Health management platform for new energy power generation equipment based on large models

Abstract

The present invention discloses a health management platform for new energy power generation equipment based on a large model, and relates to the technical field of health management of power generation equipment. The present invention comprises a multivariate data acquisition and regional identification module, a data preprocessing and regional adaptation unit, a regional perception large model analysis module, an intelligent operation and maintenance decision-making and scheduling module, and a spare parts inventory optimization sharing module; the multivariate data acquisition and regional identification module deploys multiple types of sensors at key equipment locations of photovoltaic power generation stations in different regions, collects equipment operating parameters and environmental status data, and assigns regional unique codes and feature labels to ensure that data is traceable and associated with regional characteristics; the present invention significantly improves the health management efficiency and accuracy of new energy power generation equipment; it not only realizes real-time monitoring of equipment status and fault warning, but also reduces maintenance costs and improves system operation efficiency by optimizing operation and maintenance resource allocation and intelligent spare parts inventory management.

Description

Technical Field

[0001] The present invention relates to the technical field of health management of power generation equipment, and in particular to a health management platform for new energy power generation equipment based on a large model. Background Art

[0002] With the rapid development of new energy power generation technology, especially the large-scale application in the field of photovoltaic power generation, equipment health management has become the key to ensuring power generation efficiency and system reliability. However, traditional equipment management methods often rely on manual inspections and regular maintenance, which is not only inefficient but also difficult to cope with complex and changing environmental conditions and equipment status. In addition, since new energy power generation equipment is usually distributed in a wide geographical area, the environment varies significantly, resulting in significant differences in equipment failure modes and maintenance requirements. Therefore, there is an urgent need for a health management platform that can adapt to the characteristics of different regions, monitor equipment status in real time, and provide accurate maintenance decisions.

[0003] Although there are some solutions on the market that try to improve the intelligence level of equipment management through sensors and data analysis, these systems often have the following limitations: first, they lack comprehensive consideration of regional environmental characteristics, resulting in limited accuracy and applicability of analysis results; second, data processing and analysis methods are often single and difficult to adapt to changing equipment status and environmental changes; finally, the existing system lacks intelligence and automation in operation and maintenance decision-making and spare parts management, and cannot achieve rapid response and optimize resource allocation. These problems limit the efficiency and effectiveness of health management of new energy power generation equipment, and a more advanced and comprehensive technical solution is urgently needed to solve them.

[0004] In order to solve the above defects, a technical solution is now provided. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of health management of new energy power generation equipment in different regional environments, and to propose a new energy power generation equipment health management platform based on a large model.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The health management platform for new energy power generation equipment based on large models includes:

[0008] The multi-dimensional data collection and regional identification module deploys multiple types of sensors at the equipment locations of photovoltaic power stations in different regions to collect equipment operating parameters and environmental status data, and assigns unique regional codes and feature labels;

[0009] The data preprocessing and regional adaptation unit cleans the collected data by region, processes outliers and missing values, standardizes the format and dimensions, adapts different rules to the regional environment, and organizes the processed data into a structured data set for the large model to call;

[0010] The regional perception large model analysis module builds a large language model, integrates domain knowledge and trains for regional working conditions. It mines associations and analyzes fault logic by inputting data with regional identification, and outputs multi-dimensional analysis results including equipment health scores and fault risk levels.

[0011] The intelligent operation and maintenance decision-making and scheduling module generates differentiated operation and maintenance strategies based on the analysis results of the big model, combines the operation and maintenance resource library and the geographic information system, plans the optimal maintenance path, sends operation and maintenance instructions through multiple channels, tracks progress and updates data to optimize decisions;

[0012] The spare parts inventory optimization sharing module collects and analyzes regional fault repair data to determine the spare parts demand list, monitors inventory based on the Internet of Things, predicts demand trends and formulates procurement plans, searches for matches and coordinates distribution when there is a shortage, and regularly evaluates the effectiveness of strategies and continuously optimizes them.

[0013] Furthermore, the multivariate data collection and region identification module performs the following steps:

[0014] Deploy multiple types of sensors at the solar panels, inverters, junction boxes, and brackets of photovoltaic power stations in each region, including irradiance sensors, temperature sensors, humidity sensors, wind speed and direction sensors, micro vibration sensors, current and voltage sensors, and image acquisition cameras to capture equipment operating parameters and environmental conditions in real time;

[0015] Assign unique digital geographic codes and environmental feature labels to each regional station;

[0016] The collected data are packaged with regional identification and aggregated to the data preprocessing unit via wireless transmission or wired communication links to ensure that the data source can be traced and associated with regional characteristics.

