New energy power station master data management method and system based on artificial intelligence

Abstract

The invention discloses a new energy power station main data management method and system based on artificial intelligence, and relates to the technical field of new energy power station management, and the method comprises the steps that a data collection module collects the operation data of new energy power station equipment; the fault identification module analyzes an equipment fault mode based on a fault identification algorithm and judges the risk level of the equipment; the data management module generates a health state report according to the risk level, and performs classified management on the equipment; and the production strategy optimization module carries out production task distribution by using a reinforcement learning algorithm and optimizes the operation load of the equipment. According to the new energy power station main data management method based on artificial intelligence, through combination of the generative adversarial network and the reinforcement learning algorithm, intelligent closed-loop management of equipment fault prediction, load management and system optimization is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy power station management, and specifically to a new energy power station master data management method and system based on artificial intelligence. Background Art

[0002] With the rapid global adoption of new energy power plants, the number of new energy devices such as photovoltaic panels, wind turbines, and battery systems continues to increase. While the widespread deployment of these devices brings clean energy to the power system, it also raises complex equipment management and O&M challenges. Traditional equipment management methods rely on preset maintenance cycles and manual monitoring methods, making it difficult to promptly identify potential equipment failure risks. This leads to inefficient O&M, delayed fault warnings, and shortened equipment lifespans. Furthermore, existing equipment management systems often lack the ability to comprehensively analyze real-time environmental conditions and historical equipment data, making it impossible to dynamically adjust equipment load distribution based on actual operating conditions. This limits the overall operational efficiency of new energy power plants.

[0003] To achieve intelligent management and efficient fault prediction for renewable energy power station equipment, equipment management methods based on artificial intelligence algorithms have emerged. Generative adversarial network (GAN) algorithms and reinforcement learning algorithms, currently cutting-edge technologies in the field of artificial intelligence, can conduct in-depth analysis of equipment operating data and provide intelligent load adjustment strategies. However, existing technologies lack a new energy power station equipment management method that combines GAN and reinforcement learning to achieve closed-loop control for fault prediction and load distribution.

[0004] To address the above problems, an artificial intelligence-based master data management method for new energy power stations is proposed. It can effectively collect equipment data, identify fault risks, and dynamically optimize equipment load distribution to achieve intelligent management of new energy power station equipment. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is that traditional methods are difficult to fully utilize the equipment operating status, environmental conditions and historical data to achieve real-time prediction of faults, resulting in the equipment continuing to operate in a potential fault state, exacerbating the risk of failure.

[0007] Existing equipment management methods typically use static load distribution methods, which cannot dynamically adjust the load based on the real-time health status and risk level of the equipment, limiting the overall operating efficiency of the power station.

[0008] In existing technologies, equipment management systems usually lack real-time data collection and analysis, making it difficult to form a closed-loop system for data collection, fault prediction, load adjustment, and status feedback, resulting in insufficient equipment status management and optimization.

[0009] To solve the above technical problems, the present invention provides the following technical solutions: a new energy power station master data management method based on artificial intelligence, comprising:

[0010] The data acquisition module collects the operating data of new energy power station equipment;

[0011] The fault identification module analyzes the equipment failure mode based on the fault identification algorithm and determines the risk level of the equipment;

[0012] The data governance module generates health status reports based on risk levels and manages devices by category;

[0013] The production strategy optimization module uses reinforcement learning algorithms to allocate production tasks and optimize the operating load of equipment.

[0014] As a preferred embodiment of the artificial intelligence-based master data management method for a new energy power station according to the present invention, the collecting of operating data of the new energy power station equipment includes collecting equipment operating status data, environmental condition data, and equipment history data from photovoltaic modules, wind turbines, and battery systems in the new energy power station to generate a first data set;

[0015] Performing a multi-level analysis on the first data set using a feature extraction algorithm to assign an acquisition risk level to the device, the acquisition risk level including high acquisition risk, medium acquisition risk, and low acquisition risk;

[0016] The acquisition strategy is formulated according to the acquisition risk level, and a third data set containing multi-level acquisition content is generated and transmitted to the fault identification module.

[0017] As a preferred solution of the artificial intelligence-based master data management method for a new energy power station of the present invention, wherein: the analyzing the equipment failure mode based on the fault identification algorithm includes inputting the data of the third data set into a generative adversarial network algorithm to construct a multi-level fault prediction model;

[0018] The data in the third data set are processed level by level according to the order of acquisition risk levels to generate a fourth data set including fault type, fault probability and feature weight.

