New energy equipment health prediction and intelligent maintenance method based on multi-source data fusion

Abstract

The invention discloses a new energy equipment health prediction and intelligent maintenance method based on multi-source data fusion, and the method comprises the steps: building a dynamic health state evaluation model through a system in combination with equipment operation parameters and environment variables; monitoring the running state of the equipment in real time, and grading the health degree of the equipment; based on fault prediction of data, a deep learning algorithm is introduced to carry out modeling on time series data, and possible fault occurrence time and types of equipment are predicted by capturing potential trends and change modes in operation of the equipment; a dynamic prediction function is provided, a prediction result is updated in time, and synchronism of a model and an actual equipment state is ensured; and generating an optimized maintenance plan according to the health diagnosis and prediction result. Through innovative design of a multi-source data fusion technology and an intelligent prediction algorithm, a core pain point in operation and maintenance of current new energy equipment is solved, a feasible path is provided for full-life-cycle management of the equipment, and the method has remarkable economic value and industry popularization potential.

Description

Technical Field

[0001] The present invention relates to the field of new energy intelligent operation and maintenance technology, and specifically to a new energy equipment health prediction and intelligent maintenance method based on multi-source data fusion. Background Art

[0002] As the global demand for renewable energy continues to grow, photovoltaic, wind power, energy storage and charging equipment are increasingly used in new energy systems. However, these devices face many challenges in operation and maintenance, and traditional operation and maintenance methods can no longer meet the needs of modern operation and maintenance management, which are mainly manifested in the following aspects:

[0003] 1. Reliance on manual inspections leads to low O&M efficiency

[0004] The traditional operation and maintenance management method of new energy equipment is mainly based on manual inspection, which is to regularly dispatch operation and maintenance personnel to the site to check the operating status of the equipment, including equipment appearance inspection, manual data recording and operating status evaluation. This method has the following problems:

[0005] Low coverage efficiency: New energy equipment is usually widely distributed, such as wind farms located in remote mountainous areas or coastal areas, and photovoltaic power stations distributed in vast deserts. The geographical distribution characteristics of the equipment lead to inspection work that consumes a lot of time and resources.

[0006] Reliance on experience: Operation and maintenance personnel usually judge the operating status of the equipment based on subjective experience. The limitations and differences in experience may lead to hidden dangers not being discovered in time or misjudgment of equipment status.

[0007] Delayed response: Manual inspections often require waiting for the equipment to become obviously abnormal or completely shut down before intervention, missing the best time for maintenance and further aggravating the extent of equipment damage.

[0008] Shortage of human resources: With the rapid expansion of the new energy industry, the training speed of operation and maintenance personnel is far behind the growth rate of new equipment, and the shortage of high-skilled talents is particularly prominent.

[0009] The above problems make it difficult for manual inspection methods to adapt to the operation and maintenance needs of large-scale new energy equipment, and more intelligent and automated solutions are urgently needed.

[0010] 2. Equipment data is not fully utilized and the value of information is not mined

[0011] New energy equipment will generate a large amount of real-time data during operation, including voltage, current, temperature, vibration, wind speed, radiation intensity and other multi-dimensional operating parameters. These data have the following characteristics:

[0012] Large amount of data: Modern photovoltaic power stations, wind farms and energy storage equipment are equipped with high-precision sensors, which collect thousands of data per second. The amount of data collected by a large power station can reach several TB or even PB levels every day.

[0013] Complex data: Data comes in various forms, including time series data, event-driven data, and image data. The data protocols and collection standards between devices are not unified, which makes data analysis difficult.

[0014] Strong real-time performance: operating data is generated in seconds or minutes, and traditional data analysis methods cannot process these dynamically changing signals in real time.

[0015] Despite the huge amount of data and its high potential value, the current industry generally has the following problems:

[0016] Data island phenomenon: Device data is stored in different devices or platforms in a scattered manner. There is a lack of effective integration mechanism, making it difficult to form a complete data chain.

