding, no early warning is performed. If yes, early warning is carried out, and false early warning is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed photovoltaic technology, and particularly to an intelligent operation warning method for a distributed photovoltaic power station. Background Art

[0002] With the continuous growth of the global demand for clean energy, distributed photovoltaic power stations have been widely applied and developed. However, during their operation, they face many challenges. Especially when a fault occurs in a photovoltaic power station, the ability to give a timely warning plays an important role in subsequent timely maintenance.

[0003] Traditional intelligent warnings for photovoltaic power stations are all based on machine learning. Its core lies in using historical operation data to construct a prediction model. The data involved covers various information such as power output data and environmental monitoring data. Then, through machine learning algorithms such as support vector machines and neural networks for model training, these algorithms can deeply mine the internal laws and characteristics of the data, so as to construct an accurate model representation under normal operating conditions. During actual operation, newly collected real-time data is input into the trained model. The model deeply analyzes and compares the data based on the learned normal mode characteristics. Once it is found that the change trend of the data significantly deviates from the range set by the normal mode, for example, the power output data shows abnormal fluctuations and continuously deviates from the predicted value, it can be inferred that there may be a potential fault in the photovoltaic power station.

[0004] This method of giving a warning by constructing a machine learning model prediction is extremely prone to false warnings. When clouds block sunlight and cause the output power of a photovoltaic power station to drop, the error with the value predicted based on machine learning is relatively large, and the warning device starts to alarm, resulting in false warnings and wasting human and material resources. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an intelligent operation warning method for a distributed photovoltaic power station, which solves the problem of false warnings of the warning device caused by the decrease in output power of the photovoltaic power station due to cloud occlusion.

[0006] To achieve the above object, the present invention is realized through the following technical solutions: An intelligent operation warning method for a distributed photovoltaic power station includes the following steps:

[0007] S1. Real-time collect the output power Pnow of each photovoltaic power station, and obtain a power anomaly threshold Th at this time according to the adaptive power anomaly threshold. If Pnow < Th, it is initially determined that a fault has occurred in the photovoltaic power station;

[0008] S2. For the normal photovoltaic power stations around the faulty photovoltaic power station, find out the common characteristic attributes with the faulty photovoltaic power station, calculate the similarity between the characteristic attributes of the two, and select a suitable photovoltaic power station as a backup comparison photovoltaic power station according to the similarity;

[0009] S3, further processing the standby photovoltaic power station using the Pareto optimal solution method to obtain an actual comparison photovoltaic power station that is more consistent with the faulty photovoltaic power station;

[0010] S4. Calculate the output power changes of the actual photovoltaic power station and the faulty photovoltaic power station one by one, and obtain the correlation coefficient between them. According to the correlation coefficient, determine whether the fault is caused by natural reasons or equipment problems.

[0011] As a further solution of the present invention, a method for setting an adaptive power abnormality threshold is:

[0012] The average output power Parg(i) of the photovoltaic power station in the i-th season is calculated based on the historical operation data of the photovoltaic power station, and the standard deviation is Pstd(i). The initial power threshold of the i-th photovoltaic power station is calculated according to the formula Thorg=Parg(i)-k*Pstd(i);

[0013] The current light intensity is collected as l cur, and the historical average light intensity is l arg. The corrected power threshold is obtained according to the formula Thcorr = Thorg*(l cur / l arg);

[0014] The actual output power of the current photovoltaic power station collected is Pcur, and the historical average output power under the same weather conditions is Ph is. The performance coefficient of the current photovoltaic power station is obtained according to the formula SPF = Pcur / Ph is;

[0015] The final adaptive power abnormality threshold is obtained according to the formula Th=Thorg*(l cur / l arg)*SPF.

[0016] As a further solution of the present invention, the characteristic attributes include the number of historical failures of the photovoltaic power station, the average maintenance time after the failure, and the capacity of the power station.

[0017] As a further solution of the present invention, the specific method for calculating the similarity between feature attributes is:

[0018] Normalizing the characteristic values ​​in the characteristic attributes of the faulty photovoltaic power station and the surrounding photovoltaic power stations;

[0019] The normalized characteristic attribute vector of the faulty photovoltaic power station is A = (a1, a2, a3), and the characteristic attribute vector of the surrounding j-th photovoltaic power station is Bj = (bj1, bj2, bj3);

[0020] The similarity of characteristic attributes between the faulty PV power station and the jth surrounding PV station is calculated according to the formula R(A,Bj)=(A*Bj) / (||A||*||Bj||).