[0019] As a preferred solution of the artificial intelligence-based master data management method for new energy power plants of the present invention, the classification management of the equipment includes calculating a health score using the fault probability and feature weights in the fourth data set, and constructing a fifth data set based on the fault type and health score of each device;

[0020] Generate a master data mapping structure based on health scores and classify equipment risks into high-failure-risk equipment, medium-failure-risk equipment, and low-failure-risk equipment;

[0021] Generate task assignment list based on master data mapping structure and fault type;

[0022] The task allocation list includes a load allocation requirement interval for equipment at each risk level.

[0023] As a preferred solution of the artificial intelligence-based master data management method for a new energy power station of the present invention, wherein: the production task allocation using a reinforcement learning algorithm includes determining a load allocation value for each device within a load allocation requirement interval using the reinforcement learning algorithm;

[0024] Evaluate the health status of the equipment after the re-load distribution, update the master data mapping structure, and if the equipment is still at high risk of failure or the risk classification level has increased, execute the load secondary distribution strategy until the risk classification level decreases;

[0025] If the equipment is not judged as a high-failure-risk equipment or the risk classification level is reduced, the risk classification level is fed back to the data acquisition module to adjust the frequency of operation data collection.

[0026] As a preferred solution of the master data management method of a new energy power station based on artificial intelligence according to the present invention, wherein: the execution of the load secondary distribution strategy includes reducing the equipment load to the minimum load mode when the equipment risk level is still high failure risk;

[0027] When the risk level rises from low failure risk equipment to medium failure risk equipment, immediately adjust the load to the lower limit of the medium risk range.

[0028] Another object of the present invention is to provide an artificial intelligence-based new energy power station master data management system, which can solve the problems of inaccurate fault prediction and lack of flexibility in load distribution in existing new energy power station equipment management systems by constructing an artificial intelligence-based new energy power station master data management system.

[0029] To solve the above technical problems, the present invention provides the following technical solutions: an artificial intelligence-based new energy power station master data management system, comprising: a data acquisition module, a fault identification module, a data governance module and a production strategy optimization module; the data acquisition module collects operating data of new energy power station equipment; the fault identification module analyzes equipment failure modes based on a fault identification algorithm and determines the risk level of the equipment; the data governance module generates a health status report according to the risk level and classifies and manages the equipment; the production strategy optimization module uses a reinforcement learning algorithm to allocate production tasks and optimize the operating load of the equipment.

[0030] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned artificial intelligence-based new energy power station master data management method are implemented.

[0031] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned artificial intelligence-based new energy power station master data management method.

[0032] A computer program product includes a computer program, which, when executed by a processor, implements the steps of the artificial intelligence-based new energy power station master data management method described in any one of the first aspects.

[0033] Beneficial effects of the present invention: The artificial intelligence-based master data management method for new energy power stations provided by the present invention realizes intelligent closed-loop management of equipment fault prediction, load management and system optimization through the combination of generative adversarial networks and reinforcement learning algorithms. The GAN algorithm is based on real-time analysis of multi-dimensional data, predicts faults several hours in advance, and reduces the risk of equipment loss during the latent period; the reinforcement learning algorithm dynamically adjusts the load, accurately reduces the load under high-risk conditions, and gradually optimizes the load distribution after restoration of stability, thereby reducing the failure frequency of equipment under high load and extending the service life of the equipment. The closed-loop management structure of the system ensures real-time feedback of data collection, analysis and load adjustment, making equipment management more efficient and robust. The overall solution has significant effects in improving the accuracy of fault prediction, optimizing load distribution, extending equipment life and improving the operation and maintenance efficiency of new energy power stations. It is suitable for the equipment management needs of new energy power stations under complex operating environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 An overall flow chart of a master data management method for a new energy power station based on artificial intelligence provided by one embodiment of the present invention.

[0036] Figure 2 This is an overall structural diagram of the artificial intelligence-based new energy power station master data management system provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0037] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0038] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0039] Example 1

[0040] Reference Figure 1 , which is an embodiment of the present invention, provides a new energy power station master data management method based on artificial intelligence, including:

[0041] S1: The data acquisition module collects the operating data of the new energy power station equipment;

[0042] S2: The fault identification module analyzes the equipment failure mode based on the fault identification algorithm and determines the risk level of the equipment;

[0043] S3: The data governance module generates health status reports based on risk levels and manages devices by category.

[0044] S4: The production strategy optimization module uses reinforcement learning algorithms to allocate production tasks and optimize the operating load of equipment.