[0017] Lack of in-depth analysis: Most data is only used for simple status monitoring or historical records, but fails to achieve higher-level analysis goals such as equipment operation trend prediction and health status assessment through in-depth mining.

[0018] Ignoring environmental impacts: Equipment operating data is closely related to the external environment (such as climate conditions and power grid fluctuations). The lack of comprehensive analysis capabilities for environmental variables may lead to inaccurate prediction results.

[0019] Therefore, how to fully tap the value of data and establish an intelligent analysis model for multi-dimensional data has become a key technical problem in the field of new energy equipment operation and maintenance.

[0020] 3. Delayed response to failures and lack of predictive maintenance capabilities

[0021] The core goal of new energy equipment operation and maintenance management is to ensure the stability and safety of the equipment. However, the traditional operation and maintenance management model tends to be more passive maintenance, that is, repairing the equipment after it fails, rather than intervening in advance when potential risks accumulate. The disadvantages of this approach mainly include:

[0022] Long downtime: Equipment failure usually requires multiple steps such as problem confirmation, spare parts deployment and maintenance operations. Especially for widely distributed equipment, transportation and coordination take longer. This directly leads to long-term equipment inoperability, resulting in loss of power generation or reduced energy storage capacity.

[0023] High maintenance costs: In reactive maintenance, equipment is often not discovered until the fault has expanded to a serious level, resulting in an expanded maintenance scope, which may require the replacement of expensive core components or even the entire machine.

[0024] Hidden dangers of secondary failures: Failures that are not maintained in time may trigger a chain reaction. For example, a bearing failure in a wind turbine may cause abnormal vibration of the entire machine, or even damage the main shaft; a failure in the cooling system of a photovoltaic inverter may cause the circuit to overheat, leading to more serious problems.

[0025] Lack of global planning: The operation and maintenance strategy lacks scientific planning and is based only on the randomness of failures, resulting in unreasonable scheduling of maintenance resources and easily causing excessive operation and maintenance load in a short period of time.

[0026] Currently, the industry is gradually aware of the importance of predictive maintenance, but faces the following challenges:

[0027] Data modeling is difficult: The operating environment of new energy equipment is complex and the influencing factors are diverse. Existing prediction models often cannot accurately describe the dynamic changes in the health status of equipment.

[0028] Lack of a universal method: The fault characteristics of different types of equipment (photovoltaic, wind power, energy storage, charging equipment) vary significantly, making it difficult to build a universal prediction algorithm.

[0029] Insufficient updating and adaptability: Existing prediction models are mostly based on offline training, which makes it difficult to dynamically adapt to changes in equipment operating conditions, causing the models to gradually become ineffective in actual applications.

[0030] Summary of current situation:

[0031] In summary, the current operation and maintenance management model of new energy equipment has significant limitations: low efficiency of manual inspection, insufficient utilization of equipment data, and delayed response to faults. These problems not only restrict the operating efficiency of new energy equipment, but also directly affect the economic benefits and return on investment of new energy power stations. Against the background of the rapid expansion of the new energy industry, it is urgent to introduce intelligent and automated operation and maintenance technologies to achieve real-time monitoring, accurate prediction and scientific maintenance of the health status of equipment, so as to improve the overall operation and maintenance management level. Summary of the invention

[0032] The present invention proposes a new energy equipment health prediction and intelligent maintenance method based on multi-source data fusion, especially an intelligent operation and maintenance management method and system based on equipment health diagnosis and prediction algorithm for photovoltaic, wind power, energy storage and charging equipment, aiming to realize health status monitoring, fault prediction and intelligent preventive maintenance of new energy equipment. To achieve the above purpose, the present invention adopts the following technical solutions:

[0033] A method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion, which performs the following steps through computer equipment:

[0034] Including, multi-dimensional data fusion: the system combines equipment operating parameters and environmental variables to build a dynamic health status assessment model;

[0035] Health status classification: real-time monitoring of equipment operating status and grading of equipment health;

[0036] Data-based fault prediction introduces deep learning algorithms to model time series data, and predicts the time and type of possible equipment failures by capturing potential trends and change patterns in equipment operation. It provides dynamic prediction functions, updates prediction results in a timely manner, and ensures the synchronization of the model with the actual equipment status.