[0017] Furthermore, the specific operation steps of the data preprocessing and area adaptation unit are as follows:

[0018] The received multi-source heterogeneous data is cleaned by regional classification, outliers are removed based on statistical analysis and threshold setting, missing data fragments are supplemented through interpolation algorithms, and data formats and dimensions are standardized according to device types; the interpolation algorithm uses the Lagrange interpolation method, using the formula: ,in is the interpolation point, is the function value of the known data point, is the value of the independent variable at a known data point, is the number of known data points; Indicates that at the interpolation point The interpolation results at are used to estimate the values ​​of missing data; is the function value corresponding to the known data point; is the corresponding independent variable, is the function value Independent variables that directly correspond to known data points; missing data are estimated based on the information of known data points through Lagrange interpolation polynomial calculation;

[0019] Different data processing rules are adapted to the environmental characteristics of different regions; including: in desert areas, strengthening image recognition algorithms to quantitatively evaluate the dust coverage on the surface of solar panels; in coastal salt fog areas, focusing on electrical parameter correction to offset the measurement deviation caused by humidity and salt fog corrosion; in mountainous areas, considering the conversion and adjustment of equipment performance by temperature and air pressure, the processed data is organized into structured data sets by time series and equipment level for large model calls.

[0020] Furthermore, the specific operation steps of the regional perception large model analysis module are as follows:

[0021] Build a large language model based on the Transformer architecture, pre-train and integrate global photovoltaic power generation knowledge, and carry out adjustment training for different regional working conditions;

[0022] Input a data set with regional identification, the model deeply mines the relationship between regional illumination change patterns and equipment operation, and analyzes the potential fault logic of equipment; train the fault prediction model based on historical data and regional feature data, and input real-time monitoring data using the formula: Calculate the probability of equipment failure and realize early warning of potential equipment failures. Indicates that given a feature vector The probability of equipment failure under the condition of k is the failure state, is the intercept term, Features The weight coefficient of is the equipment operation feature data, and m is the number of features;

[0023] With the help of multimodal data fusion, the fault type and fault location are determined, and the output includes analysis results such as regional adaptation health score, fault risk level, and remaining reliable operation time estimation. The output regional adaptation health score is comprehensively scored using the following formula: , where S is the regional adaptation health score, is the weight of the jth health indicator, is the normalized score of the jth health indicator, and p is the number of health indicators. The scoring standard is formulated based on factors such as regional power generation potential and equipment tolerance. The equipment health score is calculated through this formula to achieve quantitative evaluation and comparison of the health status of equipment in different regions.

[0024] Furthermore, the specific operation steps of the intelligent operation and maintenance decision-making and scheduling module are as follows:

[0025] Based on the analysis results of the large model, the differentiated operation and maintenance strategies are generated by combining the operation and maintenance resource library of each regional station with the path planning of the geographic information system. The path planning is based on the regional station geographic information and traffic condition data provided by the geographic information system, and the Dijkstra algorithm is used to calculate the optimal maintenance path. The formula is: ,in is the shortest path length from the starting point to node v, u is the node with the shortest path determined, V is the set of all nodes, is the path weight from node u to node v; this formula is used to determine the shortest or fastest path from the location of the operation and maintenance personnel to the faulty equipment; it is used to shorten the time it takes for maintenance personnel to arrive at the site, improve the fault response speed, reduce equipment downtime, and improve the overall operating efficiency of the photovoltaic power generation system;

[0026] In the case of sudden sandstorms in desert areas, where a large number of solar panels are blocked by dust, we dispatched personnel from the surrounding areas with cleaning equipment to clean the power generation arrays first; in coastal areas, we dispatched moisture-proof spare parts and electrical maintenance personnel to deal with inverter failures after typhoons;

[0027] Operation and maintenance instructions are sent to the corresponding regional operation and maintenance team through multiple channels such as SMS, APP push, and in-site alarms. The work order system tracks the progress, and data is fed back after maintenance to update the equipment health file and large model training samples, continuously optimizing the accuracy of operation and maintenance decisions.

[0028] Furthermore, the specific operation steps of the spare parts inventory optimization sharing module are as follows:

[0029] Collect historical fault data and maintenance records of each region, analyze the relationship between different fault types and regional environmental characteristics, and determine the spare parts demand list for each region; establish a spare parts inventory management unit based on the Internet of Things to monitor the inventory quantity, in and out time, and usage status information of spare parts in each region in real time;

[0030] Use predictive inventory management algorithms, combined with equipment operation data, environmental data, and historical maintenance data, to predict the demand trends of different spare parts in various regions and make procurement plans in advance;

[0031] When a spare part shortage occurs in a certain area, the intelligent search and matching algorithm is used to find spare parts in other areas and coordinate logistics distribution to achieve rapid sharing and replenishment. When the spare parts inventory management unit in a certain area detects that the inventory quantity of a certain spare part is lower than the preset shortage threshold, a shortage warning is immediately triggered.

