Fault detection and root cause analysis system for multi-source energy system

Abstract

The invention discloses a fault detection and root cause analysis system for a multi-source energy system, and relates to the technical field of energy management and fault diagnosis. The system obtains various operation data and image data of the multi-source energy system in real time; preprocessing the operation data and the image data; inputting the preprocessed data into a hybrid algorithm combining wavelet transform, CNN and LSTM, and detecting an abnormal mode and a potential fault in the multi-source energy system; the digital twinborn model is dynamically updated according to data acquired in real time; and in combination with the feature extraction capability of a digital twin model and a hybrid algorithm, positioning key causes causing the fault, simulating a system operation state, analyzing potential causes of the fault, and outputting an analysis result. According to the invention, the comprehensive analysis capability of the system on various data sources is effectively improved, and a comprehensive and accurate solution is provided for intelligent operation and maintenance of the energy system.

Description

Technical Field

[0001] The present invention relates to the technical field of energy management and fault diagnosis, and in particular to a multi-source energy system fault detection and root cause analysis system. Background Art

[0002] With the continuous optimization of industrial fields, especially renewable energy, energy storage equipment and traditional power generation technology, factories and other industrial sites are increasingly relying on the coordinated work of multiple energy sources such as solar power generation, grid power supply, lithium battery energy storage equipment and traditional power generation to improve energy efficiency, reduce costs and achieve green production. Intelligent and efficient management and fault prediction of multi-source energy systems are of great significance in ensuring system stability and improving the intelligence of energy management. However, there are still some problems and challenges in the current fault detection technology and diagnostic methods in this field.

[0003] 1. Current status of fault detection technology for lithium battery energy storage systems

[0004] At present, fault detection of lithium battery energy storage systems mainly relies on traditional hardware monitoring methods, such as temperature, voltage, current and other sensors, and models based on data analysis, such as statistical analysis, time series analysis, etc. For example, through real-time monitoring of lithium battery voltage, current, temperature and other data, traditional fault detection methods use threshold judgment or deviation detection to identify abnormal conditions of equipment. However, this approach has many problems.

[0005] Problems with existing technologies:

[0006] The data is single and cannot fully reflect the system operation status

[0007] Most existing systems only focus on the physical parameters of lithium batteries, such as voltage and current, as a basis for judging faults, but ignore many important factors. Environmental factors such as temperature and humidity have a significant impact on the performance of lithium batteries. The aging and wear of the equipment appearance may indicate potential internal problems, and there is a lack of monitoring of external factors such as grid load and power fluctuations. For example, in a high temperature environment, the performance of lithium batteries may decline, but it is difficult to detect only by relying on voltage and current data. For example, grid load fluctuations may cause lithium batteries to charge and discharge frequently, accelerating battery aging, but the existing system cannot effectively capture this related information, which may miss the opportunity to discover some hidden faults early.

[0008] Lack of dynamic model update mechanism

[0009] Traditional fault detection methods are usually based on fixed models and cannot adapt to system state changes and dynamic fluctuations in the external environment in real time. Taking environmental changes as an example, changes in temperature and humidity, load fluctuations, and load fluctuations in the power grid will affect the performance and working state of lithium batteries. However, the existing monitoring model lacks an effective dynamic adjustment and optimization mechanism and cannot be updated in time according to actual conditions, resulting in a significant reduction in the accuracy and timeliness of fault detection.

[0010] Insufficient fault location capability

[0011] Existing technologies can generally only provide preliminary fault detection results, and it is difficult to accurately identify the root cause of the fault. In particular, when facing the impact of external grid load fluctuations on lithium battery equipment, it is impossible to accurately locate the fault point. This limitation seriously limits the accuracy and efficiency of equipment fault diagnosis, making subsequent system optimization and maintenance work lack a clear direction and specific basis, and unable to effectively solve the fundamental problem.

