Distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis

Abstract

The invention discloses a distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis. The method comprises the following steps: S1, constructing a distributed photovoltaic power station fault tree model; s2, collecting operation data of the distributed photovoltaic power station in real time; s3, calculating the occurrence probability of each bottom event of the fault tree according to the collected operation data by using a fault tree analysis method; s4, identifying key fault factors according to the bottom event occurrence probability; and S5, making an intelligent operation and maintenance strategy according to the key fault factors. According to the distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis, key fault factors are accurately identified, the operation and maintenance strategy is dynamically adjusted, the fault positioning and repairing efficiency is improved, the operation and maintenance cost is reduced, and the distributed photovoltaic power station is promoted to be upgraded to active intelligent operation and maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of distributed photovoltaic power station operation and maintenance technology, and particularly relates to a distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis. BACKGROUND

[0002] A distributed photovoltaic power station is composed of photovoltaic modules, inverters, combiner boxes, monitoring systems and power transmission lines, etc. Various faults may occur in the operation process of each part of the equipment, thereby affecting the normal power generation of the power station. The traditional operation and maintenance of the distributed photovoltaic power station mainly relies on manual inspection and simple monitoring systems. This operation and maintenance method has many shortcomings.

[0003] Firstly, manual inspection is low in efficiency and high in cost. Since the distributed photovoltaic power stations are widely distributed, manual inspection requires a large amount of manpower, material resources and time cost, and the working intensity of the inspection personnel is large. At the same time, manual inspection is greatly affected by subjective factors of the inspection personnel, and there are situations of missed inspection and misinspection, which makes it difficult to guarantee the accuracy and comprehensiveness of the inspection. For example, in some complex terrain areas, the inspection personnel may not be able to reach some photovoltaic power stations, resulting in that the faults of these power stations cannot be discovered in time.

[0004] Secondly, the existing monitoring systems can only perform simple monitoring and alarm on the operation parameters of the power station, and lack the ability of in-depth analysis and diagnosis of faults. When the monitoring system issues an alarm, the operation and maintenance personnel are difficult to quickly and accurately locate the root cause of the fault, and often need to spend a lot of time to investigate, resulting in a long fault repair time and affecting the power generation efficiency of the power station. For example, when the power generation capacity of the photovoltaic power station decreases, the monitoring system may only issue an alarm of abnormal power, but cannot determine whether it is caused by photovoltaic module failure, inverter failure or other equipment failure.

[0005] Furthermore, the faults of the distributed photovoltaic power station have diversity and complexity. Different fault causes may lead to the same fault phenomenon, and the same fault cause may also exhibit different fault phenomena under different environmental conditions. In addition, with the continuous expansion of the scale of the distributed photovoltaic power station, the amount of operation data generated is also increasing. The traditional operation and maintenance method cannot fully utilize these data, cannot mine valuable information from a large amount of data, and cannot realize intelligent operation and maintenance of the power station.

[0006] Therefore, a new operation and maintenance method is needed to effectively solve the above problems and improve the operation and maintenance efficiency and management level of the distributed photovoltaic power station. SUMMARY

[0007] The application aims to provide a distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis, which can realize efficient intelligent operation and maintenance, improve power generation efficiency and reliability, and reduce operation and maintenance cost by constructing a fault tree model, combining real-time and historical data, diagnosing faults and identifying key factors by using analysis algorithms, and formulating intelligent operation and maintenance strategies accordingly.

[0008] To achieve the above-mentioned purpose, the application provides a distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis, which comprises the following steps:

[0009] S1, constructing a distributed photovoltaic power station fault tree model;

[0010] S2, collecting real-time operation data of the distributed photovoltaic power station;

[0011] S3, using the fault tree analysis method to calculate the occurrence probability of each bottom event of the fault tree according to the collected operation data;

[0012] S4, identifying key fault factors according to the bottom event occurrence probability;

[0013] S5, formulating intelligent operation and maintenance strategies according to the key fault factors.

[0014] 2. The distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis according to claim 1, wherein S1 comprises the following steps:

[0015] S11, collecting historical fault data of multiple distributed photovoltaic power stations in the target area, including basic fault data, operation parameter data and environmental parameter data; preprocessing the collected historical fault data, including data cleaning, removing missing values and abnormal values, and data standardization to uniformly convert different dimension data to a specific interval;

[0016] S12, calculating the occurrence frequency of each type of fault within a certain time period, and screening out high-frequency fault types as important candidate events in the fault tree model:

[0017]

[0018] wherein f F represents the occurrence frequency of the fault type F, n F represents the number of occurrences of the fault type F within the statistical period, and N represents the total number of occurrences of all faults within the statistical period;

[0019] S13, using a decision tree algorithm to analyze the correlation between the fault type and the operation parameter and the environmental parameter;

[0020] S14, determining the fault tree events, including the top event, the intermediate event and the bottom event.

[0021] Preferably, in S11, the basic fault data includes the fault occurrence timestamp t, the fault type F (including photovoltaic module fault, inverter fault, combiner box fault, and transmission line fault), and the power station number Sid where the fault occurred; the operating parameter data includes the output voltage V of the photovoltaic module. module Output current I module The input voltage V of the inverter inverter-in Input current I inverter-in Output voltage V inverter-out Output current I inverter-out The input current I of the combiner box combiner-in Output current; environmental parameter data include ambient temperature T, light intensity S, and humidity H.

[0022] Preferably, the specific steps in S13 are as follows:

[0023] S131. First, divide the historical fault data after S11 preprocessing into feature set A and label set Y; feature set A includes operating parameters and environmental parameters, each row represents a data record, and each column corresponds to a parameter; label set Y is the fault type, and each element corresponds to the fault type F of a data record.

[0024] S132. Using information gain ratio as the metric for selecting the splitting attribute, first calculate the information entropy Ent(D) of dataset D, as follows:

[0025]

[0026] Where |C| represents the number of fault types in the dataset, p i It represents the proportion of samples in the dataset that belong to the i-th type of fault.

[0027] For each attribute 'a', calculate its information gain Gain(D,a) with respect to dataset D, using the following formula:

[0028]

[0029] Where V is the number of possible values ​​for attribute a, and D n It is a subset of samples in dataset D where attribute a takes the value v, |D v | and |D| represent subsets D, respectively. v And the number of samples in dataset D;

[0030] Next, calculate the intrinsic value of attribute a, using the formula IV(a), which is:

[0031]

[0032] Finally, the information gain ratio GainRatio(D, a) of attribute a is obtained, and the formula is as follows:

[0033]

[0034] The attribute with the maximum information gain ratio is selected as the partition attribute of the current node;

[0035] The data set D is taken as a root node, the data set is divided into multiple subsets according to the selected partition attribute, each subset corresponds to a child node, and the process of selecting a partition attribute and dividing a data set is repeated for each child node, and a decision tree is recursively constructed.