[0045] The collecting of the operating data of the new energy power station equipment includes collecting equipment operating status data, environmental condition data and equipment history data from photovoltaic components, wind turbines and batteries in the new energy power station to generate a first data set;

[0046] It should be noted that the equipment operating status data of the present invention includes the current power output, temperature and vibration conditions of the equipment, the environmental condition data includes the real-time meteorological parameters of the location of the equipment, and the equipment history data includes the historical fault records and operating performance information of the equipment.

[0047] Performing a multi-level analysis on the first data set using a feature extraction algorithm to assign an acquisition risk level to the device, the acquisition risk level including high acquisition risk, medium acquisition risk, and low acquisition risk;

[0048] It should be noted that the feature extraction algorithm steps of the present invention are as follows:

[0049] First, the parameters in each data set are standardized to ensure the consistency of the data dimensions.

[0050] The K-means clustering algorithm is then used to group the data into clusters based on feature similarity. The equipment's fault records and feature weights are used as input variables for clustering to identify the most representative equipment operating modes.

[0051] The characteristic points of the equipment are calculated by weight (different weights are assigned according to the fault frequency and key parameter fluctuation rate of historical data) to extract the key characteristic points that reflect the current status.

[0052] Based on these key features, devices are assigned to three acquisition risk levels: high risk (feature points significantly exceeding the standard value), medium risk (some feature points are close to the standard value), and low risk (all feature points are below the standard value). The generated second dataset contains each device's acquisition risk level, acquisition frequency, and corresponding feature point parameters.

[0053] The acquisition strategy is formulated according to the acquisition risk level, and a third data set containing multi-level acquisition content is generated and transmitted to the fault identification module.

[0054] It should be noted that for high-risk devices, data for all key features and some minor features is collected every 5 seconds. For medium-risk devices, data for key features and some minor features is collected every 10 seconds. For low-risk devices, data for some key features is collected every 15 seconds. The resulting third dataset contains detailed collection content and frequency, and this data is transmitted to the fault identification module for fault prediction analysis.

[0055] Analyzing the equipment failure mode based on the fault identification algorithm includes inputting the third data set data into a generative adversarial network algorithm to construct a multi-level fault prediction model;

[0056] The data in the third data set are processed level by level according to the order of acquisition risk levels to generate a fourth data set including fault type, fault probability and feature weight.

[0057] It should be noted that the fault identification module receives the third data set, and performs standardization and denoising on the equipment's operating status data, environmental condition data, and equipment history data to obtain a structured feature matrix.

[0058] In a generative adversarial network (GAN) model, a generator and a discriminator are trained to identify key fault signatures in high-risk equipment data. The generator creates virtual fault samples with different characteristics, while the discriminator compares the real data with the generated data and gradually optimizes the generator's parameters to produce samples that are closer to the actual fault manifestations.

[0059] During training, feature weights are calculated based on the influence of each fault feature on the fault prediction results. Specifically, through multiple iterations, the sensitivity of each feature to the model output error is measured and corresponding weights are assigned. Features with higher weights are considered to have a greater impact on the occurrence of faults, forming the final feature weight matrix.

[0060] Based on the trained GAN model, the team first identified key features of the high-risk equipment data. Parameters with significant fluctuations, such as temperature and vibration, were marked as key features.

[0061] During the processing of medium- and low-risk data, the GAN model sequentially identifies the most relevant feature points for each risk level based on the feature weight matrix and creates a list of feature points for each risk level. Detailed data on key feature points is generated for each of the three risk levels: high, medium, and low. Anomaly thresholds are set for each feature point to enable real-time monitoring of device failures during the data collection process.

[0062] The resulting fourth dataset contains the fault type, probability of failure, and feature weight matrix for each device, along with key feature points at different risk levels. This dataset will be transferred to the data governance module for health score calculation and classification management.

[0063] Furthermore, the feature weight is the importance parameter of each equipment fault feature (such as temperature, vibration, humidity, etc.) in the fault prediction model, which is used to reflect the degree of influence of different features on the equipment failure risk.

[0064] Weights are determined through data analysis and model training. Specifically, during GAN model training, each feature is weighted based on its contribution to fault prediction accuracy. Features with higher contribution rates are assigned higher weights, indicating a greater impact on fault occurrence. Feature weights range from 0 to 1, with larger weights indicating a more significant impact on the device health score.

[0065] Key feature points (KFPs) are abnormal parameters or highly sensitive features that appear when a device experiences different fault levels. The GAN model monitors these KFPs frequently to improve fault identification accuracy. Key feature point identification criteria include: Points within the fault sample data where characteristic values frequently and significantly deviate are defined as KFPs. For example, when the frequency and amplitude of temperature and vibration fluctuations are significantly higher than normal, these parameters are marked as KFPs. The threshold for each KFP is determined through statistical analysis of historical device data to ensure that the KFPs accurately reflect the device's abnormal state.