[0032] The intelligent search and matching algorithm conducts a comprehensive evaluation based on spare parts type, quantity, distance between the area and the shortage area, and logistics and distribution cost factors to select the optimal supply area and supply plan. After determining the supply area, the platform connects with the logistics and distribution system to coordinate logistics and distribution matters, including arranging transportation vehicles, determining transportation routes, and estimating transportation time. The logistics and distribution cost is calculated through the formula Get the total logistics cost, m is the number of available delivery areas, is the available quantity in the ith region;

[0033] During the logistics and distribution process, the location of transport vehicles and the status of goods are tracked in real time through IoT technology to ensure that spare parts are delivered safely and promptly to the shortage areas;

[0034] At the same time, we regularly evaluate the effectiveness of spare parts inventory strategies in each region, adjust inventory structure and quantity according to actual conditions, and ensure that inventory costs are minimized while meeting equipment maintenance needs.

[0035] Furthermore, the specific operation steps of the spare parts inventory optimization sharing module to regularly evaluate the effectiveness of the spare parts inventory strategy of each region are as follows:

[0036] Set inventory strategy evaluation indicators, including:

[0037] Inventory Turnover Ratio: By Formula Calculate the inventory turnover rate TR, where The cost of selling spare parts in a certain period, i.e. the cost of using spare parts for maintenance. is the average inventory level;

[0038] Out of stock rate: by formula Calculate the out-of-stock rate SR, where is the number of out-of-stock times, is the total number of demands;

[0039] Inventory cost: includes the sum of holding cost and stock-out cost and is recorded as ;

[0040] The obtained inventory turnover rate TR, out-of-stock rate SR and inventory cost Compare with the corresponding standard values ​​and analyze the corresponding scores, which are the inventory turnover rate scores , out-of-stock rate score and inventory cost score After normalization, the following formula is entered: To obtain the comprehensive judgment value ZPZ, the obtained comprehensive judgment value ZPZ is compared with the preset comprehensive judgment threshold. When the comprehensive judgment value ZPZ is greater than the preset comprehensive judgment threshold, it is judged that the effectiveness of the regional spare parts inventory strategy has reached the standard. Otherwise, it is judged that the standard has not been reached, and the inventory strategy evaluation index is retrospectively analyzed;

[0041] When the inventory turnover rate is lower than the standard or the inventory cost is higher than the standard, it is considered as inventory backlog and you need to consider reducing the purchase volume or adjusting the safety stock threshold. When the out-of-stock rate is higher than the standard, you need to increase the safety stock or optimize the procurement plan.

[0042] By simulating different inventory strategies and using historical data for simulation tests, the comprehensive judgment values ​​ZPZ under different strategies are compared, and the one with the highest comprehensive judgment value ZPZ is selected as the optimal inventory strategy for adjustment and implementation; at the same time, the adjusted strategy is fed back and relevant parameters are updated to achieve continuous optimization of the inventory strategy.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] (1) The present invention realizes comprehensive real-time monitoring of the operating parameters and environmental status of new energy power generation equipment through multi-dimensional data collection and regional identification modules. By deploying multiple types of sensors at photovoltaic power generation sites in different regions, key parameters such as irradiance, temperature, and humidity are collected. The unique regional code and feature tags are combined to ensure the traceability of data and the association with regional characteristics. This refined data collection method enables the equipment health management platform to more accurately capture equipment status changes and respond to potential failure risks in a timely manner, thereby significantly improving the accuracy and real-time performance of equipment health management.

[0045] (2) In the present invention, the data preprocessing and regional adaptation unit and the regional perception large model analysis module deeply mine and analyze data with regional identification, and output multi-dimensional analysis results including equipment health score, fault risk level, etc.; the intelligent operation and maintenance decision-making and scheduling module generates differentiated operation and maintenance strategies based on these analysis results, combines the operation and maintenance resource library with the geographic information system, and plans the optimal maintenance path. It also sends operation and maintenance instructions through multiple channels and tracks the progress in real time. This not only improves the fault response speed and equipment operation efficiency, but also realizes the optimal allocation of operation and maintenance resources, reduces equipment downtime, and reduces maintenance costs;

[0046] (3) The present invention, the spare parts inventory optimization sharing module, collects and analyzes regional fault maintenance data, determines the spare parts demand list, monitors the inventory based on the Internet of Things, predicts demand trends and formulates procurement plans. When spare parts are in short supply, the intelligent search and matching algorithm can quickly find and coordinate spare parts distribution in other regions to achieve rapid sharing and replenishment. Regularly evaluate the effectiveness of each region's spare parts inventory strategy, and adjust the inventory structure and quantity according to actual conditions to ensure that inventory costs are minimized while meeting equipment maintenance needs; this intelligent spare parts inventory management not only improves the response speed and flexibility of spare parts supply, but also effectively reduces the risk of inventory backlogs and achieves cost control. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0048] Figure 1 This is the overall system block diagram of the present invention. DETAILED DESCRIPTION

[0049] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] It should be understood that the terms “include” and “comprising” used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0051] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations.