[0036] Preferably, in S14, the power generation efficiency E of the distributed photovoltaic power station is lower than a normal threshold value E th The top event is set, and the calculation formula of the power generation efficiency E is as follows:

[0037]

[0038] Wherein, P out is the actual output power of the power station, P in is the theoretical maximum input power; when E < E th , the top event of power generation efficiency reduction is triggered, and it is determined that the top-level fault of the distributed photovoltaic power station occurs;

[0039] The intermediate event is determined by combining the high-frequency fault type screened out in S12 and the correlation between the fault type and the operation parameter and the environmental parameter obtained in S13, and the specific process is as follows:

[0040] First, the high-frequency fault type in the photovoltaic module fault, the inverter fault, the combiner box fault and the transmission line fault is taken as a first-level intermediate event;

[0041] For the photovoltaic module fault, the photovoltaic module output voltage anomaly and the photovoltaic module output current anomaly are taken as second-level intermediate events;

[0042] For the inverter fault, the inverter input voltage anomaly, the inverter input current anomaly, the inverter output voltage anomaly and the inverter output current anomaly are taken as second-level intermediate events;

[0043] The bottom event is determined according to the equipment itself defect, the operation parameter anomaly and the environmental factor, and the correlation between the fault type and each parameter obtained by the decision tree analysis in S13 is further determined, and the specific process is as follows:

[0044] In terms of the equipment itself defect, for the photovoltaic module, the photovoltaic module aging, the photovoltaic module surface damage, the inverter cooling fan damage and the inverter circuit board fault are taken as bottom events;

[0045] In terms of operating parameter abnormalities, the output voltage of the photovoltaic module exceeding the normal range or the input current of the inverter fluctuating too much is taken as the bottom event;

[0046] In terms of environmental factors, the bottom event is taken as the bottom event, such as the high ambient temperature causing the inverter to overheat, the sudden drop in light intensity affecting the power generation of the photovoltaic module, and the high humidity causing the insulation performance of the power transmission line to decrease.

[0047] Preferably, S2 comprises the following steps:

[0048] S21, installing voltage and current sensors at the photovoltaic module array for collecting the output voltage V module and output current I module of the photovoltaic module;

[0049] Installing voltage and current sensors at the input and output sides of the inverter, collecting the input voltage V inverter-in , input current I inverter-in , output voltage V inverter-out and output current I inverter-out of the inverter;

[0050] Installing current sensors at the input and output ends of the junction box, collecting the input current I combiner-in and output current I combiner-out of the junction box, and monitoring the current transmission of the junction box in real time;

[0051] Installing environmental monitoring equipment in the power station area, including temperature sensors, light intensity sensors and humidity sensors, for measuring the ambient temperature T, real-time acquisition of light intensity and monitoring of ambient humidity H;

[0052] S22, using LoRa technology to build a data transmission network to transmit the collected data to the data processing center in real time; the data transmission network includes LoRa device deployment, channel and frequency band setting, LoRa device spreading factor SF and transmission power adjustment, LoRa network ID and key configuration, data transmission and verification, and network monitoring and optimization;

[0053] S23, the data acquisition equipment collects each parameter in real time according to the set sampling frequency, and preliminarily processes the collected data, including analog-to-digital conversion of the collected analog signals, average calculation of n data continuously collected for each collection parameter, obtaining filtered data, time stamp marking of the preliminarily processed data, and recording the time of data collection;

[0054] S24, storing the preliminarily processed and time-stamped data in the local cache, and uploading the data to the database of the data processing center at certain time intervals.

[0055] Preferably, S3 comprises the following steps:

[0056] S31, based on the historical fault data collected in S11, count the occurrence frequency of each bottom event B i under different operating state combinations S j , construct a bottom event probability table, and record the co-occurrence sample number of the bottom event B i and the total sample number of the operating state combination S j ;

[0057] S32, map the real-time operating parameter vector X real collected in S2, including parameters such as photovoltaic module output voltage, current, inverter input and output voltage, current, ambient temperature, and light intensity, to the discretized operating state S current through equal frequency binning or decision tree division rules;

[0058] S33, according to the historical data, calculate the prior probability of the bottom event using the conditional probability formula:

[0059]

[0060] If , use Laplace smoothing method to correct:

[0061]

[0062] where M is the total number of bottom events in the fault tree;

[0063] Combine the decision tree model in S13, input the real-time operating parameter vector X real into the model, and record the probability of the i-th bottom event B i occurring under the current parameters output by the model as P dt (B i |X real ), and use this probability value to weight and correct the prior probability:

[0064] P'(B i )=α·P(B i |S current )+(1-α)P dt (B i |X real );

[0065] where α is the historical data weight coefficient;

[0066] S34, calculate the probability of the occurrence of intermediate events according to the AND gate and OR gate logic relationship, and the calculation formula is as follows:

[0067] ​When the intermediate event is caused by k bottom events B1, B2,..., B k When connected by AND logic, the intermediate event will only occur when the k bottom events occur simultaneously. The probability of occurrence of the intermediate event connected by AND gate P(AND) is calculated as follows:

[0068]

[0069] where P(B i ) represents the probability of occurrence of the i-th bottom event, and ∏ is the multiplication symbol.

[0070] When the intermediate event is caused by k bottom events B1, B2,..., B k When connected by OR logic, the intermediate event will occur as long as any one of the k bottom events occurs. The probability of occurrence of the intermediate event connected by OR gate P(OR) is calculated as follows:

[0071]

[0072] where 1-P(B i ) represents the probability of non-occurrence of the i-th bottom event.

[0073] S35, calculate the top event occurrence probability P(E) by using the minimal cut sets. Suppose there are m minimal cut sets C1, C2,..., C m in the fault tree, the occurrence probability of the l-th minimal cut set is P(C l ). The calculation of the top event occurrence probability P(E) uses the total probability formula:

[0074]

[0075] where C o represents the 0-th minimal cut set, C l represents the l-th minimal cut set, and ∑ 1≤l＜O≤m P(C l ∩C o ) represents the summation of the probabilities of simultaneous occurrence of two different minimal cut sets.

[0076] S36, continuously update historical statistical samples using the sliding time window method, and optimize and adjust the bottom event probabilities. The bottom event B i and the running state S j co-occurrence sample number is dynamically updated as follows:

[0077]

[0078] where n is the original co-occurrence sample number, n is the expired sample, n is the new sample, and n To update the post-sample.

[0079] Preferably, S4 comprises the following steps:

[0080] S41, according to the actual operation and maintenance experience and historical data of the power station, the probability threshold beta of the key degree of the bottom event is set, which is used for preliminary screening of the bottom event with higher probability;

[0081] Set the weight coefficient omega for different types of bottom events i , which reflects the difference degree of the influence of each bottom event on the operation of the power station;

[0082] S42, for each bottom event B i , combined with its occurrence probability P'(B i ) and weight coefficient omega" i , the comprehensive influence value I i is calculated, and the calculation formula is as follows:

[0083] I i = omega" i * P'(B i );

[0084] Wherein, P'(B i ) is the bottom event occurrence probability after S34 correction, through the formula, the comprehensive influence degree of each bottom event on the operation of the power station is quantified;

[0085] S43, the comprehensive influence value I i of all bottom events is sorted in descending order, and the bottom events with the comprehensive influence value greater than the probability threshold beta are screened out, which are the key failure factors;

[0086] According to the actual demand, the top n' bottom events are selected as the key failure factors to be focused on, and a key failure factor list n'={N'1, N'2,... N' n'} is formed, which provides a basis for subsequent development of operation and maintenance strategy;

[0087] S44, combined with the bottom event probability data updated in S36, the comprehensive influence value of the bottom event is recalculated regularly, and the key failure factor list is updated.