[0066] The classified management of the equipment includes calculating a health score using the fault occurrence probability and feature weight in the fourth data set, and constructing a fifth data set based on the fault type and health score of each device;

[0067] Generate a master data mapping structure based on health scores and classify equipment risks into high-failure-risk equipment, medium-failure-risk equipment, and low-failure-risk equipment;

[0068] Generate task assignment list based on master data mapping structure and fault type;

[0069] The task allocation list includes a load allocation requirement interval for equipment at each risk level.

[0070] It should be noted that after receiving the fourth data set, the data governance module first calculates the health score of each device one by one based on the fault type, occurrence probability and weight value in the feature weight matrix.

[0071] The health score calculation process is as follows: the probability of occurrence of the main equipment fault types is weighted using feature weight values. Specifically, the weight of each fault feature is multiplied by its probability of occurrence, and all weighted values are accumulated to form the fault score.

[0072] For devices that exhibit multiple major fault types, the fault scores are normalized to ensure that the score range is between 0 and 100, with lower values indicating a higher risk of failure.

[0073] The generated fifth dataset contains the health score, main fault type, feature weight, and key feature point information for each device. This fifth dataset will be used for risk level classification and task assignment in subsequent steps.

[0074] Based on the health scores in the fifth data set, the risk level of the equipment is classified and a master data mapping structure is generated.

[0075] The specific classification criteria are as follows: High-failure-risk equipment: Scored between 0 and 30, marked as high-risk, suitable for low-load operation. The main failure characteristics of the equipment are bound to the master data mapping structure and configured as a high-priority monitoring object.

[0076] Medium Failure Risk Equipment: Scored between 30 and 70 points, marked as medium risk, suitable for medium-load operation. The primary failure characteristics of this equipment will be recorded within the mapping structure to maintain a moderate monitoring frequency during task allocation.

[0077] Low-failure-risk equipment: Scored between 70 and 100 points, marked as low-risk and suitable for high-load operation. The mapping structure of low-risk equipment records the main fault characteristics of low-frequency monitoring.

[0078] After the master data mapping structure is generated, each type of equipment is assigned a task label, which contains risk level and key fault feature monitoring information to ensure the accuracy of equipment classification management and the integrity of feature monitoring.

[0079] Based on the master data mapping structure, load distribution requirement intervals are configured for each risk level equipment, and the feature monitoring frequency is bound to the task allocation list.

[0080] To avoid load management inconsistency caused by vacant intervals, the setting of load intervals is rationally adjusted according to the following standards:

[0081] High-failure-risk equipment is set to operate at a low load of 0% to 30%. Key characteristic points (such as temperature and vibration) are collected every 5 seconds to quickly monitor changes in equipment status.

[0082] Equipment with a medium failure risk is set to operate at a medium load between 30% and 70%. Key feature point data collection for the equipment is set to once every 10 seconds to ensure that medium-risk features can be effectively monitored at a lower frequency.

[0083] The load range for low-failure-risk equipment is set between 70% and 100%, operating at high load. The collection frequency of key feature points is set to once every 15 seconds, allowing the equipment to achieve efficient load operation at a lower collection frequency.

[0084] The task allocation list details each machine's load range, key fault signatures, and acquisition frequency, providing clear load management requirements for the production strategy optimization module. Transmitting the task allocation list to the production strategy optimization module ensures that each machine receives appropriate load control and monitoring support based on its risk level and health status during the load allocation process.

[0085] The utilizing the reinforcement learning algorithm to allocate production tasks includes determining a load allocation value for each device within a load allocation requirement interval using the reinforcement learning algorithm;

[0086] Evaluate the health status of the equipment after the re-load distribution, update the master data mapping structure, and if the equipment is still at high risk of failure or the risk classification level has increased, execute the load secondary distribution strategy until the risk classification level decreases;

[0087] If the equipment is not judged as a high-failure-risk equipment or the risk classification level is reduced, the risk classification level is fed back to the data acquisition module to adjust the frequency of operation data collection.

[0088] It should be noted that the production strategy optimization module receives the task allocation list generated in step 3 and sets initial load distribution values based on load intervals of different risk levels. The system uses a reinforcement learning algorithm to determine the load distribution for each device and sets the target value for the device load.