[0052] like Figure 1 As shown in the figure, the health management platform of new energy power generation equipment based on the big model includes a multivariate data collection and regional identification module, a data preprocessing and regional adaptation unit, a regional perception big model analysis module, an intelligent operation and maintenance decision-making and scheduling module, and a spare parts inventory optimization sharing module;

[0053] The multi-dimensional data collection and regional identification module deploys multiple types of sensors at key equipment locations in photovoltaic power stations in different regions to collect equipment operating parameters and environmental status data, and assigns unique regional codes and feature labels to ensure that data is traceable and associated with regional characteristics;

[0054] Deploy multiple types of sensors at key equipment parts of solar panels, inverters, junction boxes, and brackets of photovoltaic power stations in various regions (including desert power stations, coastal power stations, high mountain power stations, and urban distributed power stations); including high-precision irradiance sensors, temperature sensors, humidity sensors, wind speed and direction sensors, micro vibration sensors, current and voltage sensors, and image acquisition cameras to capture equipment operating parameters and environmental conditions in real time; assign unique digital geographic codes and environmental feature labels to each regional station (such as "Northwest Desert Dry and High Temperature Area - Strong Wind and Sand - Strong Light and Small Duration Fluctuation" and "Southeast Coast Humid and Salty Fog Area - Frequent Typhoons - Light Affected by Monsoon"); the definition process of environmental feature labels is as follows:

[0055] Light conditions include: Annual sunshine duration statistics: Through the irradiance sensors and sunshine duration recorders deployed at the site (with automatic recognition of light threshold trigger timing function), the sunshine duration is recorded cumulatively in the natural year, accurate to the hour, and the effective sunshine duration throughout the year is counted ,when , marked as “long daylight zone”; , marked as “Zhongrizhao District”; , marked as "short-day area"; average irradiance calculation: multiple irradiance sensors are evenly distributed at different positions and heights of the solar panels on the site (with a spacing of no more than 50 meters to ensure regional representativeness), and data is collected once a minute. The system calculates the average irradiance on a quarterly basis ; The calculation formula is: ,in is the irradiance value collected each time, It is the total number of collections within a quarter. The levels are defined as follows: , classified as “high irradiance area”; , defined as the “medium irradiance zone”; , identified as a "low irradiance area"; Light stability assessment: Using the time series analysis algorithm, the irradiance sensor data is segmented and analyzed at intervals of 15 minutes, and the standard deviation of irradiance in each period is calculated. , reflecting the degree of light fluctuation. The larger the standard deviation, the worse the stability. Specific classification: , marked as “photostable zone”; , defined as the “relatively stable light zone”; , classified as the “light unstable zone”;

[0056] Temperature and humidity environment includes: Temperature parameter setting: Calculate the average value of temperature data collected throughout the year ,when , marked as "high temperature zone"; , define the “medium temperature zone”; , marked as "low temperature zone"; extreme temperature range: record the highest and lowest temperatures throughout the year, evaluate the risk of cold and heat shock faced by the equipment, when the temperature difference between the highest and lowest temperatures is greater than 50°C, mark "temperature fluctuation zone", otherwise it is "relatively stable temperature zone"; humidity: calculate the annual average relative humidity, if the annual average relative humidity is greater than 80%, it is classified as "high humidity zone"; when the annual average relative humidity is between 50%-80%, it is defined as "medium humidity zone"; when the annual average relative humidity is less than 50%, it is identified as "low humidity zone"; at the same time, record the highest relative humidity of the year, when it is greater than 95%, mark "high humidity extreme risk zone" to remind the equipment of moisture-proof needs;

[0057] The frequency of meteorological disasters includes: typhoon ( ): According to local meteorological data, = "high incidence", 1-2 = "frequent occurrence", and none = "no typhoon"; heavy rain and floods ( , ): Statistics of the number and days of annual heavy rain (daily precipitation > 50 mm), and days is "frequent", 2-4 times and 5-9 days is "regular", and the rest is "rare"; sandstorm: annual sandstorm For "frequent", hail is "often present", and vice versa is "rarely present";

[0058] Geographical features include: terrain slope ( ): LiDAR mapping, It is a steep slope. For "gentle slope", belongs to "flat"; altitude ( ): Determine the altitude based on GPS. Meters is "high altitude", 1000-1999 meters is "medium altitude", Meters for "low altitude"; soil corrosiveness ( ): Quarterly soil sampling to measure the corrosion index, 0 is "strong corrosion", 2-4 is "medium corrosion", It is "weakly corrosive".

[0059] The collected data are packaged with regional identification and aggregated to the data preprocessing unit via wireless transmission (4G / 5G networks are prioritized to ensure stable and high-speed data return) or wired (internal local area network of the station) communication link to ensure that the data source can be traced and associated with regional characteristics.