[0021] As a further solution of the present invention, if R(A, Bj)>nbd, the j-th surrounding photovoltaic power station is added to the standby comparison photovoltaic power stations; if R(A, Bj)<=nbd, it is not added, where nbd represents the similarity threshold of characteristic attributes.

[0022] As a further solution of the present invention, the specific method for further processing the standby comparison photovoltaic power stations by using the Pareto optimal solution method is as follows:

[0023] Calculate that the number of standby comparison photovoltaic power stations is n, the distance from the faulty photovoltaic power station is d = [d1, d2,..., dn], and the similarity of characteristic attributes with the faulty photovoltaic power station is r = [r1, r2,..., rn];

[0024] Create an n×n comparison matrix M, where the element m pq is used to compare photovoltaic power station p and photovoltaic power station q, p∈[1, n], q∈[1, n];

[0025] According to M and the formula Determine the number of times each photovoltaic power station is dominated by other photovoltaic power stations;

[0026] If Np<mi u, add the p-th standby comparison photovoltaic power station to the actual comparison photovoltaic power stations; if Np>=miu, do not add it, where mi u represents the threshold of the number of domination times.

[0027] As a further solution of the present invention, the specific method for obtaining the element m pq in M by comparing photovoltaic power station p and photovoltaic power station q is as follows:

[0028] If dq<dp and rq>rp, then m pq =1, indicating that photovoltaic power station q is superior to photovoltaic power station p; otherwise, m pq =0.

[0029] As a further solution of the present invention, the specific steps for calculating the correlation coefficient between the output power change of the actual comparison photovoltaic power stations and the faulty photovoltaic power station are as follows:

[0030] Take out photovoltaic power station s (s is the photovoltaic power station number) one by one from the actual comparison photovoltaic power stations, and calculate its output power change as △Ps(t);

[0031] Calculate the output power change of the faulty photovoltaic power station as △Pe(t);

[0032] According to the formula Calculate the correlation coefficient between the output power change of the faulty photovoltaic power station and the actual comparison photovoltaic power stations;

[0033] Among them, represents the average value of △Pe(t), represents the average value of △Ps(t), and T is the number of sampling points within the comparison time period.

[0034] As a further solution of the present invention, if r(e, s) > U, it indicates that the output power change trends of the faulty photovoltaic power station and the actual comparison photovoltaic power station s are highly consistent, and it is concluded that the decrease in the output power of the faulty photovoltaic power station is caused by cloud shading. If r(e, s) < U1, it indicates that there is a large difference in the output power between the faulty photovoltaic power station and the actual comparison photovoltaic power station s, and it is concluded that the decrease in the output power of the faulty photovoltaic power station is caused by a malfunction of the equipment of the photovoltaic power station itself. Among them, U and U1 are correlation coefficient thresholds.

[0035] As a further solution of the present invention, the number of actual comparison photovoltaic power stations with r(e, s) > U calculated is f, and the number of actual comparison photovoltaic power stations with r(e, s) < U1 calculated is h. If (h / (f + h)) > η, it indicates that the fault is caused by the abnormal equipment of the photovoltaic power station itself and needs to be warned. If (f / (f + h)) > η, it indicates that the fault of the photovoltaic power station is caused by cloud shading and does not need to be warned. Among them, η is the decision threshold.

[0036] The present invention provides an intelligent operation warning method for a distributed photovoltaic power station, which has the following beneficial effects compared with the prior art:

[0037] (1) When initially judging the fault of the photovoltaic power station in this application, an adaptive power abnormality threshold is set. This threshold fully considers the influence of the current photovoltaic power station in three aspects: season, weather, and performance, making the initial fault judgment based on the output power of the photovoltaic power station more accurate;

[0038] (2) By comparing the current faulty photovoltaic power station with the surrounding normal photovoltaic power stations in this application, it can more accurately determine whether the decrease in the output power of the faulty photovoltaic power station is caused by cloud shading or the reason of its own equipment. If it is caused by cloud shading, no warning is given. If it is the reason of the equipment of the photovoltaic power station itself, a warning is given, effectively reducing the probability of false warnings. Description of the Drawings