[0012] 2. Application of digital twin technology in multi-source energy systems

[0013] As an innovative monitoring and analysis method, digital twin technology has been widely used in many fields, especially in industrial manufacturing, energy management and other fields. It builds a virtual model that corresponds to the actual system in a virtual environment to achieve real-time monitoring and prediction of the system status, thereby providing strong support for the optimization and adjustment of the actual system. In a multi-source energy system, digital twin technology can simulate the operating status of the energy system and update and correct the virtual model based on real-time data, providing important support for fault detection and prediction. However, the current application of digital twin technology in multi-source energy systems still faces some challenges.

[0014] Problems with existing digital twin technology:

[0015] Although digital twins can monitor systems through virtual models, most existing digital twin systems lack an automated update mechanism and are unable to update virtual models in real time based on multimodal data, such as sensor data, image data, grid load data, etc. This means that the model often lags behind actual system state changes and cannot reflect system anomalies and failures in a timely manner, thus affecting the timeliness and accuracy of fault detection.

[0016] Weak data fusion capability

[0017] Traditional digital twin technology often lacks effective fusion strategies when processing data from different data sources, such as batteries, images, environment, and grid connection. In multi-source energy systems, there is a large amount of data from lithium battery management systems, various sensors, cameras, traditional power generation equipment, and power grids. The effective fusion of these data is crucial for accurate fault detection. However, existing digital twin systems usually cannot fully integrate these different types of data, resulting in insufficient utilization of data resources and failure to give full play to the advantages of digital twin technology in fault detection.

[0018] 3. Application of multimodal data fusion technology

[0019] Multimodal data fusion technology aims to extract information from multiple different types of energy and obtain a more comprehensive and accurate understanding of the system status through comprehensive analysis. In a multi-source energy system, in addition to collecting parameters such as voltage, current, and temperature of lithium batteries through sensors, image recognition technology can also be used to obtain information on changes in the appearance of the equipment, thereby achieving a more comprehensive monitoring of the operating status of energy equipment. By fusing these different types of data, it is expected to improve the accuracy and diagnostic capabilities of fault detection. However, there are still some shortcomings in the application of this technology.

[0020] Problems with existing multimodal data fusion technology:

[0021] Fusion algorithm is not mature enough

[0022] Existing multimodal data fusion technologies have not been widely used in the field of lithium battery energy storage systems, and most of them focus on basic feature extraction and fusion algorithms, such as weighted average, decision tree, etc. There are still technical bottlenecks for deep fusion and automated analysis of high-dimensional and complex data, including power grid load data. Especially in energy systems, the data types are diverse and the data volume is huge, which requires more complex and efficient algorithms to process in order to fully explore the potential information in the data.

[0023] Insufficient feature extraction and pattern recognition capabilities

[0024] Although traditional multimodal data fusion technology can process different types of data, it usually lacks sufficiently powerful feature extraction and pattern recognition capabilities. The operating status of multi-source energy systems is complex, and simple data fusion-based methods often find it difficult to extract the deep characteristics of system failures, especially under the influence of external factors such as grid load fluctuations. It is difficult to achieve accurate fault prediction and root cause analysis, thus affecting the reliability and stability of the system.

[0025] 4. Application of hybrid algorithms

[0026] In order to overcome the above problems, researchers have gradually begun to use more advanced algorithms, such as CNN and LSTM in deep learning algorithms to analyze and fuse multi-source data. CNN is a convolutional neural network, and LSTM is a long short-term memory network. These algorithms have powerful feature extraction capabilities when processing time series data and image data, and can more effectively mine useful information from a variety of data. However, in practical applications, hybrid algorithms still have some problems.

[0027] Application Problems of Existing Hybrid Algorithms

[0028] Over-reliance on feature extraction from a single data source Although algorithms such as CNN and LSTM have achieved certain success in their respective fields, in practical applications, many studies rely too much on a single data source, such as using only sensor data or image data, and ignore the potential of data fusion. Especially in multi-source energy systems, the occurrence of faults may be caused by a variety of factors. A single data source often cannot accurately describe the system status, especially under the influence of grid-connected data. Considering only a single data source will result in an inability to fully consider the impact of external factors on the system, thereby reducing the accuracy of fault detection.