[0060] The data preprocessing and regional adaptation unit cleans the collected data by region, processes outliers and missing values, standardizes the format and dimensions, adapts special rules to the regional environment, and organizes the processed data into a structured data set for the large model to call;

[0061] The massive multi-source heterogeneous data received is cleaned by region classification, outliers (such as outrageous data points caused by sensor failure) are eliminated based on statistical analysis and threshold setting, missing data fragments are supplemented through interpolation algorithms, and data formats and dimensions are standardized according to device types; the interpolation algorithm uses the Lagrange interpolation method, using the formula: ,in is the interpolation point, is the function value of the known data point, is the value of the independent variable at a known data point, is the number of known data points; Indicates that at the interpolation point The interpolation results at are used to estimate the values ​​of missing data; is the function value corresponding to the known data point (such as the value collected by sensors such as temperature and irradiance); is the corresponding independent variable (such as time, location, etc.), is the function value Independent variables that directly correspond to known data points; through Lagrange interpolation polynomial calculation, reasonable estimates of missing data are made based on the information of known data points;

[0062] Special data processing rules are adapted according to the environmental characteristics of different regions. In windy desert areas, image recognition algorithms are strengthened to quantitatively evaluate the dust coverage on the surface of solar panels. In coastal salt fog areas, focus is placed on electrical parameter correction to offset measurement deviations caused by humidity and salt fog corrosion. In high-altitude mountainous areas, the effects of temperature and air pressure on equipment performance are adjusted, and the processed data are organized into structured data sets by time series and equipment level for large model calls.

[0063] The regional perception big model analysis module builds an improved big language model, integrates domain knowledge and fine-tunes the training according to regional working conditions. It mines associations and analyzes fault logic by inputting data with regional identification, and outputs multi-dimensional analysis results including equipment health scores, fault risk levels, etc.

[0064] Build a large language model based on the improved Transformer architecture, pre-train and integrate global photovoltaic power generation knowledge (equipment principles, fault cases, meteorological photovoltaic correlation theory), and carry out special fine-tuning training for different regional working conditions; input a data set with regional identification, and the model deeply explores the relationship between regional light change rules (sudden rise and fall of irradiance in desert areas, seasonal light shielding in coastal areas) and equipment operation, and analyzes the potential fault logic of equipment; train a fault prediction model based on historical data and regional feature data, and input real-time monitoring data, using the formula: Calculate the probability of equipment failure and realize early warning of potential equipment failures. Indicates that given a feature vector The probability of equipment failure under the condition of k is the fault state (0 means normal, 1 means fault), is the intercept term, Features The weight coefficient of is the equipment operation characteristic data, such as electrical parameters, temperature change rate, irradiance fluctuation, etc., and m is the number of characteristics;

[0065] With the help of multimodal data fusion (electrical parameters, thermal imaging, appearance images), the fault type (hidden cracks in solar panel cells, overheating and damage of inverter power tubes, poor contact of combiner boxes) and fault location (which solar panel in which array of a certain regional station, which power module of the inverter) are judged, and the output includes regional adaptation health score (scoring standards are formulated according to regional power generation potential and equipment tolerance), fault risk level (high, medium, and low correspond to emergency repairs, regular maintenance, and continuous observation), and remaining reliable operation time estimation. The output regional adaptation health score is comprehensively scored using the following formula: , where S is the regional adaptation health score, is the weight of the jth health indicator, is the normalized score of the jth health indicator, and p is the number of health indicators. The scoring standard is formulated based on factors such as regional power generation potential and equipment tolerance. The equipment health score is calculated through this formula to achieve quantitative evaluation and comparison of the health status of equipment in different regions.

[0066] The intelligent operation and maintenance decision-making and scheduling module combines the operation and maintenance resource library and geographic information system based on the analysis results of the big model to generate differentiated operation and maintenance strategies, plan the optimal maintenance path, send operation and maintenance instructions through multiple channels, track progress and update data to optimize decision-making;

[0067] Based on the analysis results of the big model, combined with the regional station operation and maintenance resource library (personnel skill distribution, spare parts inventory location and quantity, maintenance tool equipment) and geographic information system (GIS) path planning, intelligently generate differentiated operation and maintenance strategies; the path planning is based on the regional station geographic information and traffic status data provided by the geographic information system (GIS), and the Dijkstra algorithm is used to calculate the optimal maintenance path, through the formula: ,in is the shortest path length from the starting point to node v, u is the node with the shortest path determined, V is the set of all nodes, is the path weight from node u to node v, such as distance and traffic congestion level; this formula is used to determine the shortest or fastest path from the location of the operation and maintenance personnel to the faulty equipment; it helps to shorten the time for maintenance personnel to arrive at the site, improve the fault response speed, reduce equipment downtime, and improve the overall operating efficiency of the photovoltaic power generation system; especially when dealing with faults in photovoltaic power generation sites distributed over a large area, fast and efficient path planning is particularly important;