[0039] Figure 1 It is a flowchart of the present invention. Detailed Embodiments

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0041] As Figure 1 , this application provides a method for intelligent operation warning of a distributed photovoltaic power station, including:

[0042] S1. Real-time collect the output power Pnow of each photovoltaic power station, and obtain a power anomaly threshold Th at this time according to the adaptive power anomaly threshold. If Pnow < Th, it is initially determined that the photovoltaic power station has a fault;

[0043] This adaptive power anomaly threshold can be adaptively adjusted according to the historical operation data, weather conditions and system performance of the photovoltaic power station;

[0044] First, consider the influence of seasons. Set the initial power threshold according to the season where the photovoltaic power station is located:

[0045] Collect the historical operation data of the photovoltaic power station to calculate the average output power Parg(i) of the photovoltaic power station in the i-th season, and the standard deviation is Pstd(i). Calculate the initial power threshold of the i-th photovoltaic power station according to the formula Thorg = Parg(i) - k * Pstd(i), where k is an adjustment coefficient and is set according to the actual situation of the photovoltaic power station;

[0046] Then, consider the influence of the current weather conditions. For example, when it is cloudy or rainy, the light intensity weakens, and the output power of the power station will decrease accordingly. Therefore, when dynamically adjusting the power threshold, it is necessary to consider the current weather conditions and correct the initial power threshold through the light intensity:

[0047] Collect the current light intensity as l cur and the historical average light intensity as l arg, and obtain the corrected power threshold according to the formula Thcorr = Thorg * (l cur / l arg);

[0048] Finally, consider the performance coefficient of the photovoltaic power station. This coefficient is used to measure the degree of decline of the current performance relative to its initial design or historical performance, and can further improve the corrected initial power threshold:

[0049] Collect the actual output power Pcur of the current photovoltaic power station, and the historical average output power Ph i s under the same weather conditions. Obtain the performance coefficient of the current photovoltaic power station according to the formula SPF = Pcur / Ph i s;

[0050] The final adaptive power abnormal threshold formula is Th=Thorg*(l cur / l arg)*SPF;

[0051] For example, assuming that the historical operation data of a photovoltaic power station shows that under spring conditions, the average output power of the power station is 100kW and the standard deviation is 10kW, and according to the actual situation of the power station and the warning requirements, the adjustment coefficient k is set to 1.5, then the initial power threshold Thorg = 100-10*1.5 = 85kW;

[0052] On a certain day, the light intensity in the area where the photovoltaic power station is located is 600W / m 2 , while the historical average light intensity is 800W / m 2 , the corrected power threshold is Thcorr = (600 / 800) * 85 = 63.75 kW;

[0053] At the same time, the actual output power of the photovoltaic power station is currently 75kW, and the historical average output power under the same weather conditions is 90kW, and the final adjusted power threshold Th is Th = 63.75*(75 / 90) = 42.45kW;

[0054] If the real-time output power of the photovoltaic power station is lower than 42.45kW at this time, it is preliminarily determined that the photovoltaic power station has failed, and whether to activate the early warning system needs further confirmation;

[0055] S2. For normal photovoltaic power stations around the faulty photovoltaic power station, find out the characteristic attributes shared by the normal photovoltaic power station and the faulty photovoltaic power station, calculate the similarity between the characteristic attributes of the two, and select a suitable photovoltaic power station as a backup comparison photovoltaic power station based on the similarity;

[0056] Finding out the normal photovoltaic power stations around the faulty photovoltaic power station is mainly to further identify the faulty photovoltaic power station by using the normal photovoltaic power stations around it, so as to obtain more accurate early warning;

[0057] Find out the characteristic attributes shared by the faulty PV power station, including the number of historical failures of the PV power station, the average maintenance time after a failure, and the capacity of the power station;

[0058] Perform Z-score normalization on the characteristic values ​​in the characteristic attributes of the faulty photovoltaic power station and the normal photovoltaic power station;

[0059] The specific formula for Z-score normalization is X_norm = (Xorg-mean) / std, where X_norm is the normalized data, Xorg is the original data, mean is the mean of the original data, and std is the standard deviation of the original data;

[0060] The obtained characteristic attribute vector of the normalized faulty photovoltaic power station is A = (a1, a2, a3), and the characteristic attribute vector of the j-th photovoltaic power station around it is Bj = (bj1, bj2, bj3);