[0088] Therefore, the present application adopts the above-mentioned one kind based on fault tree analysis's distributed photovoltaic power station intelligent operation and maintenance method, and the beneficial effects are as follows:

[0089] (1) the present application constructs a fault tree model, combines historical fault data and real-time parameters, accurately identifies key failure factors such as photovoltaic module aging and inverter overtemperature, traces from bottom event to top event, solves the traditional operation and maintenance fault positioning problem, early warns high-risk failure, reduces the shutdown probability, guarantees the stable power generation of the power station, and improves the reliability.

[0090] (2)The application relies on the dynamically updated fault probability and key factor list, collects and transmits data in real time through LoRa technology, quickly matches operation and maintenance work orders in combination with the fault tree algorithm, shortens the fault response and repair time, reduces the labor input, and reduces the operation and maintenance cost.

[0091] (3)The application fully utilizes historical fault data and real-time operation data, correlates the fault and parameter relationship through the decision tree algorithm, dynamically optimizes the probability model through the sliding time window, breaks through the traditional operation and maintenance "data cannot be used" dilemma, mines the fault rules and equipment performance trends from massive data, provides data support for long-term planning of power stations, and promotes the distributed photovoltaic power station from "passive operation and maintenance" to "active intelligent operation and maintenance".

[0092] The technical solutions of the application will be further described below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0093] Figure 1 is the overall flowchart of an embodiment of the intelligent operation and maintenance method of the distributed photovoltaic power station based on the fault tree analysis of the application;

[0094] Figure 2 is the flowchart of constructing the fault tree model of an embodiment of the intelligent operation and maintenance method of the distributed photovoltaic power station based on the fault tree analysis of the application;

[0095] Figure 3 is the flowchart of collecting the operation data of the distributed photovoltaic power station in real time of an embodiment of the intelligent operation and maintenance method of the distributed photovoltaic power station based on the fault tree analysis of the application;

[0096] Figure 4 is the process diagram of calculating the occurrence probability of each bottom event of the fault tree by the fault tree analysis method of an embodiment of the intelligent operation and maintenance method of the distributed photovoltaic power station based on the fault tree analysis of the application. DETAILED DESCRIPTION

[0097] The technical solutions of the application will be further described below with the help of the accompanying drawings and examples.

[0098] Unless otherwise defined, the technical terms or scientific terms used in the application should be understood as the usual meanings understood by those skilled in the art to which the application belongs.

[0099] As shown in Figure 1 , an intelligent operation and maintenance method of a distributed photovoltaic power station based on fault tree analysis, characterized in that it comprises the following steps:

[0100] S1, constructing a fault tree model of a distributed photovoltaic power station, comprising the following steps:

[0101] AsFigure 2 As shown, S11, historical fault data collection and pretreatment: collect historical fault data of multiple distributed photovoltaic power stations in the target area, including basic fault data, operation parameter data and environmental parameter data, the basic fault data includes the timestamp t of fault occurrence, which is used for subsequent time dimension analysis; the fault type F includes photovoltaic component fault, inverter fault, combiner box fault and power transmission line fault, which is used for classifying and identifying the fault; the power station number Sid of fault occurrence is used to distinguish the fault data of different power stations.

[0102] The operation parameter data includes the output voltage V module of the photovoltaic component, the output current I module , which is used to reflect the power generation capacity of the photovoltaic component; the input voltage I inverter-in , the input current I inverter-in , the output voltage V inverter-out , the output current I inverter-out of the inverter, which embodies the working state of the inverter; the input current I combiner-in , the output current of the combiner box, which is used to show the current transmission of the combiner box. The environmental parameter data includes the environmental temperature T, the light intensity S and the humidity H, and the environmental factors have a great influence on the operation of the photovoltaic power station equipment, which needs to be recorded in detail.

[0103] The collected historical fault data is pretreated, including data cleaning, removing missing values and abnormal values, and data standardization, which uniformly converts different dimensional data to a specific interval, facilitating subsequent analysis, and the minimum-maximum standardization formula is adopted in the application:

[0104]

[0105] Wherein, X is the original data, X min and X max are the minimum value and the maximum value of the data feature, respectively, and X norm is the standardized data.

[0106] S12, calculate the occurrence frequency of each type of fault in a certain time period, the formula is:

[0107]

[0108] Wherein, f F represents the occurrence frequency of the fault type F, n F represents the number of occurrences of the fault type F in the statistical period, and N is the total number of occurrences of all faults in the statistical period. Through the formula, the relative frequency of occurrence of different fault types is determined, and the high-frequency fault type is selected as an important candidate event in the fault tree model.

[0109] S13, using a decision tree algorithm to analyze the correlation between the fault type and the operating parameters and the environmental parameters, and the specific process is as follows:

[0110] Data preparation: first, divide the historical fault data preprocessed by S11 into a feature set A and a label set Y; wherein the feature set A includes operating parameters such as the output voltage Vmodule of the photovoltaic module, the output current Imodule, and the like, and environmental parameters such as the ambient temperature T, the light intensity S, and the like, and each row represents a data record, and each column corresponds to a parameter; the label set Y is the fault type, and each element corresponds to the fault type F of a data record.

[0111] Select the division attribute: use information gain ratio as an index for selecting the division attribute:

[0112] First, calculate the information entropy Ent(D) of the data set D, and the formula is:

[0113]

[0114] Wherein, |C| represents the number of types of fault types in the data set, p i is the proportion of samples belonging to the i-th fault type in the data set; for each attribute a, calculate the information gain Gain(D, a) of the data set D, and the formula is:

[0115]

[0116] Wherein, V is the number of possible values of attribute a, D n is the sample subset of attribute a with value v in the data set D, |D v | and |D| represent the number of samples in the subset D v and the data set D respectively.

[0117] Then calculate the intrinsic value of attribute a, and the formula is IV(a), and the formula is:

[0118]

[0119] Finally, the information gain ratio GainRatio(D, a) of attribute a is obtained, and the formula is:

[0120]

[0121] Select the attribute with the maximum information gain ratio as the division attribute of the current node.

[0122] Take the data set D as the root node, divide the data set into multiple subsets according to the selected division attribute, and each subset corresponds to a child node. For each child node, repeat the above process of selecting the division attribute and dividing the data set, and recursively construct the decision tree.

[0123] Construction will stop when one of the following conditions is met: all samples in the subset belong to the same fault type, the attribute set is empty, or the number of samples in the subset is less than a preset threshold N. threshold .