[0068] In the case of sudden sandstorms in desert areas, which caused dust to cover a large number of solar panels, we deployed personnel with sand clearing experience in the surrounding area with large cleaning equipment to prioritize cleaning of key power generation arrays. In the case of inverter failures in coastal areas after typhoons, we urgently dispatched moisture-proof spare parts and professional electrical maintenance personnel. We sent detailed operation and maintenance instructions (fault diagnosis details, maintenance steps, safety precautions) to the operation and maintenance teams in the corresponding areas through multiple channels such as text messages, APP push notifications, and in-station alarms. The work order system tracks the progress, and the data is sent back after maintenance to update the equipment health file and large model training samples, continuously optimizing the accuracy of operation and maintenance decisions.

[0069] The spare parts inventory optimization sharing module collects and analyzes regional fault repair data to determine the demand list of special spare parts, monitors inventory based on the Internet of Things, predicts demand trends and formulates procurement plans, intelligently searches for matching and coordinates distribution when there is a shortage, and regularly evaluates the effectiveness of strategies and continuously optimizes them;

[0070] Collect historical fault data and maintenance records from each region, analyze the relationship between different fault types and regional environmental characteristics, and determine the spare parts demand list for each region; establish a spare parts inventory management unit based on the Internet of Things (IoT) to monitor the inventory quantity, in and out time, and usage status information of spare parts in each region in real time; use predictive inventory management algorithms, combined with equipment operation data, environmental data, and historical maintenance data, to predict the demand trend of different spare parts in each region and formulate procurement plans in advance;

[0071] When a spare part shortage occurs in a certain area, the intelligent search and matching algorithm is used to find available spare parts in other areas, and the logistics distribution is coordinated to achieve rapid sharing and replenishment. When the spare parts inventory management unit in a certain area detects that the inventory quantity of a certain spare part is lower than the preset shortage threshold, a shortage warning is immediately triggered. The intelligent search and matching algorithm conducts a comprehensive evaluation based on the spare part type, quantity, distance between the area where the spare part is located and the shortage area (which can be calculated by the geographic information system), and logistics distribution cost factors to select the optimal supply area and supply plan. After determining the supply area, the platform connects with the logistics distribution system to coordinate logistics distribution matters, including arranging transportation vehicles, determining transportation routes, and estimating transportation time. The logistics distribution cost is calculated by the formula Get the total logistics cost, where m is the number of available delivery areas, The quantity of goods available for the i-th region; in the process of logistics and distribution, the location of transport vehicles and the status of goods are tracked in real time through the Internet of Things technology to ensure that spare parts are delivered safely and timely to the shortage areas;

[0072] At the same time, the effectiveness of spare parts inventory strategies in each region is regularly evaluated, and the inventory structure and quantity are adjusted according to actual conditions to ensure that inventory costs are minimized while meeting equipment maintenance needs. Specific inventory strategy evaluation indicators include:

[0073] Inventory Turnover Ratio: By Formula Calculate the inventory turnover rate TR, where The cost of selling spare parts in a certain period, i.e. the cost of using spare parts for maintenance. is the average inventory level; out-of-stock rate: through the formula Calculate the out-of-stock rate SR, where is the number of out-of-stock times, is the total number of demands; inventory cost: includes the sum of holding cost and out-of-stock cost, and is recorded as ;

[0074] The obtained inventory turnover rate TR, out-of-stock rate SR and inventory cost Compare with the corresponding standard values ​​and analyze the corresponding scores, which are the inventory turnover rate scores , out-of-stock rate score and inventory cost score After normalization, the following formula is entered: To obtain the comprehensive judgment value ZPZ, the obtained comprehensive judgment value ZPZ is compared with the preset comprehensive judgment threshold. When the comprehensive judgment value ZPZ is greater than the preset comprehensive judgment threshold, it is judged that the effectiveness of the regional spare parts inventory strategy has reached the standard. Otherwise, it is judged that the standard has not been reached, and the inventory strategy evaluation index is retrospectively analyzed;

[0075] When the inventory turnover rate is lower than the standard or the inventory cost is higher than the standard, it means that there is an inventory backlog, and it is necessary to consider reducing the purchase volume or adjusting the safety stock threshold; if the out-of-stock rate is higher than the standard, it is necessary to increase the safety stock or optimize the purchase plan; by simulating different inventory strategies (such as adjusting the purchase batch, safety stock level, replenishment trigger point, etc.), using historical data for simulation testing, comparing the comprehensive judgment values ​​ZPZ under different strategies, and selecting the one with the highest comprehensive judgment value ZPZ as the optimal inventory strategy for adjustment and implementation. At the same time, the adjusted strategy is fed back, and the relevant parameters are updated to achieve continuous optimization of the inventory strategy.