[0037] Generate an optimized maintenance plan based on the health diagnosis and prediction results, including recommended inspection time, required spare parts and maintenance operation guidelines.

[0038] Furthermore, it also includes using concept drift detection technology to monitor changes in device data distribution and dynamically adjust model parameters to adapt to environmental changes.

[0039] Furthermore, a dynamic health status assessment model is constructed, including:

[0040] Data acquisition and preprocessing module, health status assessment module, fault prediction and early warning module, intelligent maintenance decision module and dynamic optimization and learning module.

[0041] Furthermore, the data acquisition and preprocessing module steps include:

[0042] The system collects equipment operation data and environmental data through the sensor network deployed on the equipment;

[0043] X run ={x v , x i , x t , x p , x a}, X env ={x irr , x wind , x hum , x press},in:

[0044] XrunX_{run}: equipment operation data set, including voltage (xvx_{v}), current (xix_{i}), temperature (xtx_{t}), power (xpx_{p}) and vibration (xax_{a});

[0045] XenvX_{env}: Environmental dataset, including light intensity (xirrx_{irr}), wind speed (xwindx_{wind}), humidity (xhumx_{hum}), and air pressure (xpressx_{press});

[0046] Data cleaning and preprocessing:

[0047] Missing value completion includes using the mean filling method for missing data xmissx_{miss}:

[0048] Or the interpolation algorithm:

[0049] Normalization:

[0050] Use Min-Max normalization to map the data to [0, 1][0, 1]:

[0051]

[0052] Use Kalman filter to eliminate noise:

[0053] x k =x k-1 +K k (z k -Hx k-1 ),

[0054] Where KkK_k is the Kalman gain, zkz_k is the measurement value, and HH is the measurement matrix.

[0055] Generate time series samples using a sliding window:

[0056] X window ={x t , x t+1 , ..., x t+W},

[0057] Where WW is the window length;

[0058] Data integration and storage:

[0059] The edge computing node performs the following data storage operations locally:

[0060] X local =Process(X raw ),

[0061] Among them, XlocalX_{local} is the locally processed data, and XrawX_{raw} is the original collected data; the processing results are regularly uploaded to the cloud for in-depth analysis and backup.

[0062] Furthermore, the health status assessment module processing steps include:

[0063] Multi-dimensional feature fusion: Health feature vector F{health} extracted through feature engineering:

[0064] F health =φ(X run , X env ),

[0065] in is the feature mapping function;

[0066] Real-time status diagnosis: Use classification models to evaluate the health status of equipment in real time;

[0067] The mathematical expression of status diagnosis is:

[0068]

[0069] Where P(C|Fhealth)P(C|F_{health}) is the probability of state CC, i.e. normal, subhealthy, and faulty, and fc(·)f_c(\cdot) is the score function of the classifier;

[0070] Dynamic visualization of health status:

[0071] I health =Plot(F health , t),

[0072] Where IhealthI_{health} is a graph showing the health index changing over time.

[0073] Furthermore, the fault prediction and warning module processing steps include:

[0074] Prediction model construction: Use LSTM model to predict the time T{fault} when the fault occurs:

[0075] h t =σ(W h ·h t-1 +W x ·x t +b h ), y t =W y ·h t +b y ,

[0076] Where hth_t is the hidden state, yty_t is the predicted output, WhW_h, WxW_x, WyW_y, bhb_h, byb_y are parameters;

[0077] Fault type identification:

[0078] Fault classification algorithm based on decision tree:

[0079]

[0080] Where P(C|Xfault)P(C|X_{fault}) is the category probability of the fault feature vector XfaultX_{fault};

[0081] Threshold warning mechanism:

[0082] Dynamic threshold setting,

[0083]

[0084] Where θ\theta is the dynamic threshold.