[0089] The specific implementation process of the reinforcement learning model is as follows:

[0090] The state space includes the current health score of the equipment, the frequency of key feature point acquisition, environmental conditions, and the current load distribution value.

[0091] The action space is set to three operations: load increase, load maintenance, or load reduction. The load range and the health status of the equipment jointly determine the range of load adjustment.

[0092] The reward function system calculates rewards based on changes in the device's health score and fluctuations in its characteristic points after load allocation. If the device's health score remains or improves under the current load, a positive reward is given. If the health score decreases, the system will give a negative reward based on the magnitude of the decrease, adjusting subsequent load decisions.

[0093] The reinforcement learning model gradually optimizes the load distribution strategy based on reward feedback, configuring low loads for high-risk equipment, medium loads for medium-risk equipment, and high loads for low-risk equipment. It also records the changes in the health status of the equipment and key feature points after allocation in real time.

[0094] The execution of the load secondary distribution strategy includes reducing the equipment load to the minimum load mode when the equipment risk level is still high failure risk;

[0095] When the risk level rises from low failure risk equipment to medium failure risk equipment, immediately adjust the load to the lower limit of the medium risk range.

[0096] It should be noted that if the equipment risk level remains high after the initial load adjustment, the equipment load is reduced to the minimum safe value, and the load adjustment module activates a "minimum load lockout" mechanism to maintain the load at or below the minimum safe value for five minutes. During this period, the acquisition module activates full-parameter monitoring mode. In addition to high-risk characteristics such as temperature and vibration, it also monitors secondary fault-related parameters such as humidity and voltage fluctuations, with a set acquisition frequency of every five seconds. The acquisition module transmits data in real time to the GAN model, which uses data from all characteristic parameters to generate updated fault predictions, which are continuously updated to the health scoring module. The health scoring module assesses the equipment health status every five minutes. If a significant increase in the equipment health score is detected, the load control module releases the lockout and gradually increases the load level to return to the median load. If there is no significant change in the score, the load remains at the minimum level and locked until the next assessment.

[0097] When the equipment risk level rises from low failure risk to medium failure risk, the equipment load is adjusted to the lower limit of the medium risk range, and the load control module is set to allow the equipment load to recover dynamically as the score changes. The acquisition module analyzes the main fault characteristics that caused the score to drop, and increases the frequency of data collection related to this feature to once every 3 seconds (such as temperature or vibration). The GAN model updates the feature priority weight in real time based on this data, and directly feeds back the collected feature data to the load control module during the operation of the equipment. The load control module continuously monitors the health score trend within 5 minutes; if the score tends to stabilize or increase, the load is gradually increased in increments of 10% to 15%, and continues to monitor the fault feature in real time when it returns to the normal load level to ensure that the equipment remains stable when it returns to the target load.

[0098] Example 2

[0099] Reference Figure 2 , which is an embodiment of the present invention, provides a new energy power station master data management system based on artificial intelligence, including:

[0100] Data collection module 100, fault identification module 200, data governance module 300 and production strategy optimization module 400;

[0101] The data acquisition module 100 collects the operating data of the new energy power station equipment;

[0102] First, we normalize the equipment's operating status data (power output, temperature, vibration), environmental condition data (temperature, humidity, wind speed), and historical equipment data (historical fault records and operating performance information) to ensure that all data fits within the normalized range of 0 to 1. This step ensures that all feature points are scaled consistently, preventing the subsequent clustering algorithm from being affected by different dimensions.

[0103] The standardized first data set was grouped using the K-means clustering algorithm, with the specific clustering parameter set to K=3, where K represents three cluster centers, representing high risk, medium risk, and low risk levels, respectively.

[0104] Weighted formula: Use the formula

[0105]

[0106] Among them, W i is the weighted value of the i-th feature, f i is the frequency of feature occurrence (based on historical data statistics), A i is the fluctuation amplitude of the feature. This formula ensures that high-frequency and high-fluctuation features are given higher weights, making these features account for a larger proportion in fault prediction.

[0107] Devices were categorized as high, medium, and low risk based on K-means clustering. High-risk devices had feature points significantly exceeding the standard (with a deviation greater than 2 standard deviations), medium-risk devices had feature points close to the standard (with a deviation between 1 and 2 standard deviations), and low-risk devices had feature points below the standard (with a deviation less than 1 standard deviation). The resulting second dataset contained detailed information about each device's risk level, collection frequency, and feature points.

[0108] Through dynamic weighting and a multi-level collection strategy, the system can adjust collection resources for devices with different risk levels, avoiding resource waste and focusing on high-risk devices. This step expands fault identification from a single-dimensional analysis to a comprehensive multi-dimensional analysis, improving the system's flexibility and efficiency in real-time monitoring.