[0076] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A health management platform for new energy power generation equipment based on a large model, characterized by: include: The multi-dimensional data collection and regional identification module deploys multiple types of sensors at the equipment locations of photovoltaic power stations in different regions to collect equipment operating parameters and environmental status data, and assigns unique digital geographic codes and environmental feature labels to the regions; The data preprocessing and regional adaptation unit cleans the collected data by region, processes outliers and missing values, standardizes the format and dimensions, adapts different rules to the regional environment, and organizes the processed data into a structured data set for the large model to call; The regional awareness large model analysis module builds a large language model, integrates domain knowledge and trains for regional working conditions. It mines associations and analyzes fault logic by inputting data with regional identification, and outputs multi-dimensional analysis results including equipment health scores and fault risk levels. The specific operation steps are as follows: Build a large language model based on the Transformer architecture, pre-train and integrate global photovoltaic power generation knowledge, and carry out adjustment training for different regional working conditions; Input a data set with regional identification, the model deeply mines the relationship between regional illumination change patterns and equipment operation, and analyzes the potential fault logic of equipment; train the fault prediction model based on historical data and regional feature data, and input real-time monitoring data using the formula: Calculate the probability of equipment failure and realize early warning of potential equipment failures. Indicates that given a feature vector The probability of equipment failure under the condition of k is the failure state, is the intercept term, Features The weight coefficient of is the equipment operation feature data, and m is the number of features; With the help of multimodal data fusion, the fault type and fault location are determined, and the output includes regional adaptation health score, fault risk level, and estimated analysis results of the remaining reliable operating time. The output regional adaptation health score is comprehensively scored using the formula: , where S is the regional adaptation health score, is the weight of the jth health indicator, is the normalized score of the jth health indicator, and p is the number of health indicators. The scoring standard is formulated based on the regional power generation potential and equipment tolerance factors, and the equipment health score is calculated through the formula to achieve quantitative evaluation and comparison of the health status of equipment in different regions. The intelligent operation and maintenance decision-making and scheduling module generates differentiated operation and maintenance strategies based on the analysis results of the big model, combines the operation and maintenance resource library and the geographic information system, plans the optimal maintenance path, sends operation and maintenance instructions through multiple channels, tracks progress and updates data to optimize decisions; The spare parts inventory optimization sharing module collects and analyzes regional fault repair data to determine the spare parts demand list, monitors inventory based on the Internet of Things, predicts demand trends and formulates procurement plans, searches for matches and coordinates distribution when there is a shortage, and regularly evaluates the effectiveness of strategies and continuously optimizes them.

2. The large model-based new energy power generation equipment health management platform according to claim 1 is characterized in that: The multivariate data collection and region identification module performs the following steps: Deploy multiple types of sensors at the solar panels, inverters, junction boxes, and brackets of photovoltaic power stations in each region, including irradiance sensors, temperature sensors, humidity sensors, wind speed and direction sensors, micro vibration sensors, current and voltage sensors, and image acquisition cameras to capture equipment operating parameters and environmental conditions in real time; Assign unique digital geographic codes and environmental feature labels to each regional station; The collected data are packaged with regional identification and aggregated to the data preprocessing unit via wireless transmission or wired communication links to ensure that the data source can be traced and associated with regional characteristics.

3. The large model-based new energy power generation equipment health management platform according to claim 1 is characterized in that: The specific operation steps of the data preprocessing and area adaptation unit are as follows: The received multi-source heterogeneous data is cleaned by regional classification, outliers are removed based on statistical analysis and threshold setting, missing data fragments are supplemented through interpolation algorithms, and data formats and dimensions are standardized according to device types; the interpolation algorithm uses the Lagrange interpolation method, using the formula: ,in is the interpolation point, is the function value of the known data point, is the value of the independent variable at a known data point, is the number of known data points; Indicates that at the interpolation point The interpolation results at are used to estimate the values ​​of missing data; is the function value corresponding to the known data point; is the corresponding independent variable, is the function value Independent variables that directly correspond to known data points; missing data are estimated based on the information of known data points through Lagrange interpolation polynomial calculation; Different data processing rules are adapted to the environmental characteristics of different regions; including: in desert areas, strengthening image recognition algorithms to quantitatively evaluate the dust coverage on the surface of solar panels; in coastal salt fog areas, focusing on electrical parameter correction to offset the measurement deviation caused by humidity and salt fog corrosion; in mountainous areas, considering the conversion and adjustment of equipment performance by temperature and air pressure, the processed data is organized into structured data sets by time series and equipment level for large model calls.