[0124] S14. Identify fault tree events, including top events, intermediate events, and bottom events.

[0125] The apex event is the final result of fault tree analysis, representing the most undesirable fault state for a distributed photovoltaic (PV) power station. Based on the core objectives of power station power generation efficiency and normal operation, it represents the state where the power generation efficiency E of the distributed PV power station falls below the normal threshold E0. th Set as the top event; the formula for calculating power generation efficiency E is as follows:

[0126]

[0127] Among them, P out P represents the actual output power of the power plant. in This is the theoretical maximum input power, calculated based on factors such as illumination and equipment rated parameters. When E < E0 th When a top-level failure is triggered, the power generation efficiency decline event is determined, indicating a top-level failure in the distributed photovoltaic power station. The occurrence of a top-level failure directly affects the normal operation and power generation efficiency of the power station, and is a failure state that needs to be focused on and avoided during operation and maintenance.

[0128] Intermediate events are the direct causes of top events. Intermediate events are determined by combining the high-frequency fault types screened in S12 with the correlations between fault types and operating and environmental parameters obtained in S13. The specific process is as follows:

[0129] First, high-frequency fault types among photovoltaic module faults, inverter faults, combiner box faults, and transmission line faults are classified as first-level intermediate events. For photovoltaic module faults: based on the analysis in S13, the intermediate events are further refined, if... or Abnormal output voltage and abnormal output current of photovoltaic modules are then considered as secondary intermediate events; among them, ΔV and ΔI are the average values ​​of voltage and current when the photovoltaic module is operating normally, and ΔV and ΔI are the set voltage and current fluctuation thresholds.

[0130] For inverter faults: when

[0131]

[0132] or hour,

[0133] The inverter input voltage anomaly, inverter input current anomaly, inverter output voltage anomaly, and inverter output current anomaly are taken as secondary intermediate events, which can more specifically reflect possible factors leading to inverter failure.

[0134] wherein, or is the average value of the input voltage V inverter-in , input current I inverter-in , output voltage V inverter-out , and output current I inverter-out ; and inverter-in , ΔI inverter-in , ΔV inverter-out , and ΔI inverter-out are fluctuation thresholds of corresponding parameters.

[0135] The bottom event is the most basic event in the fault tree, which is the root cause of the occurrence of the intermediate event. The bottom event is determined according to the device defects, abnormal operation parameters, and environmental factors, and further clarified according to the association between the fault type obtained by the decision tree analysis in S13 and each parameter.

[0136] In terms of device defects, the photovoltaic module aging, photovoltaic module surface damage, inverter cooling fan damage, and inverter circuit board failure of the photovoltaic module are taken as the bottom events. In terms of abnormal operation parameters, the output voltage of the photovoltaic module exceeding the normal range, i.e., V module >V max or V module <V min , wherein V max , V min are the upper and lower limits of the voltage normal range; or the inverter input current fluctuation being too large, i.e., , is taken as the bottom event, wherein is the input current fluctuation, and σ th is the set fluctuation threshold.

[0137] In terms of environmental factors, the inverter overheating caused by high environmental temperature, i.e., T>T th , T th is the upper limit of the temperature for normal operation of the inverter, the sudden drop of light intensity affecting the power generation of the photovoltaic module, i.e., S<S th , S th is the threshold of light intensity affecting power generation, and high humidity leading to the decrease of insulation performance of the power transmission line, i.e., H>H th , H th is the threshold of humidity affecting insulation performance, are taken as the bottom events. For example, the decision tree shows that in a high-temperature and high-humidity environment, i.e., T>T th and H>H th .If the inverter is more prone to failure, then the "high temperature and humidity environment" can be used as a bottom event related to inverter failure.

[0138] S2, real-time acquisition of operation data of the distributed photovoltaic power station, including the following steps:

[0139] As shown in Figure 3 S21, deploying corresponding data acquisition sensors and devices at each key equipment and environmental monitoring point of the distributed photovoltaic power station: installing voltage sensors and current sensors at the photovoltaic module array for acquiring the output voltage V module and output current I module of the photovoltaic module. Among them, the voltage sensor adopts a high-precision Hall voltage sensor, which can monitor the change of the output voltage of the photovoltaic module in real time, and the current sensor selects a Rogowski coil current sensor, which can accurately measure the output current.

[0140] Voltage and current sensors are installed at the input and output sides of the inverter to acquire the input voltage V inverter-in , input current I inverter-in , output voltage V inverter-out and output current I inverter-out of the inverter to obtain the working state parameters of the inverter. These sensors all have high precision and high reliability, and can adapt to complex power station operating environments.

[0141] Current sensors are installed at the input and output ends of the bus box to acquire the input current I combiner-in and output current I combiner-out of the bus box to monitor the current transmission of the bus box in real time and ensure accurate acquisition of current data.

[0142] Environmental monitoring devices, including temperature sensors, light intensity sensors and humidity sensors, are installed at appropriate positions in the power station area. The temperature sensor adopts a platinum resistance temperature sensor to accurately measure the environmental temperature T; the light intensity sensor selects a silicon photocell type sensor to obtain the light intensity S in real time; and the humidity sensor adopts a capacitive humidity sensor to accurately monitor the environmental humidity H. These environmental parameters are crucial for analyzing the operating state of the power station.

[0143] S22, constructing a stable and reliable data transmission network to transmit the collected data to the data processing center in real time, specifically:

[0144] A data transmission network is built using LoRa technology to transmit the collected data to the data processing center in real time. Building the data transmission network includes the following aspects:

[0145] First, the deployment of LoRa devices includes the deployment of terminal nodes: at each data collection point of the distributed photovoltaic power station, deploy LoRa terminal node devices, connect the LoRa terminal node with the voltage sensor and current sensor of the photovoltaic module, the inverter parameter sensor, and the environmental monitoring sensor, and ensure that the running data and environmental data collected by the sensor can be accurately obtained. Configure a unique device ID for each LoRa terminal node for identity recognition and data differentiation in the network; for example, assign the LoRa terminal node in the photovoltaic module area an ID of "PV-001", and assign the inverter area an ID of "INV-001", etc., to facilitate subsequent data processing and management.

[0146] Gateway deployment: According to the scale and geographical distribution of the power station, deploy LoRa gateways reasonably; for power stations with large area or complex terrain, multiple gateways can be set up to expand the signal coverage range and ensure that all LoRa terminal nodes can establish stable connections with the gateway. The gateway is usually deployed in a relatively high and open location in the power station, such as the roof of the power station monitoring room or a dedicated signal tower, to reduce signal shielding and improve signal transmission quality. The gateway is connected to the data processing center through wired methods such as Ethernet or optical fiber to realize further data forwarding.

[0147] Second, network parameter configuration includes channel and frequency band setting: according to local radio management regulations and actual application needs, select appropriate LoRa channels and frequency bands. Common LoRa frequency bands include 433MHz, 868MHz (Europe), 915MHz (North America), etc. In the same area, if there are multiple LoRa networks running simultaneously, the channels need to be planned reasonably to avoid channel conflicts; through the management interface of the LoRa gateway, set the channel and frequency band parameters used for communication between the terminal node and the gateway to ensure that all devices transmit data on the same wireless frequency.