[0061] Then, use the formula R(A, Bj) = (A * Bj) / (||A|| * ||Bj||) to calculate the similarity of characteristic attributes between the faulty photovoltaic power station and the j-th photovoltaic power station around it;

[0062] If R(A, Bj) > nbd, it indicates that the current normal photovoltaic power station is highly similar to the faulty photovoltaic power station in terms of overall structure and usage, and is suitable as a comparison power station for the faulty photovoltaic power station. Therefore, add the j-th photovoltaic power station to the standby comparison photovoltaic power stations. If R(A, Bj) <= nbd, it indicates that there are significant differences between the current normal photovoltaic power station and the faulty photovoltaic power station in terms of overall structure and usage, and forced addition for comparison is likely to cause misjudgment. Therefore, it is not added. Here, nbd represents the threshold of characteristic attribute similarity.

[0063] S3. Further process the standby photovoltaic power stations using the Pareto optimal solution method to obtain a more suitable actual comparison photovoltaic power station for the faulty photovoltaic power station;

[0064] The Pareto optimal solution is in a multi-objective optimization problem. For a set of solutions, if there is no other solution that makes all objectives better simultaneously, then this set of solutions is the Pareto optimal solution. In the problem of selecting an actual comparison photovoltaic power station, the goal is to find a photovoltaic power station with a short distance and high similarity of characteristic attributes. When a photovoltaic power station is Pareto optimal, it means that there is no other photovoltaic power station with a closer distance and higher similarity of characteristic attributes;

[0065] Calculate that the number of standby comparison photovoltaic power stations is n, the distance from them to the faulty photovoltaic power station is d = [d1, d2,..., dn], and the similarity of characteristic attributes with the faulty photovoltaic power station is r = [r1, r2,..., rn];

[0066] Create an n×n comparison matrix M, where the element m pq is used to compare photovoltaic power station p and photovoltaic power station q, p ∈ [1, n], q ∈ [1, n];

[0067] If dq < dp and rq > rp, it means that the distance of photovoltaic power station q is closer than p and the similarity of its characteristic attributes is also higher than p. Therefore, set m pq = 1, indicating that photovoltaic power station q is superior to photovoltaic power station p. If photovoltaic power station q is weaker than p in either distance or similarity of characteristic attributes, then set m pq = 0

[0068] According to M and the formula Determine the number of times each PV power station is dominated by other PV power stations;

[0069] If Np < mi u, it indicates that the number of times the p-th PV power station is dominated by other PV power stations is small. This PV power station is not only close to the faulty PV power station (being close means a higher similarity in the surrounding environment, greatly excluding the influence of the environment on the PV power station) but also very similar in overall performance and usage. Therefore, the p-th backup comparison PV power station can be added to the actual comparison PV power stations. If Np >= mi u, it indicates that the number of times the p-th PV power station is dominated by other PV power stations is large and its performance is worse than that of other PV power stations. Therefore, it is not added. Here, mi u represents the domination times threshold.

[0070] For example, assume there are n = 5 backup comparison PV power stations, and their distances from the faulty PV power station are d = [10, 12, 8, 15, 9], and the similarity of characteristic attributes is r = [0.92, 0.95, 0.8, 0.96, 0.93];

[0071] Construct a 5×5 comparison matrix M. Compare backup comparison PV power stations 1 and 2. d1 = 10, d2 = 12, r1 = 0.92, r2 = 0.95. Since d2 > d1 and r2 > r1, it fails to meet the condition of smaller distance and higher similarity at the same time. Therefore, set m 12 = 0; Then compare backup comparison PV power stations 1 and 5. d1 = 10, d5 = 9, r1 = 0.92, r5 = 0.93. Since d5 < d1 and r5 > r3, it meets the condition of smaller distance and higher similarity at the same time. Therefore, set m 15 = 1; According to such rules, the constructed comparison matrix M is

[0072]

[0073] For backup comparison PV power stations 1, 2, 3, 4, 5, the number of times they are dominated is 0, 0, 0, 0, 1 respectively. Set mi u = 1. Add backup comparison PV power stations 1, 2, 3, 4 to the actual comparison PV power stations and remove the backup comparison PV power stations;

[0074] S4. Calculate the output power change of each actual comparison PV power station and the faulty PV power station one by one, and obtain the correlation coefficient between them. Judge whether the fault is caused by cloud occlusion or its own equipment problem according to the correlation coefficient;

[0075] Take out the PV power station s (s is the PV power station number) from the actual comparison PV power stations one by one, and calculate its output power change as △Ps(t);

[0076] Calculate the output power change of the faulty PV power station as △Pe(t);

[0077] The change in output power refers to calculating the change in output power between two consecutive measurements taken at 5 - second intervals within a fixed period of time.