4. The large model-based new energy power generation equipment health management platform according to claim 1 is characterized in that: The specific operation steps of the intelligent operation and maintenance decision-making and scheduling module are as follows: Based on the analysis results of the large model, the differentiated operation and maintenance strategies are generated by combining the operation and maintenance resource library of each regional station with the path planning of the geographic information system. The path planning is based on the regional station geographic information and traffic condition data provided by the geographic information system, and the Dijkstra algorithm is used to calculate the optimal maintenance path. The formula is: ,in is the shortest path length from the starting point to node v, u is the node with the shortest path determined, V is the set of all nodes, is the path weight from node u to node v; the formula is used to determine the shortest or fastest path from the location of the operation and maintenance personnel to the faulty equipment; it is used to shorten the time for maintenance personnel to arrive at the site, improve the fault response speed, reduce equipment downtime, and improve the overall operation efficiency of the photovoltaic power generation system; When a large number of solar panels are blocked by dust after a sudden sandstorm in a desert area, surrounding personnel are deployed with cleaning equipment to clean the power generation array first; Inverter failures in coastal areas after typhoons led to the dispatch of moisture-proof spare parts and electrical maintenance personnel; Operation and maintenance instructions are sent to the corresponding regional operation and maintenance team through multiple channels such as SMS, APP push, and in-site alarms. The work order system tracks the progress, and data is fed back after maintenance to update the equipment health file and large model training samples, continuously optimizing the accuracy of operation and maintenance decisions.

5. The large model-based new energy power generation equipment health management platform according to claim 1 is characterized in that: The specific operation steps of the spare parts inventory optimization sharing module are as follows: Collect historical fault data and maintenance records of each region, analyze the relationship between different fault types and regional environmental characteristics, and determine the spare parts demand list for each region; establish a spare parts inventory management unit based on the Internet of Things to monitor the inventory quantity, in and out time, and usage status information of spare parts in each region in real time; Use predictive inventory management algorithms, combined with equipment operation data, environmental data, and historical maintenance data, to predict the demand trends of different spare parts in various regions and make procurement plans in advance; When a spare part shortage occurs in a certain area, the intelligent search and matching algorithm is used to find spare parts in other areas and coordinate logistics distribution to achieve rapid sharing and replenishment. When the spare parts inventory management unit in a certain area detects that the inventory quantity of a certain spare part is lower than the preset shortage threshold, a shortage warning is immediately triggered. The intelligent search and matching algorithm conducts a comprehensive evaluation based on spare parts type, quantity, distance between the area and the shortage area, and logistics and distribution cost factors to select the optimal supply area and supply plan. After determining the supply area, the platform connects with the logistics and distribution system to coordinate logistics and distribution matters, including arranging transportation vehicles, determining transportation routes, and estimating transportation time. The logistics and distribution cost is calculated through the formula Get the total logistics cost, m is the number of available delivery areas, is the available quantity in the ith region; During the logistics and distribution process, the location of transport vehicles and the status of goods are tracked in real time through IoT technology to ensure that spare parts are delivered safely and promptly to the shortage areas; At the same time, we regularly evaluate the effectiveness of spare parts inventory strategies in each region, and adjust inventory structure and quantity according to actual conditions to ensure that inventory costs are minimized while meeting equipment maintenance needs.

6. The large model-based new energy power generation equipment health management platform according to claim 5 is characterized in that: The specific operation steps of the spare parts inventory optimization sharing module to regularly evaluate the effectiveness of the spare parts inventory strategy of each region are as follows: Set inventory strategy evaluation indicators, including: Inventory Turnover Ratio: By Formula Calculate the inventory turnover rate TR, where The cost of selling spare parts during the set period, i.e. the cost of spare parts used for maintenance, is the average inventory level; Out of stock rate: by formula Calculate the out-of-stock rate SR, where is the number of out-of-stock times, is the total number of demands; Inventory cost: includes the sum of holding cost and stock-out cost and is recorded as ; The obtained inventory turnover rate TR, out-of-stock rate SR and inventory cost Compare with the corresponding standard values ​​and analyze the corresponding scores, which are the inventory turnover rate scores , out-of-stock rate score and inventory cost score After normalization, the following formula is entered: To obtain the comprehensive judgment value ZPZ, the obtained comprehensive judgment value ZPZ is compared with the preset comprehensive judgment threshold. When the comprehensive judgment value ZPZ is greater than the preset comprehensive judgment threshold, it is judged that the effectiveness of the regional spare parts inventory strategy has reached the standard. Otherwise, it is judged that the standard has not been reached, and the inventory strategy evaluation index is retrospectively analyzed; When the inventory turnover rate is lower than the standard or the inventory cost is higher than the standard, it is considered as inventory backlog and you need to consider reducing the purchase volume or adjusting the safety stock threshold. When the out-of-stock rate is higher than the standard, you need to increase the safety stock or optimize the procurement plan. By simulating different inventory strategies and using historical data for simulation tests, the comprehensive judgment values ​​ZPZ under different strategies are compared, and the one with the highest comprehensive judgment value ZPZ is selected as the optimal inventory strategy for adjustment and implementation; at the same time, the adjusted strategy is fed back and relevant parameters are updated to achieve continuous optimization of the inventory strategy.

Images