[0148] Third, spread spectrum factor and transmission power adjustment: according to the data transmission distance and network load, adjust the spread spectrum factor SF and transmission power of the LoRa device. The spread spectrum factor SF usually has a value range of 6-12. A larger spread spectrum factor can improve the signal's anti-interference ability and transmission distance, but will reduce the data transmission rate; a smaller spread spectrum factor is the opposite. For terminal nodes that are far from the gateway or have severe signal shielding, increase the spread spectrum factor (such as setting it to 10 or 12) and transmission power to ensure reliable data transmission. For nodes that are close, reduce the spread spectrum factor and transmission power to reduce power consumption and interference. Through the configuration tool of the gateway, the spread spectrum factor and transmission power of each terminal node can be set individually.

[0149] Fourth, network ID and key configuration: To protect network security, set a unique network ID for the LoRa network, that is, NetID, and all terminal nodes and gateways need to use the same NetID to join the network; At the same time, configure the network communication key, including the session key (used for data encryption) and the application key (used for application layer data decryption and verification); In the terminal node and gateway device, respectively input the correct NetID, session key and application key, ensure that the data is encrypted during transmission to prevent data from being stolen or tampered with.

[0150] Fifth, data transmission and verification including data packaging and sending: The LoRa terminal node packages the collected operation data and environmental data, adds device ID, timestamp and other information, and generates a data packet according to the format specified by the LoRa protocol. The CRC cyclic redundancy check algorithm is used to calculate the check code of the data packet, and the calculated CRC check code is appended to the end of the data packet. After packaging and check code calculation, the terminal node sends the data packet to the gateway through the LoRa wireless channel according to the set sending period (such as once every 10 seconds).

[0151] Sixth, data reception and forwarding: The LoRa gateway listens to the wireless channel in real time and receives data packets from the terminal node; The gateway parses the received data packet and extracts the device ID, data content and CRC check code; Recalculate the CRC check code of the data packet and compare it with the received check code: if the check codes are consistent, the data transmission is correct, the gateway extracts the effective data (such as photovoltaic module voltage, inverter current, etc.) in the data packet, and forwards the data to the data processing center through the wired network (such as Ethernet) according to the pre-set protocol format; If the check codes are inconsistent, the gateway sends a retransmission request to the terminal node, requiring the terminal node to resend the data packet until the data transmission is correct.

[0152] Seventh, network state monitoring and optimization: During data transmission, the LoRa gateway monitors the network state in real time, including the connection status of the terminal node, signal strength, data transmission rate and packet loss rate, etc. Through the management interface of the gateway or the supporting monitoring software, the network state data is visualized; When the signal strength of a certain terminal node is too low or the packet loss rate is too high, the parameters such as the spreading factor and transmission power of the node can be adjusted remotely, or the maintenance personnel can be notified to check and maintain the node device, to ensure the stable operation of the LoRa data transmission network.

[0153] S23, real-time data acquisition and preliminary processing: the data acquisition equipment collects each parameter in real time according to the set sampling frequency (such as 1 time per second), and performs preliminary processing on the collected data, including analog-to-digital conversion of signals such as voltage, current, temperature, etc. collected, converting them into digital signals for computer processing. Then perform data filtering processing to remove noise interference in the collected data; use the sliding average filtering algorithm to average the n data collected continuously for each collection parameter to obtain the filtered data.

[0154] For example, for the output voltage V module of the photovoltaic module, its filtered value V filter is calculated as follows:

[0155]

[0156] Where V module (t-i) represents the photovoltaic module output voltage value collected at time t-i, and n is the filter window size, which can be adjusted according to actual conditions. The preliminarily processed data is marked with a time stamp to accurately record the data collection time and ensure the time accuracy of the data, providing accurate time dimension information for subsequent data analysis.

[0157] S24, data storage and upload: store the preliminarily processed and time-stamped data in the local cache, and upload the data to the database of the data processing center at certain time intervals, with the embodiment set to upload data every 1 minute. In the data storage process, efficient data storage formats such as CSV, JSON, etc. are used to facilitate data storage and reading. At the same time, to prevent data loss, a data backup mechanism is set locally to regularly backup and store the collected data, ensuring the safety and integrity of the data.

[0158] S3, using fault tree analysis method, calculate the occurrence probability of each bottom event of the fault tree according to the collected operation data, including the following steps:

[0159] As shown in Figure 4 S31, establish a bottom event probability database: based on the historical fault data collected in S11, count the occurrence frequency of each bottom event B i under different operating state combinations S j , construct a bottom event probability table, record the co-occurrence sample number of bottom event B i and operating state combination S j and the total sample number of operating state combination Where B i represents the i-th bottom event (such as high ambient temperature). S j ​represents the discretization interval of the jth operating state combination containing operating parameters and environmental parameters, such as "environmental temperature T≥35℃" and illumination intensity S>800lx.

[0160] S32, real-time state mapping: mapping the real-time operating parameter vector X real collected in S2 to the discretized operating state S current , including parameters such as photovoltaic module output voltage, current, inverter input and output voltage, current, environmental temperature, illumination intensity, etc., through equal frequency binning or decision tree division rule.

[0161] S33, prior probability calculation and correction: according to historical data, the conditional probability formula is used to calculate the prior probability of the bottom event:

[0162]

[0163] If , the Laplace smoothing method is used for correction:

[0164]

[0165] Where M is the total number of bottom events in the fault tree.

[0166] Combined with the decision tree model of S13, the real-time operating parameter vector X real is input into the model, and the probability of the ith bottom event B i occurring under the current parameters output by the model is recorded as P dt (B i |X real ), and the prior probability is weighted and corrected using this result:

[0167] P'(B i )=α·P(B i |S current )+(1-α)P dt (B i |X real );

[0168] Where α is the historical data weight coefficient, usually taken as 0.7.

[0169] S34, logic gate probability calculation: in the fault tree, intermediate events are connected by bottom events through AND gate or OR gate, etc. According to these logical relationships, the probability of the occurrence of the intermediate event is calculated, and the calculation formula is as follows:

[0170] AND gate probability calculation: when the intermediate event is composed of k bottom events B1, B2,..., B kBy and logic connection, it means that only when the k bottom events occur simultaneously, the intermediate event will occur; according to the multiplication principle of probability, the occurrence probability P(AND) of the intermediate event is calculated as follows:

[0171]

[0172] Wherein, P(B i ) represents the probability of occurrence of the i-th bottom event, and ∏ is a multiplication symbol, and the occurrence probability of the intermediate event connected by the AND gate can be obtained by multiplying the occurrence probabilities of the k bottom events.