[0085] Furthermore, the processing steps of the intelligent maintenance decision module include:

[0086] Maintenance plan generation:

[0087] Optimization goal: Minimize the total maintenance cost CtotalC_{total}:

[0088]

[0089] Where Ccheck,iC_{check,i} and Crepair,iC_{repair,i} are the inspection and repair costs respectively;

[0090] Multi-device collaborative optimization: Establish scheduling optimization problems,

[0091]

[0092] Where Tdown,iT_{down,i} is the downtime of device ii.

[0093] Furthermore, the dynamic optimization and learning module processing steps include:

[0094] Model adaptive update:

[0095] Update model parameters using online learning algorithms:

[0096]

[0097] Where θt\theta_t is the updated parameter, η\eta is the learning rate, and LL is the loss function;

[0098] Concept Drift Detection:

[0099] Use KL divergence to determine data distribution changes:

[0100]

[0101] Where P(i)P(i) and Q(i)Q(i) are the distributions of current and historical data respectively;

[0102] Feedback learning mechanism: Use maintenance data to train new models,

[0103] Train(M,X new )→M′,

[0104] Where MM is the original model, M′M' is the updated model, and XnewX_{new} is the new dataset.

[0105] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0106] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0107] It can be seen from the above technical solution that the present invention proposes an equipment health diagnosis and prediction system and method based on intelligent algorithms. By deeply mining the potential value of equipment operation data and external environment data, a predictive maintenance framework is established to improve the intelligence level of equipment operation and maintenance management, and to achieve active discovery, precise positioning and scientific maintenance of equipment faults.

[0108] This paper proposes a new energy equipment health prediction and intelligent maintenance method and system based on multi-source data fusion, aiming to solve the problems of low efficiency, delayed fault response and difficulty in quantifying equipment health status in the operation and maintenance management of new energy equipment. Through the deep combination of multi-dimensional data fusion and intelligent algorithms, the system has the following technical advantages:

[0109] 1. Data-driven equipment health prediction and maintenance

[0110] The present invention builds a comprehensive health prediction model by integrating equipment operation data (such as voltage, current, power, etc.) and environmental impact data (such as light intensity, wind speed, humidity, etc.), achieving a leap from single data dimension to multi-dimensional data fusion.

[0111] The fusion of multi-source data makes equipment health prediction more comprehensive, and can more accurately capture the correlation between the equipment's operating status and environmental factors, thereby improving prediction accuracy and reliability.

[0112] 2. Real-time health status monitoring and accurate diagnosis

[0113] By constructing a multi-dimensional feature vector of the equipment health status and combining it with a machine learning algorithm, the system can classify the equipment operation status in real time and generate real-time assessment results of "normal", "sub-healthy" and "faulty" status.

[0114] The system uses dynamic data streams to continuously monitor the health status of equipment, ensuring early detection of problems and providing sufficient time and basis for subsequent maintenance decisions.

[0115] 3. Intelligent fault prediction capability

[0116] The present invention uses time series data modeling technology (such as LSTM or random forest) to perform intelligent analysis on equipment historical data and real-time data to predict the time and type of potential failure of the equipment in the future.

[0117] By accurately predicting fault trends, the system can effectively avoid the expansion of faults, reduce the probability of unplanned downtime, and provide users with efficient predictive maintenance solutions.

[0118] 4. Efficient maintenance plan generation

[0119] Based on the equipment health prediction results, the system generates an optimized maintenance plan, including inspection time, operation recommendations and a list of required spare parts.

[0120] By introducing a multi-device collaborative optimization algorithm, it is possible to optimize maintenance resource allocation in large-scale scenarios such as photovoltaic power plants or wind farms, minimizing downtime and the impact on energy output.

[0121] 5. Strong dynamic adaptability

[0122] The algorithm design of the present invention can dynamically adapt to changes in the equipment operating environment, such as fluctuations in external climate conditions or performance degradation caused by equipment aging.