[0109] The fault identification module 200 analyzes the equipment failure mode based on the fault identification algorithm and determines the risk level of the equipment;

[0110] The generator in the GAN model is used to generate virtual fault samples, and the discriminator is used to compare real data with generated data. The generator loss function is set as:

[0111] Loss G = -log(D(G(z)))

[0112] Among them, G(z) is the output of the generator and D is the discriminator. The loss function of the discriminator is set as:

[0113] Loss D = -log(D(x))-log(1-D(G(z)))

[0114] The generator inputs noise and device characteristics, and simulates the performance of devices at different risk levels through the generated virtual fault samples.

[0115] The discriminator compares the generated samples with the actual samples and adjusts the parameters of the generator so that it can generate virtual samples that are close to real device failures.

[0116] Feature weight allocation and sensitivity calculation: During the training process, the impact of each feature on the output error of the GAN model is calculated using the formula:

[0117]

[0118] Among them, x i is the value of the i-th feature. The sensitivity of the feature is obtained by calculating the partial derivative, and the corresponding weight is assigned. The weight range is 0 to 1, ensuring that features with a greater impact on the fault receive higher priority.

[0119] Based on the risk level of the device (high, medium, low), different training strategies are gradually introduced:

[0120] The number of training times for high-risk devices is set to 200 times, and the loss function weight is higher to ensure more accurate recognition of key feature points.

[0121] The number of training times for medium-risk equipment is set to 150 times, and a lower-weight loss function is used to ensure the recognition of moderate fault features.

[0122] The number of training times for low-risk equipment is set to 100 times, and only basic feature points are trained to identify basic fault characteristics of the equipment.

[0123] The GAN model's step-by-step training method provides the system with a flexible, multi-level fault prediction solution, improving the accuracy of fault identification at different risk levels. By calculating the sensitivity of feature weights, equipment fault analysis is more accurate, providing precise data for subsequent health scoring.

[0124] The data governance module 300 generates health status reports based on risk levels and manages devices by category;

[0125] Weighted average calculation: Combine the feature weights and the probability of failure to calculate the health score using the formula:

[0126] Health Score=100-∑(P i ×W i )

[0127] Among them, P i is the probability of occurrence of the i-th fault feature, W i is the feature weight. The results are normalized to between 0 and 100, with lower values indicating higher risks.

[0128] The classification criteria for health scores are as follows: high failure risk is 0-30 points; medium failure risk is 31-70 points; and low failure risk is 71-100 points.

[0129] The combination of health scores and a master data mapping structure enables the system to accurately manage devices in a hierarchical manner. This mapping structure further refines task allocation requirements, tying device monitoring to load intervals and providing a clear basis for load allocation within the reinforcement learning module.

[0130] The production strategy optimization module 400 uses a reinforcement learning algorithm to allocate production tasks and optimize the operating load of the equipment.

[0131] Specific parameters of the reinforcement learning model:

[0132] The state space includes health score, key feature collection frequency, environmental conditions, current load, etc.

[0133] The range of motion allows you to increase, maintain, or decrease the load. The load range determines the range within which the load can be adjusted.

[0134] Reward function When the health score increases, the positive reward is:

[0135] R=+10×(Health Score Improvement)

[0136] When the rating decreases, the negative reward is:

[0137] R=-20×(Health Score Decrease)

[0138] The reinforcement learning algorithm dynamically optimizes the load distribution value using a reward function in each evaluation and adjusts the load based on the score.

[0139] The combination of reinforcement learning and a secondary load adjustment strategy ensures optimal load management through real-time control when equipment failure risks occur. This closed-loop control strategy enables load distribution to flexibly respond to changes in equipment status, ultimately achieving a balance between equipment operating efficiency and safety.

[0140] Example 3

[0141] An embodiment of the present invention is different from the previous two embodiments in that:

[0142] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0143] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0144] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0145] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0146] Example 4

[0147] As an embodiment of the present invention, a master data management method for a new energy power station based on artificial intelligence is provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0148] This experiment divided 30 devices (10 each of photovoltaic panels, wind turbines, and battery systems) in a new energy power station into two groups: a traditional method group and an innovative method group. The experiment lasted four months and covered various environmental conditions (such as temperature, humidity, and wind speed). The innovative method group employed a combination of generative adversarial networks (GANs) and reinforcement learning algorithms to achieve real-time fault prediction and dynamic load adjustment, while the traditional method group used static load management and fixed-cycle fault detection. During the experiment, the innovative method group implemented a high-frequency data collection mechanism, collecting and analyzing data on device operating status, environmental conditions, and historical faults every five seconds; the traditional method group adopted a daily static data recording and processing strategy. The fault detection and load adjustment strategies also differed: the innovative method group used GANs to identify features and dynamically adjust the load upon detecting an increased risk level, activating load monitoring mode; the traditional method group implemented a pre-defined load handling plan upon detecting an anomaly. Through multi-dimensional data collection and feature analysis, this experiment compared the two methods in terms of fault prediction accuracy, response speed, failure rate, and equipment lifespan.