[0173] For example, if an intermediate event is connected by the AND gate by two bottom events of "dust accumulation on the surface of the photovoltaic module" and "light intensity lower than the threshold value", and the occurrence probabilities of the two bottom events are P(B1) = 0.1 and P(B2) = 0.2 respectively, then the occurrence probability P(AND) of the intermediate event is 0.1 x 0.2 = 0.02.

[0174] Or gate probability calculation: if an intermediate event is connected by k bottom events B1, B2,..., B k By or logic connection, as long as any one of the k bottom events occurs, the intermediate event will occur, and the present application calculates the probability that all bottom events do not occur, and then subtracts the probability from 1 to obtain the occurrence probability P(OR) of the intermediate event connected by the OR gate, and the calculation formula is as follows:

[0175]

[0176] Wherein, (1-P(B i ) represents the probability that the i-th bottom event does not occur, and the occurrence probability of the intermediate event connected by the OR gate can be obtained by multiplying the probabilities that all bottom events do not occur and then subtracting the product from 1. For example, if an intermediate event is connected by two bottom events of "inverter cooling fan damage" and "inverter circuit board failure" through the OR gate, the probabilities that they do not occur are (1-P(B1) = 0.9 and (1-P(B2) = 0.8 respectively, and the occurrence probability P(OR) of the intermediate event is 1-0.9 x 0.8 = 0.28.

[0177] S35, top event verification: the top event represents the most undesirable fault state of the distributed photovoltaic power station, and the occurrence probability P(E) of the top event is calculated by the minimal cut set, assuming that there are m minimal cut sets C1, C2,..., C m , and the probability of each cut set is P(C l ), and the calculation of the occurrence probability P(E) of the top event adopts the total probability formula:

[0178]

[0179] Wherein, C oC l C 1≤l＜O≤m P(C l ∩C o ) represents the sum of the probabilities of two different minimum cut sets occurring simultaneously, because when the cut set probabilities are individually accumulated in advance, the case of two cut sets occurring simultaneously is calculated repeatedly (each is calculated once in P(C l ) and P(C O ), so the repeated value needs to be subtracted to correct the probability calculation result.

[0180] S36, dynamic updating: due to the fact that the operation and failure rules of the power station will change over time, a sliding time window method is used to update the historical statistical samples, and the probability calculation is optimized. In this embodiment, the window size Z is set to 30 days. The co-occurrence sample number of the bottom event B i and the operating state S j is updated according to the following formula:

[0181]

[0182] wherein, is the original co-occurrence sample number, is the expired sample, is the new sample, is the updated sample; by continuously updating the samples, the probability of the bottom event is adjusted in real time, so that the fault tree analysis can reflect the operation status of the power station in real time.

[0183] S4, identifying key failure factors according to the probability of the bottom event, specifically including the following steps:

[0184] S41, setting a probability threshold and a weight coefficient: according to the actual operation and maintenance experience and historical data of the power station, a probability threshold β of the criticality of the bottom event is set, which is used to preliminarily screen out bottom events with a higher probability of occurrence. A weight coefficient ω i is set for different types of bottom events to reflect the difference in the influence of each bottom event on the operation of the power station.

[0185] Specifically: the bottom events of the device itself defect type have a greater impact on the long-term stable operation of the power station, and can be given a higher weight; the weight of the environmental factor type bottom event is relatively low. Wherein, 0≤ω i ≤1, and M is the total number of bottom events in the fault tree.

[0186] S42, calculating the comprehensive influence value of the bottom event: for each bottom event B i , the comprehensive influence value I i is calculated in combination with its probability of occurrence P'(B i ) and weight coefficient ω i , and the calculation formula is as follows:

[0187] I i =ω″ i ×P'(B i );

[0188] wherein P'(B i ) is the bottom event occurrence probability after correction by S34, through the formula, the comprehensive influence degree of each bottom event on the operation of the power station is quantified.

[0189] S43, ranking and screening key failure factors: ranking all bottom event comprehensive influence values I i in descending order, and screening out bottom events with a comprehensive influence value greater than a probability threshold β, which are key failure factors. Meanwhile, according to actual requirements, the top n' bottom events can be selected as key failure factors for attention, forming a key failure factor list N' = {N'1, N'2,..., N' n'}, which provides a basis for subsequent development of operation and maintenance strategies. For example, if the comprehensive influence values of multiple bottom events are calculated, and it is found after ranking that the comprehensive influence values of 7 bottom events are greater than β = 0.05, then the 7 bottom events are determined as key failure factors; if the top 5 are selected, then the top 5 bottom events in terms of comprehensive influence value are taken as core key failure factors.

[0190] S44, dynamically updating the key failure factor list: as time goes by and the operation state of the power station changes, the occurrence probability and influence degree of the bottom events will also change. Therefore, combined with the dynamically updated bottom event probability data in S37, the comprehensive influence values of the bottom events are recalculated regularly, such as once a month, and the key failure factor list is updated to ensure that the operation and maintenance strategy can always target the key failure factors that actually exist at present, improving the accuracy and effectiveness of operation and maintenance.

[0191] S5, developing intelligent operation and maintenance strategies according to key failure factors: establishing an operation and maintenance strategy knowledge base, and matching corresponding operation and maintenance measures according to key failure factors; when the key failure factor is "dust accumulation on the surface of photovoltaic components", triggering a cleaning work order and arranging manual cleaning; when the key failure factor is "overheating of the inverter", triggering a heat dissipation system inspection work order and adjusting the speed of the heat dissipation fan or cleaning the heat dissipation fins; when the key failure factor is "poor line contact", triggering a line inspection work order and checking and repairing the line connection.

[0192] Embodiment one

[0193] I. Scene and target There are multiple distributed photovoltaic power stations in an industrial park, and components, inverters and other equipment have been running for a long time, resulting in frequent failures affecting power generation. Through the method proposed in the present application, accurate fault diagnosis and intelligent operation and maintenance can improve power generation efficiency and reliability.

[0194] II. Specific steps:

[0195] S1, constructing a fault tree model comprises:

[0196] S11, data collection and preprocessing: collect 1-year historical fault data of 5 power stations in the park, including fault timestamp, component, inverter fault type, power station number, and component voltage and current, inverter electrical parameters, and environmental temperature and humidity. Clean the data, remove missing and abnormal values, and use minimum-maximum standardization to unify the dimension.

[0197] S12, screening high-frequency faults: calculate the fault frequency, and screen out components with a fault rate of 35% and inverters with a fault rate of 30% as high-frequency faults as model candidates.

[0198] S13, decision tree analysis correlation: operating and environmental parameters are divided into a feature set and fault types are divided into a label set. The information gain rate is used to divide the attributes, a decision tree is constructed, and it is found that "high environmental temperature + large inverter current fluctuation" easily triggers inverter failure.

[0199] S14, determine the fault tree event: the top event is set to "power generation efficiency <80% or failure downtime"; the first-level intermediate event is component failure and inverter failure; the second-level intermediate event is, for example, component voltage anomaly; and the bottom event includes component aging, inverter fan damage, and high temperature environment causing inverter overheating.