[0123] The system's continuous learning mechanism allows the prediction model to be automatically updated after new equipment operation data is collected, so that the model always maintains efficient and accurate prediction capabilities.

[0124] 6. Full life cycle health management

[0125] The system not only focuses on the current operating status of the equipment, but also evaluates the remaining service life of the equipment through predictive algorithms, providing support for the health management of the equipment throughout its life cycle.

[0126] Through the quantified health index, users can intuitively understand the usage status of the equipment and formulate more scientific operation and maintenance strategies.

[0127] 7. Improve operation and maintenance efficiency and economy

[0128] The present invention significantly reduces the reliance on manual inspections, realizes all-weather equipment status monitoring and automatic prediction through intelligent algorithms, and improves operation and maintenance efficiency.

[0129] The system reduces the total cost of operation and maintenance by identifying potential equipment hazards in advance, reducing the damage rate of expensive components and avoiding large-scale equipment failures.

[0130] 8. Modular architecture design facilitates expansion

[0131] The system adopts a modular design, and each functional module (such as data collection, health assessment, fault prediction and maintenance decision) can be deployed independently or used in combination according to needs to adapt to new energy project scenarios of different scales and complexities.

[0132] It has strong scalability and is suitable for a variety of new energy equipment such as photovoltaic, wind power, energy storage and charging equipment, and has wide industry adaptability.

[0133] 9. Maximize the use of data value

[0134] The system integrates and analyzes a large amount of equipment operation data and environmental data, providing panoramic data insights into the operation of new energy equipment.

[0135] Through deep data mining and algorithm optimization, not only can accurate prediction of equipment health status be achieved, but also valuable data information support is provided for future energy equipment research and development and operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0136] Figure 1 A schematic diagram of a health feature vector extracted by feature engineering in the present invention;

[0137] Figure 2 This is a schematic diagram of using maintenance data to train a new model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0138] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0139] The present invention aims to achieve the following key objectives:

[0140] 1. Intelligent equipment health management

[0141] The primary purpose of this invention is to build an intelligent system that can monitor the operating status of equipment in real time. Through multi-dimensional data analysis and advanced machine learning technology, the health status of equipment can be diagnosed and possible future failure trends can be predicted, breaking the limitations of traditional manual experience and realizing the digitalization and intelligence of operation and maintenance.

[0142] 2. Accurate prediction and early warning of faults

[0143] The present invention solves the problem of insufficient fault prediction capability of current new energy equipment. It can predict potential faults of equipment and issue warnings in advance by analyzing real-time data and historical data, thus avoiding serious consequences caused by failure to handle faults in a timely manner.

[0144] 3. Improve equipment operation efficiency

[0145] By proactively discovering potential equipment hazards and optimizing maintenance plans, the present invention significantly reduces the frequency of unplanned equipment downtime and improves equipment availability and energy output efficiency.

[0146] 4. Reduce operation and maintenance costs

[0147] The present invention reduces the frequency of manual inspections, optimizes the allocation of maintenance resources, avoids the expansion of equipment damage, and thus reduces the overall operation and maintenance costs through a predictive maintenance approach.

[0148] 5. Maximize data value

[0149] By using advanced data processing technology, we can fully tap the potential of equipment operation data, integrate scattered and heterogeneous data into a unified analysis platform, give data more business value, and provide support for the full life cycle management of new energy equipment.

[0150] The embodiment of the present invention proposes a comprehensive solution around the three core issues in the operation of new energy equipment: status monitoring, fault prediction and intelligent maintenance. The core ideas are as follows:

[0151] 1. Real-time monitoring and diagnosis

[0152] Multi-dimensional data fusion: The system combines equipment operating parameters and environmental variables to build a dynamic health status assessment model.

[0153] Health status classification: monitor the equipment operating status in real time and classify the equipment health level (such as normal, sub-healthy, and faulty).