[0149] The experiment focused on four key metrics: fault prediction accuracy, load adjustment response time, equipment failure rate, and equipment lifespan. Fault prediction accuracy was assessed by real-time analysis of collected data using a GAN model, generating multi-dimensional fault predictions based on fault type, probability, and feature weights. The traditional approach relied on static data for fault prediction, recording faults based on daily inspection results. Load adjustment response time measured the time it took the system to complete load adjustment after detecting an increase in risk level. For equipment failure rate, the system counted the frequency of each device's failures during the experimental period and prioritized specific fault characteristics using the GAN model in the inventive approach. The traditional approach recorded the total number of failures and compared them with daily inspection data. Finally, the experiment also measured the extension of equipment lifespan, primarily based on changes in health scores under different load conditions. The inventive approach employed a minimum load mode and dynamic load recovery strategy, monitoring score fluctuations in real time. The traditional approach maintained a fixed load adjustment strategy and recorded changes in the health status of each device. The experimental results are shown in Table 1.

[0150] Table 1 Experimental data comparison table

[0151]

[0152] Experimental results show that the present invention is significantly superior to traditional methods in terms of fault prediction accuracy. Traditional methods rely on fixed-cycle data collection and static load adjustment, and it is difficult to integrate multi-dimensional data for real-time analysis, and therefore cannot accurately identify complex multi-fault features. The present invention uses a generative adversarial network (GAN) model to achieve high-frequency collection and fusion analysis of multi-source data, so that the weights of key fault features can be dynamically adjusted. The GAN model can process equipment with different risk levels step by step, and generate virtual fault samples under high-risk conditions, and perform comparative analysis in combination with actual data to effectively identify potential fault modes. In this way, the GAN model can not only significantly improve the accuracy of fault prediction, but also has the ability to respond to the health status of the equipment in real time. The technical solution of multi-feature fusion and dynamic weighting makes fault prediction more sensitive, avoiding the untimely prediction and misjudgment caused by data lag in traditional methods.

[0153] In addition, in terms of the response time of load adjustment, the inventive method utilizes the real-time feedback mechanism of the reinforcement learning algorithm to adjust the load immediately after detecting an increase in risk. The reinforcement learning algorithm automatically activates the load adjustment mode based on the health score and fault risk level of the current equipment, reduces the load to the minimum safe level, and continues to monitor during the lock period. The algorithm can continuously optimize the load adjustment strategy according to actual conditions, forming a closed-loop feedback system, so that the load adjustment is rapid and accurate. Traditional methods rely on fixed load management solutions for load adjustment, lack the flexibility of real-time adjustment, have a lagging response speed, and cannot respond to sudden risk changes in a timely manner. This real-time feedback mechanism enables the present invention to complete load adjustment within a short period of time after fault detection, reducing the direct impact of fault risks on equipment and ensuring that load adjustment is synchronized with the health status of the equipment.

[0154] In terms of the incidence of equipment failures, the strategy of combining GAN with reinforcement learning algorithms adopted in the present invention achieves closed-loop control through load adjustment and fault feature monitoring, significantly reducing the failure frequency of equipment under high-load conditions. Since traditional methods cannot prioritize monitoring specific fault features with increased risks, equipment is prone to failures under high-load operation due to lack of real-time monitoring. The GAN model in the inventive method prioritizes the collection and analysis of characteristic data related to faults when risks increase, and can promptly identify the potential failure trends of the equipment, and quickly adjust the load through reinforcement learning algorithms to form a multi-level risk response plan. This dynamic adjustment based on real-time data significantly alleviates the pressure on equipment components under high-load conditions and effectively reduces the loss of equipment under high-load conditions.