[0078] According to the formula Calculate the correlation coefficient between the change in output power of the faulty photovoltaic power station and that of the actual comparison photovoltaic power station.

[0079] Where represents the average value of △Pe(t), represents the average value of △Ps(t), and T is the number of sampling points within the comparison time period;

[0080] If r(e,s)>U, it means that the change trends of the output power of the faulty photovoltaic power station and the actual comparison photovoltaic power station s are highly consistent. On the basis that the overall structure, usage conditions, and surrounding environment of the faulty photovoltaic power station are highly consistent with those of the actual comparison photovoltaic power station, and the change trends of their output powers are also highly consistent, it can be judged that the decrease in the output power of the faulty photovoltaic power station is caused by cloud shading. If r(e,s)<U1, on the basis that the overall structure, usage conditions, and surrounding environment of the faulty photovoltaic power station are highly consistent with those of the actual comparison photovoltaic power station, the change in the output power between the faulty photovoltaic power station and the actual comparison photovoltaic power station s is relatively large, directly excluding the influence of the surrounding environment and cloud shading on the photovoltaic power station, and it can be more accurately concluded that the decrease in the output power of the faulty photovoltaic power station is caused by a malfunction of the equipment of the photovoltaic power station itself. Here, U and U1 are the correlation coefficient thresholds.

[0081] The correlation coefficients between the change in output power of multiple actual comparison photovoltaic power stations and the faulty photovoltaic power station are calculated and compared with U and U1, and it is determined whether to give an early warning based on the actual proportion:

[0082] The number of actual comparison photovoltaic power stations with r(e,s)>U is f, and the number of actual comparison photovoltaic power stations with r(e,s)<U1 is h. If (h / (f + h))>η, it means that the fault is caused by abnormal equipment of the photovoltaic power station itself and an early warning is needed. If (f / (f + h))>η, it means that the photovoltaic power station fault is caused by cloud shading and no early warning is needed. Here, η is the decision threshold.

[0083] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well - known to those skilled in the art.

[0084] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A distributed photovoltaic power station intelligent operation early warning method, characterized in that: It includes the following steps: S1. Collect the output power Pnow of each photovoltaic power station in real time, obtain a power anomaly threshold Th at this time according to the adaptive power anomaly threshold. If Pnow < Th, it is preliminarily determined that the photovoltaic power station has a fault; S2. For the normal photovoltaic power stations around the faulty photovoltaic power station, find out the common characteristic attributes with the faulty photovoltaic power station, calculate the similarity between the characteristic attributes of the two, and select a suitable photovoltaic power station as the backup comparison photovoltaic power station according to the similarity; S3. Further process the backup photovoltaic power station by using the Pareto optimal solution method to obtain a more suitable actual comparison photovoltaic power station for the faulty photovoltaic power station; S4. Calculate the output power change amount between the actual comparison photovoltaic power station and the faulty photovoltaic power station one by one, and obtain the correlation coefficient between them. Judge whether the fault is caused by cloud occlusion or its own equipment problem according to the correlation coefficient.

2. A distributed photovoltaic power station intelligent operation early warning method according to claim 1, characterized in that: The setting method of the adaptive power anomaly threshold is as follows: Calculate the average output power Parg(i) of the photovoltaic power station in the i-th season according to the historical operation data of the photovoltaic power station, and the standard deviation is Pstd(i). Calculate the initial power threshold of the i-th photovoltaic power station according to the formula Thorg = Parg(i) - k * Pstd(i); Collect the current light intensity lcur and the historical average light intensity larg, and obtain the corrected power threshold according to the formula Thcorr = Thorg * (lcur / larg); The actual output power of the currently collected photovoltaic power station is Pcur, and the historical average output power under the same weather conditions is Phis. Calculate the performance coefficient of the current photovoltaic power station according to the formula SPF = Pcur / Phis; Obtain the final adaptive power anomaly threshold according to the formula Th = Thorg * (lcur / larg) * SPF.