[0200] S2: real-time acquisition of operating data, as follows:

[0201] S21, deploy acquisition equipment: component array with Hall voltage and Rogowski coil current sensors; inverter input / output side with high-precision electrical parameter sensors; bus box with current sensors; power station with platinum resistance temperature, silicon photocell illumination, and capacitive humidity sensors.

[0202] S22, build LoRa network: deploy terminal nodes connected to sensors, assign ID PV-002; gateway is installed on the top of the power station monitoring building and connected to the data center through Ethernet. Configure channel 433MHz, spread factor SF=8, and transmission power, set network ID and encryption key. The terminal sends data every 10 seconds, which is forwarded after verification by the gateway, and the network status is monitored in real time.

[0203] S23, data acquisition and processing: acquire data every second, convert analog signals to digital, and use sliding average filtering with a window size of 5 and timestamp.

[0204] S24, data storage and upload: locally cache data, upload data center database every minute in CSV format, and locally back up to prevent loss.

[0205] S3, calculate the probability of bottom event occurrence;

[0206] S31, Establish probability database: count the co-occurrence frequency of bottom events in historical data, such as component aging and running state combination, such as "temperature 35℃+light 800lx", and construct a probability table.

[0207] S32, Real-time state mapping: map real-time component voltage 380V, temperature 32℃, etc. to discrete states using decision tree rules.

[0208] S33, Prior probability calculation correction: calculate prior probability using conditional probability formula, if 0, then Laplace smoothing correction, combine decision tree output real-time probability weighted correction, historical weight 0.7.

[0209] S34, Logic gate probability calculation: such as "component aging" and "voltage overrun" through AND gate to intermediate event, probability is the product of the probability of both; "inverter fan damage" or "circuit board failure" through OR gate to intermediate event, probability is 1 minus the product of the probability of both not occurring.

[0210] S35, Top event probability calculation: calculate top event probability using total probability formula through minimal cut set.

[0211] S36, Dynamic update probability: set sliding time window 30 days and update samples, adjust bottom event probability, adapt to power station running changes.

[0212] S4, Identify key failure factors;

[0213] S41, Set threshold and weight: set probability threshold 0.2 by experience, device defect class bottom event weight 0.6, environmental factor class weight 0.4.

[0214] S42, Calculate comprehensive influence value: such as "component aging" corrected probability is 0.3, weight is 0.6, comprehensive influence value is 0.3x0.6=0.18.

[0215] S43, Screen key factors: sort by comprehensive influence value, screen out "component aging", "inverter fan damage" and other 5 key failure factors, form a list.

[0216] Table 1 Key failure factor list

[0217]

[0218] S44, Dynamic update list: recalculate comprehensive influence value every month, update key factor list, ensure accurate strategy.

[0219] S5, form intelligent operation and maintenance strategy: match key failure factors and operation and maintenance measures: "component aging" triggers component replacement work order, arranges professional personnel to replace; "inverter fan damage" triggers fan repair work order, overhauls and replaces the fan, responds to the failure quickly, and improves the operation and maintenance efficiency. After the implementation, the fault positioning time is shortened by 60%, the fault repair time is reduced by 50%, the power station power generation efficiency is improved by 8%, the operation and maintenance cost is reduced by 30%, and the intelligent and efficient operation and maintenance of the distributed photovoltaic power station is realized.

[0220] Therefore, the application adopts the above-mentioned distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis, effectively breaks through the traditional operation and maintenance bottleneck, adapts to the complex operation and maintenance demand of the distributed photovoltaic power station, and provides technical support for the intelligent operation and maintenance of the industry.

[0221] Finally, it should be pointed out that: the above examples are only used to illustrate the technical solutions of the present application but not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: it can still modify or equivalently replace the technical solutions of the present application, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis, characterized in that, The method comprises the following steps: S1, constructing a fault tree model of a distributed photovoltaic power station; S2, collecting real-time operation data of the distributed photovoltaic power station; S3, calculating the occurrence probability of each bottom event of the fault tree according to the collected operation data by using a fault tree analysis method; S4, identifying key fault factors according to the occurrence probability of the bottom event; S5, formulating an intelligent operation and maintenance strategy according to the key fault factors. 2.The method of claim 1, wherein, S1 comprises the following steps: S11, collecting historical fault data of a plurality of distributed photovoltaic power stations in a target area, including basic fault data, operation parameter data and environmental parameter data; preprocessing the collected historical fault data, including data cleaning, removing missing values and abnormal values, and data standardization, so as to uniformly convert different dimension data to a specific interval; S12, calculating the occurrence frequency of each type of fault within a certain time period, and screening out high-frequency fault types as important candidate events in the fault tree model: wherein f F represents the frequency of occurrence of the fault type F, n F represents the number of occurrences of the fault type F in the statistical period, and N is the total number of occurrences of all faults in the statistical period; S13, using a decision tree algorithm to analyze the correlation between the fault type and the operation parameter and the environmental parameter; S14, determining the fault tree events, including top events, intermediate events and bottom events. 3.The method of claim 2, wherein, In S11, the basic fault data includes a timestamp t of the fault occurrence, a fault type F including photovoltaic module fault, inverter fault, combiner box fault and transmission line fault and a power station number Sid of the fault occurrence; the operating parameter data includes output voltage V module , output current I module of the photovoltaic module; input voltage V inverter-in , input current I inverter-in , output voltage V inverter-out , output current I inverter-out of the inverter; input current I combiner-in , output current of the combiner box; the environmental parameter data includes environmental temperature T, illumination intensity S, humidity H.