[0154] 2. Data-based fault prediction

[0155] The innovation of the present invention lies in the introduction of deep learning algorithms (such as long short-term memory network LSTM) to model time series data, and predict the possible time and type of equipment failure by capturing the potential trends and change patterns in equipment operation.

[0156] Provide dynamic prediction function, update prediction results in time, and ensure the synchronization between the model and the actual equipment status.

[0157] 3. Intelligent maintenance plan

[0158] Based on the health diagnosis and prediction results, the system generates an optimized maintenance plan, including recommended inspection time, required spare parts and maintenance operation guidelines.

[0159] Supports multi-device collaborative maintenance strategies to minimize the impact of maintenance on equipment operations.

[0160] 4. Data-driven continuous optimization

[0161] The system uses concept drift detection technology to monitor changes in device data distribution and dynamically adjust model parameters to adapt to environmental changes.

[0162] Specifically,

[0163] The system architecture design and implementation method proposed in this invention are based on five core modules: data acquisition and preprocessing module, health status assessment module, fault prediction and early warning module, intelligent maintenance decision module and dynamic optimization and learning module. The following is a detailed description of each module, with additional relevant mathematical models and formulas to further illustrate its implementation details.

[0164] 1. System architecture design

[0165] The overall system architecture of the present invention uses a layered modular design, combined with edge computing and cloud collaboration, to achieve efficient data collection, status assessment, fault prediction and maintenance optimization.

[0166] 1.1 Data acquisition and preprocessing module

[0167] 1.1.1 Data Source

[0168] The system collects equipment operation data and environmental data through the sensor network deployed on the equipment.

[0169] X run ={x v ,xi ， x t , x p , x a}, X env ={x irr , x wind , x hum , x press},in:

[0170] XrunX_{run}: equipment operation data set, including voltage (xvx_{v}), current (xix_{i}), temperature (xtx_{t}), power (xpx_{p}), and vibration (xax_{a}).

[0171] XenvX_{env}: Environmental dataset, including light intensity (xirrx_{irr}), wind speed (xwindx_{wind}), humidity (xhumx_{hum}), and air pressure (xpressx_{press}).

[0172] 1.1.2 Data cleaning and preprocessing, such as Figure 1 Shown

[0173] Missing value completion

[0174] For missing data xmissx_{miss}, the mean filling method is used:

[0175] Or the interpolation algorithm:

[0176] Normalization

[0177] Use Min-Max normalization to map the data to [0, 1][0, 1]:

[0178]

[0179] Noise reduction

[0180] Use Kalman filter to eliminate noise:

[0181] x k =x k-1 +K k (z k -Hx k-1 ),

[0182] Where KkK_k is the Kalman gain, zkz_k is the measurement value, and HH is the measurement matrix.

[0183] Time series data segmentation

[0184] Generate time series samples using a sliding window:

[0185] X window ={x t , x t+1 , ..., x t+W}, where WW is the window length.

[0186] 1.1.3 Data Integration and Storage

[0187] The edge computing node performs the following data storage operations locally:

[0188] X local =Process(X raw ),

[0189] Among them, XlocalX_{local} is the locally processed data, and XrawX_{raw} is the original collected data. The processing results are regularly uploaded to the cloud for in-depth analysis and backup.

[0190] 1.2 Health status assessment module

[0191] 1.2.1 Multi-dimensional feature fusion

[0192] like Figure 1 As shown, the health feature vector F{health} extracted by feature engineering is:

[0193] F health =φ(X run , X env ),

[0194] in is the feature mapping function.

[0195] 1.2.2 Real-time status diagnosis

[0196] The classification model is used to evaluate the health status of the equipment in real time. Mathematical expression of status diagnosis:

[0197]

[0198] Where P(C|Fhealth)P(C|F_{health}) is the probability of state CC (normal, subhealthy, faulty), and fc(·)f_c(\cdot) is the score function of the classifier.