[0155] Finally, in terms of extending the life of the equipment, the present invention reduces the overall wear of the equipment and extends its service life through a minimum load mode and a gradual load recovery strategy. Traditional methods rely on fixed load adjustment strategies and lack a real-time response mechanism to the equipment health score, which results in severe wear of the equipment when it operates under high load for a long time and cannot effectively extend the service life of the equipment. The present invention uses a reinforcement learning algorithm to control the load to the lowest level under high-risk conditions, and gradually restores the load when the equipment score improves, ensuring that the equipment is always in an adaptive load state. Through dynamic load management, the equipment can reduce wear in a high-load operating environment, and gradually restore the load to a safe level when the score improves, thereby effectively extending the life of the equipment. This dynamic adjustment strategy not only reduces the load pressure on key components of the equipment, but also enhances the adaptability of the equipment under complex operating conditions.

[0156] In general, the present invention forms an intelligent closed-loop management system for fault prediction and load adjustment by combining GAN with reinforcement learning algorithms. This solution can analyze the status of equipment in real time, quickly respond to risk changes, optimize the operating load of equipment, significantly improve the stability and life of the equipment, and is significantly better than the fixed adjustment strategy of traditional methods. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A new energy power station master data management method based on artificial intelligence, characterized in that: include: The data acquisition module collects the operating data of new energy power station equipment; The fault identification module analyzes the equipment failure mode based on the fault identification algorithm and determines the risk level of the equipment; The data governance module generates health status reports based on risk levels and manages devices by category; The production strategy optimization module uses reinforcement learning algorithms to allocate production tasks and optimize the operating load of equipment.

2. The artificial intelligence-based new energy power station master data management method according to claim 1, characterized in that: The collecting of the operating data of the new energy power station equipment includes collecting equipment operating status data, environmental condition data and equipment history data from photovoltaic components, wind turbines and battery systems in the new energy power station to generate a first data set; Performing a multi-level analysis on the first data set using a feature extraction algorithm to assign an acquisition risk level to the device, the acquisition risk level including high acquisition risk, medium acquisition risk, and low acquisition risk; The acquisition strategy is formulated according to the acquisition risk level, and a third data set containing multi-level acquisition content is generated and transmitted to the fault identification module.

3. The artificial intelligence-based new energy power station master data management method according to claim 2, characterized in that: Analyzing the equipment failure mode based on the fault identification algorithm includes inputting the third data set data into a generative adversarial network algorithm to construct a multi-level fault prediction model; The data in the third data set are processed level by level according to the order of acquisition risk levels to generate a fourth data set including fault type, fault probability and feature weight.

4. The artificial intelligence-based new energy power station master data management method according to claim 3, characterized in that: The classified management of the equipment includes calculating a health score using the fault occurrence probability and feature weight in the fourth data set, and constructing a fifth data set based on the fault type and health score of each device; Generate a master data mapping structure based on health scores and classify equipment risks into high-failure-risk equipment, medium-failure-risk equipment, and low-failure-risk equipment; Generate task assignment list based on master data mapping structure and fault type; The task allocation list includes a load allocation requirement interval for equipment at each risk level.

5. The artificial intelligence-based new energy power station master data management method according to claim 4, characterized in that: The utilizing the reinforcement learning algorithm to allocate production tasks includes determining a load allocation value for each device within a load allocation requirement interval using the reinforcement learning algorithm; Evaluate the health status of the equipment after the re-load distribution, update the master data mapping structure, and if the equipment is still at high risk of failure or the risk classification level has increased, execute the load secondary distribution strategy until the risk classification level decreases; If the equipment is not judged as a high-failure-risk equipment or the risk classification level is reduced, the risk classification level is fed back to the data acquisition module to adjust the frequency of operation data collection.

6. The artificial intelligence-based new energy power station master data management method according to claim 5, characterized in that: The execution of the load secondary distribution strategy includes reducing the equipment load to the minimum load mode when the equipment risk level is still high failure risk; When the risk level rises from low failure risk equipment to medium failure risk equipment, immediately adjust the load to the lower limit of the medium risk range.

7. A system using the artificial intelligence-based new energy power station master data management method according to any one of claims 1 to 6, characterized in that: include: A data acquisition module (100), a fault identification module (200), a data management module (300), and a production strategy optimization module (400); The data acquisition module (100) collects operating data of new energy power station equipment; The fault identification module (200) analyzes the equipment failure mode based on the fault identification algorithm and determines the risk level of the equipment; The data governance module (300) generates a health status report based on the risk level and manages the equipment by category; The production strategy optimization module (400) uses a reinforcement learning algorithm to allocate production tasks and optimize the operating load of equipment.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the new energy power station master data management method based on artificial intelligence according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the new energy power station master data management method based on artificial intelligence according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the new energy power station master data management method based on artificial intelligence described in any one of claims 1 to 6 are implemented.

Images