4. The distributed photovoltaic power station intelligent operation and maintenance method based on fault tree analysis according to claim 3, characterized in that, The specific steps of S13 are as follows: S131, first, divide the historical fault data preprocessed in S11 into a feature set A and a label set Y; wherein the feature set A includes operation parameters and environmental parameters, each row represents a data record, and each column corresponds to a parameter; the label set Y is the fault type, and each element corresponds to the fault type F of a data record; S132, using information gain ratio as an index for selecting a division attribute, first, calculate the information entropy Ent(D) of the data set D, the formula is: wherein |C| represents the number of types of failure in the data set, p i is the proportion of samples in the data set that belong to the i-th type of failure. for each attribute a, calculate the information gain Gain(D, a) of the data set D with respect to the attribute a, the formula is: where V is the number of possible values of attribute a, D n is the subset of samples in dataset D with attribute a taking value v, |D v | and |D| represent the number of samples in subset D v and dataset D, respectively. then, calculate the intrinsic value IV(a) of the attribute a, the formula is: finally, obtain the information gain ratio GainRatio(D, a) of the attribute a, the formula is: select the attribute with the maximum information gain ratio as the division attribute of the current node; take the data set D as the root node, divide the data set into a plurality of subsets according to the selected division attribute, each subset corresponds to a child node, for each child node, repeat the process of selecting a division attribute and dividing a data set, and recursively construct a decision tree. 5.The method of claim 3, wherein, In S14, the distributed photovoltaic power station power generation efficiency E is lower than the normal threshold E th Set as the top event; the calculation formula of power generation efficiency E is as follows: wherein P out is the actual output power of the power station, P in is the theoretical maximum input power; when E th , a top event of power generation efficiency reduction is triggered, and it is determined that the top layer fault occurs in the distributed photovoltaic power station; Determine the intermediate events by combining the high-frequency fault types screened out in S12 and the correlation between the fault type and the operation parameter and the environmental parameter obtained in S13, the specific process is as follows: first, take the high-frequency fault types in the photovoltaic component fault, the inverter fault, the combiner box fault and the power transmission line fault as first-level intermediate events; for the photovoltaic component fault, take the photovoltaic component output voltage anomaly and the photovoltaic component output current anomaly as second-level intermediate events; for the inverter fault, take the inverter input voltage anomaly, the inverter input current anomaly, the inverter output voltage anomaly and the inverter output current anomaly as second-level intermediate events; determine the bottom events according to the equipment defects, operation parameter anomalies and environmental factors, and further clarify the bottom events according to the correlation between the fault type and each parameter obtained by the decision tree analysis in S13, the specific process is as follows: In terms of device defects, the aging of the photovoltaic module, the damage of the photovoltaic module surface, the damage of the inverter cooling fan, and the failure of the inverter circuit board are taken as the bottom events; In terms of abnormal operation parameters, the output voltage of the photovoltaic module exceeding the normal range or the input current fluctuation of the inverter being too large are taken as the bottom events; In terms of environmental factors, the over-high environmental temperature leading to overheating of the inverter, the sudden drop of the light intensity affecting the power generation of the photovoltaic module, and the high humidity leading to the decrease of the insulation performance of the power transmission line are taken as the bottom events.

6. The method of claim 5, wherein, S2 includes the following steps: S21, installing voltage and current sensors at the array of photovoltaic modules for collecting the output voltage V module and the output current I module of the photovoltaic modules; Voltage and current sensors are installed at the input side and the output side of the inverter respectively to collect input voltage V inverter-in , input current I inverter-in , output voltage V inverter-out and output current I inverter-out of the inverter. Install current sensor at the input and output end of the bus box, collect the input current I of the bus box combiner-in and output current I combiner-out , real-time monitor the current transmission of the bus box; An environmental monitoring device is installed in the power station area, including a temperature sensor, a light intensity sensor, and a humidity sensor, for measuring the environmental temperature T, real-time acquisition of the light intensity, and monitoring of the environmental humidity H; S22, a data transmission network is built using LoRa technology to transmit the collected data to the data processing center in real time; the data transmission network includes LoRa device deployment, channel and frequency band setting, LoRa device spreading factor SF and transmission power adjustment, LoRa network ID and key configuration, data transmission and verification, and network monitoring and optimization; S23, the data acquisition device collects each parameter in real time according to the set sampling frequency, and preliminarily processes the collected data, including analog-to-digital conversion of the collected analog signals, average calculation of n data continuously collected for each collection parameter to obtain filtered data, and time stamp marking of the preliminarily processed data to record the data collection time; S24, the preliminarily processed and time-stamped data is stored in the local cache, and the data is uploaded to the database of the data processing center at certain time intervals.

7. The method of claim 6, wherein, S3 includes the following steps: S31, based on the historical failure data collected in S11, count the number of each bottom event B i Under different operating state combinations S j , build a bottom event probability table to record the number of co-occurrence samples of bottom event B i and operating state combination S j and the total number of samples of operating state combination ​ S32, mapping the real-time operation parameter vector X collected in S2 to a discrete operation state S real , including parameters such as photovoltaic module output voltage, current, inverter input and output voltage, current, ambient temperature, and light intensity, to a discrete operation state S current ; S33, according to the historical data, the prior probability of the bottom event is calculated using the conditional probability formula: If then the Laplacian smoothing method is used to correct: Wherein, M is the total number of bottom events in the fault tree; The decision tree model combined with S13, the real-time operating parameter vector X real Inputting the model, the i-th bottom event B i The probability of occurrence is denoted as P dt (B i |X real , the probability value is used to weight and correct the prior probability: P'(B i ) = a - P(B i | S current ) + (1 - a)P dt (B i | X real ); Wherein, α is the historical data weight coefficient; S34, the probability of occurrence of the intermediate event is calculated according to the AND gate and OR gate logic relationship, and the calculation formula is as follows: When the intermediate event is caused by k basic events B1, B2,..., B k When connected by AND, the intermediate event will only occur when the k basic events occur simultaneously. The probability of occurrence of the intermediate event connected by AND, P(AND), is calculated as follows: where P(B i ) represents the probability of the occurrence of the ith bottom event, and ∏ is the symbol for the product. If the intermediate event is caused by k basic events B1, B2,..., B k By OR logic connection, the intermediate event will occur as long as any one of the k basic events occurs. The probability of the intermediate event caused by OR connection P(OR) is calculated as follows: where 1 - P(B i ) represents the probability that the ith bottom event does not occur. S35. Calculate the probability P(E) of the top event using minimal cut sets. Suppose the fault tree has m minimal cut sets C1, C2, ..., Cn. m The probability of the occurrence of the l-th minimal cut set is P(C l If the probability of the top event P(E) is calculated using the law of total probability: where C o represents the Oth minimum cut set, C l represents the lth minimum cut set, ∑ 1≤l＜O≤m P(C l ∩C o ) represents the sum of probabilities of simultaneous occurrence of two different minimum cut sets; S36, continuously update historical statistics samples by using sliding time window method, optimize and adjust the bottom event probability, the bottom event B i The co-occurrence sample number of the running state S j The dynamic update formula is: wherein, is the number of original co-occurrence samples, is the expired sample, is the new sample, is the updated sample. 8.The method of claim 7, wherein, S4 includes the following steps: S41, according to the actual operation and maintenance experience and historical data of the power station, the probability threshold β of the key degree of the bottom event is set to preliminarily screen out the bottom events with a higher probability of occurrence; Setting weight coefficients ω for different types of bottom events i , embodying the difference degree of the influence of each bottom event on the operation of the power station; S42, for each bottom event B i , its occurrence probability P'(B i ) and weight coefficient ω" i , its comprehensive influence value I i is calculated, and the calculation formula is as follows: I i = ω" i × P'(B i ); P'(B) = P(B) - P(B) * (P(B) - P'(B)) / P'(B) (1) i ) is the probability of the bottom event after the correction in S34. Through this formula, the comprehensive influence degree of each bottom event on the operation of the power station is quantified. S43、calculate the comprehensive influence value I of all bottom events i Sort in descending order, and filter out the bottom events with the comprehensive influence value greater than the probability threshold β. These bottom events are the key failure factors. Based on actual needs, the top n' events are selected as key failure factors to focus on, forming a list of key failure factors N' = {N'1, N'2, ..., N'}. n' This provides a basis for formulating subsequent operation and maintenance strategies; S44, combined with the dynamically updated bottom event probability data in S36, the comprehensive influence value of the bottom event is recalculated regularly, and the list of key failure factors is updated.

Images