[0199] 1.2.3 Visualization

[0200] Dynamic visualization of health status:

[0201] I health =Plot(F health , t),

[0202] Where lhealthl_{health} is a chart showing the changes in health index over time.

[0203] 1_3 Fault prediction and warning module

[0204] 1_3.1 Prediction model construction

[0205] Use the LSTM model to predict the time T{fault} when a fault occurs:

[0206] h t =σ(W h ·h t-1 +W x ·x t +b h), y t =W y ·h t +b y ,

[0207] Where hth_t is the hidden state, yty_t is the predicted output, and WhW_h, WxW_x, WyW_y, bhb_h, byb_y are parameters.

[0208] 1_3.2 Fault type identification

[0209] Fault classification algorithm based on decision tree:

[0210]

[0211] Where P(C|Xfault)P(C|X_{fault}) is the category probability of the fault feature vector XfaultX_{fault}.

[0212] 1_3_3Threshold warning mechanism

[0213] Dynamic threshold setting:

[0214]

[0215] Where θ\theta is the dynamic threshold.

[0216] 1.4 Intelligent maintenance decision module

[0217] 1.4.1 Maintenance plan generation

[0218] Optimization goal: Minimize the total maintenance cost CtotalC_{total}:

[0219]

[0220] Where Ccheck,iC_{check,i} and Crepair,iC_{repair,i} are the inspection and repair costs respectively.

[0221] 1.4.2 Multi-device collaborative optimization

[0222] Establish a scheduling optimization problem:

[0223]

[0224] Where Tdown,iT_{down,i} is the downtime of device ii.

[0225] 1.5 Dynamic Optimization and Learning Module

[0226] 1.5.1 Model Adaptive Update

[0227] Update model parameters using online learning algorithms:

[0228]

[0229] Where θt\theta_t is the updated parameter, η\eta is the learning rate, and LL is the loss function.

[0230] 1.5.2 Concept Drift Detection

[0231] Use KL divergence to determine data distribution changes:

[0232]

[0233] where P(i)P(i) and Q(i)Q(i) are the distributions of current and historical data respectively.

[0234] 1.5.3 Feedback Learning Mechanism

[0235] Train a new model with maintenance data, such as Figure 2 As shown:

[0236] Train(M,X new )→M′,

[0237] Where MM is the original model, M′M' is the updated model, and XnewX_{new} is the new dataset.

[0238] In summary, the present invention has built a set of efficient and intelligent equipment health prediction and maintenance system through the innovative design of multi-source data fusion technology and intelligent prediction algorithm. It not only solves the core pain points in the current operation and maintenance of new energy equipment, but also provides a feasible path for the full life cycle management of equipment, with significant economic value and industry promotion potential.

[0239] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0240] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0241] In another embodiment provided in the present application, a computer program product comprising instructions is also provided, which, when executed on a computer, enables the computer to execute any of the new energy equipment health prediction and intelligent maintenance methods based on multi-source data fusion in the above-mentioned embodiments.

[0242] It is understandable that the system, device and storage medium provided in the embodiments of the present invention correspond to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above methods.

[0243] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0244] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0245] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0246] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion, characterized in that: include, Multi-dimensional data fusion: The system combines equipment operating parameters and environmental variables to build a dynamic health status assessment model; Health status classification: real-time monitoring of equipment operating status and grading of equipment health; Data-based fault prediction uses deep learning algorithms to model time series data, and predicts the time and type of possible equipment failures by capturing potential trends and change patterns in equipment operation. Provide dynamic prediction function, update prediction results in time, and ensure the synchronization between the model and the actual equipment status; Generate an optimized maintenance plan based on the health diagnosis and prediction results, including recommended inspection time, required spare parts and maintenance operation guidelines.

2. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 1 is characterized in that: It also includes the use of concept drift detection technology to monitor changes in device data distribution and dynamically adjust model parameters to adapt to environmental changes.

3. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 1 is characterized in that: Construct a dynamic health status assessment model, including, Data acquisition and preprocessing module, health status assessment module, fault prediction and early warning module, intelligent maintenance decision module and dynamic optimization and learning module.

4. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 3 is characterized in that: The steps of data acquisition and preprocessing module include: The system collects equipment operation data and environmental data through the sensor network deployed on the equipment; X run = {x v , x i , x t , x p , x a}, X env = {x irr , x wind , x hum , x press} where: XrunX_{run}: equipment operation data set, including voltage (xvx_{v}), current (xix_{i}), temperature (xtx_{t}), power (xpx_{p}) and vibration (xax_{a}); XenvX_{env}: Environmental dataset, including light intensity (xirrx_{irr}), wind speed (xwindx_{wind}), humidity (xhumx_{hum}), and air pressure (xpressx_{press}); Data cleaning and preprocessing: Missing value completion includes missing data xmissx _ {miss} uses the mean filling method: Or the interpolation algorithm: Normalization processing: Use Min-Max normalization to map the data to [0, 1][0, 1]: Use Kalman filter to eliminate noise: x k =x k-1 +K k (z k -Hx k-1 ), Where KkK_k is the Kalman gain, zkz_k is the measurement value, and HH is the measurement matrix. Generate time series samples using a sliding window: X window ={x t ,x t+1 ,...,x t+W }, Where WW is the window length; Data integration and storage: The edge computing node performs the following data storage operations locally: X local =Process(X raw ), Among them, XlocalX_{local} is the locally processed data, and XrawX_{raw} is the original collected data; the processing results are regularly uploaded to the cloud for in-depth analysis and backup.

5. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 3 is characterized in that: The health status assessment module processing steps include: Multi-dimensional feature fusion: Health feature vector F{health} extracted through feature engineering: F heath =φ(X run ,X env ), in is the feature mapping function; Real-time status diagnosis: Use classification models to evaluate the health status of equipment in real time; The mathematical expression of status diagnosis is: Where P(C|Fhealth)P(ClF_{health}) is the probability of state CC, i.e., normal, subhealthy, and faulty, and fc(·)f_c(\cdot) is the score function of the classifier; Dynamic visualization of health status: I health =Plot(F health ,t), Where lhealthl_{health} is a chart showing the changes in health index over time.

6. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 5 is characterized in that: The processing steps of the fault prediction and warning module include: Prediction model construction: Use LSTM model to predict the time T{fault} when the fault occurs: h t =σ(W h ·h t-1 +W x ·x t +b h ),y t =W y ·h t +b y , Where hth_t is the hidden state, yty_t is the predicted output, WhW_h, WxW_x, WyW_y, bhb_h, byb_y are parameters; Fault type identification: Fault classification algorithm based on decision tree: Where P(C|Xfault)P(C|X_{fault}) is the category probability of the fault feature vector XfaultX_{fault}; Threshold warning mechanism: Dynamic threshold setting, Where θ\theta is the dynamic threshold.

7. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 6 is characterized in that: The processing steps of the intelligent maintenance decision module include: Maintenance plan generation: Optimization goal: Minimize the total maintenance cost CtotalC_{total}: Where Ccheck,iC_{check,i} and Crepair,iC_{repair,i} are the inspection and repair costs respectively; Multi-device collaborative optimization: Establish scheduling optimization problems, Where Tdown,iT_{down,i} is the downtime of device ii.

8. The method for health prediction and intelligent maintenance of new energy equipment based on multi-source data fusion according to claim 6 is characterized in that: The dynamic optimization and learning module processing steps include: Model adaptive update: Update model parameters using online learning algorithms: Where θt\theta_t is the updated parameter, η\eta is the learning rate, and LL is the loss function; Concept Drift Detection: Use KL divergence to determine data distribution changes: Where P(i)P(i) and Q(i)Q(i) are the distributions of current and historical data respectively; Feedback learning mechanism: Use maintenance data to train new models, Train(M,X new )→M′, Where MM is the original model, M′M' is the updated model, and XnewX_{new} is the new dataset.

Images