3. A distributed photovoltaic power station intelligent operation early warning method according to claim 1, characterized in that: The characteristic attributes include the historical fault occurrence times of the photovoltaic power station, the average maintenance duration after a fault occurs, and the power station capacity.

4. A distributed photovoltaic power station intelligent operation early warning method according to claim 1, characterized in that: The specific calculation method of the similarity between characteristic attributes is as follows: Normalize the characteristic values in the characteristic attributes of the faulty photovoltaic power station and the surrounding photovoltaic power stations; Obtain the characteristic attribute vector of the faulty photovoltaic power station after normalization as A = (a1, a2, a3), and the characteristic attribute vector of the j-th surrounding photovoltaic power station as Bj = (bj1, bj2, bj3); Calculate the characteristic attribute similarity between the faulty photovoltaic power station and the j-th surrounding power station according to the formula R(A, Bj) = (A * Bj) / (||A|| * ||Bj||).

5. A distributed photovoltaic power station intelligent operation early warning method according to claim 4, characterized in that: If R(A, Bj) > nbd, add the j-th surrounding photovoltaic power station to the backup comparison photovoltaic power station. If R(A, Bj) <= nbd, do not add it, where nbd represents the characteristic attribute similarity threshold.

6. A distributed photovoltaic power station intelligent operation early warning method according to claim 1, characterized in that: The specific method of further processing the backup comparison photovoltaic power station by using the Pareto optimal solution method is as follows: Calculate the number of backup comparison photovoltaic power stations as n, the distance from the faulty photovoltaic power station as d = [d1, d2,..., dn], and the characteristic attribute similarity with the faulty photovoltaic power station as r = [r1, r2,..., rn]; Create an n×n comparison matrix M, where element m pq Used to compare PV power station p and PV power station q, p∈[1,n], q∈[1,n]; According to M and formula Determine the number of times each PV plant is dominated by other PV plants; If Np < miu, the p-th standby comparison photovoltaic power station is added to the actual comparison photovoltaic power station; if Np >= miu, it is not added, where miu represents the threshold of the number of domination times.

7. A distributed photovoltaic power station intelligent operation early warning method according to claim 6, characterized in that: Compare the photovoltaic power station p and photovoltaic power station q to get the element m in M pq The specific method is: If dq < dp and rq > rp, then m pq = 1, indicating that PV power station q is superior to PV power station p, otherwise, m pq = 0.

8. A distributed photovoltaic power station intelligent operation early warning method according to claim 1, characterized in that: The specific steps for calculating the correlation coefficient between the output power change of the actual comparison photovoltaic power station and the faulty photovoltaic power station are as follows: Take out the photovoltaic power station s (s is the photovoltaic power station number) one by one from the actual comparison photovoltaic power station, and calculate its output power change as △Ps(t); Calculate the output power change of the faulty photovoltaic power station as △Pe(t); According to the formula Calculate the correlation coefficient between the output power change of the faulty PV power station and the actual comparison PV power station; in, represents the average value of △Pe(t), It represents the average value of △Ps(t), and T is the number of sampling points in the comparison time period.

9. A distributed photovoltaic power station intelligent operation early warning method according to claim 8, characterized in that: If r(e, s) > U, it means that the output power change trends of the faulty photovoltaic power station and the actual comparison photovoltaic power station s are highly consistent, and it is concluded that the output power drop of the faulty photovoltaic power station is caused by cloud shading; if r(e, s) < U1, it means that the output power of the faulty photovoltaic power station and the actual comparison photovoltaic power station s differ greatly, and it is concluded that the output power drop of the faulty photovoltaic power station is caused by a malfunction of the photovoltaic power station's own equipment, where U and U1 are the correlation coefficient thresholds.

10. A distributed photovoltaic power station intelligent operation early warning method according to claim 9, characterized in that: The number of actual comparison photovoltaic power stations with r(e, s) > U calculated is f, and the number of actual comparison photovoltaic power stations with r(e, s) < U1 is h. If (h / (f + h)) > η, it means that the fault is caused by the abnormal equipment of the photovoltaic power station itself and needs to be warned; if (f / (f + h)) > η, it means that the photovoltaic power station fault is caused by cloud shading and does not need to be warned, where η is the decision threshold.

Images