[0029] Lack of integration and automated update mechanisms

[0030] Most existing hybrid algorithms focus on processing specific data sources, but lack the integration and automatic update mechanism of multimodal data. Especially in fault detection, the dynamic adaptability of hybrid algorithms is weak and cannot respond to changes in the system operation status in a timely manner. Especially under the influence of factors such as grid load fluctuations, the system state changes will be more complex, and existing hybrid algorithms cannot adapt quickly. Stronger dynamic adaptability is needed to ensure the stable operation of the system and the effectiveness of fault detection.

[0031] By analyzing the background of the prior art, we can clearly see that the current fault detection technology in multi-source energy systems has not yet achieved efficient multi-modal data fusion, dynamic digital twin model update, and accurate root cause analysis. Therefore, the multi-source energy system fault detection and root cause analysis method based on a hybrid algorithm of multi-modal data fusion and digital twin update proposed in the present invention is an innovative solution to the above technical bottlenecks. Summary of the invention

[0032] The purpose of the present invention is to propose a multi-source energy system fault detection and root cause analysis system. The multimodal data fusion of the present invention extracts frequency domain features through wavelet transform, extracts spatial features through convolutional neural network, and predicts time series features through long short-term memory network, so as to detect abnormal modes and potential faults in multi-source energy systems, and further judge the occurrence of faults and root cause analysis by combining digital twin technology, effectively improving the system's comprehensive analysis capabilities of multiple data sources, and solving the problem of accurate root cause analysis of multimodal data fusion in background technology.

[0033] The technical solution adopted by the present invention is as follows:

[0034] The present invention is a multi-source energy system fault detection and root cause analysis system, comprising a data acquisition module, a data preprocessing module, a data feature extraction and dimension reduction module, a digital twin module and a fault detection and root cause analysis module, wherein:

[0035] The data acquisition module acquires various operation data and image data of the multi-source energy system in real time and uploads them to the data preprocessing module;

[0036] The data preprocessing module preprocesses the various operating data and image data collected by the data acquisition module, including outlier detection, missing value filling, noise reduction and data standardization;

[0037] The data feature extraction and dimensionality reduction module extracts features and reduces the dimensionality of various pre-processed operation data and image data, and inputs the pre-processed operation data and image data into a hybrid algorithm combining wavelet transform, CNN and LSTM, wherein wavelet transform and CNN extract features of the pre-processed data, fuse the extracted features to obtain a fused feature set, perform dimensionality reduction processing on the high-dimensional data in the fused feature set, and combine LSTM for time series modeling to detect abnormal patterns and potential faults in multi-source energy systems;

[0038] The digital twin module builds a virtual model of the multi-source energy system based on the initial data, namely the digital twin model. The digital twin model is dynamically updated according to the real-time collected data, maintains consistency with the physical equipment in the multi-source energy system, reflects the status of the physical equipment, and provides a more accurate basis for fault detection and root cause analysis;

[0039] In the fault detection and root cause analysis module, a hybrid algorithm analyzes the real-time collected data, detects abnormal patterns and potential faults in the multi-source energy system, compares the prediction results of the real-time data input digital twin model, determines whether there is a deviation, and further confirms the occurrence of the fault; combines the feature extraction capabilities of the digital twin model and the hybrid algorithm to locate the key causes of the fault, simulate the system operation status, analyze the potential causes of the fault, and output the analysis results.

[0040] Furthermore, the various operating data of the multi-source energy system include photovoltaic power generation data, energy storage equipment data, traditional power generation equipment data, power load data, environmental data and grid-connected data. The various operating data of the multi-source energy system are acquired through various sensors installed in the multi-source energy system; the image data of the multi-source energy system, i.e., the appearance changes of energy equipment, are collected through a camera and transmitted to a data preprocessing module in real time.

[0041] Furthermore, the data preprocessing module preprocesses the various operation data and image data collected by the data collection module, specifically:

[0042] Outlier detection, removing obviously abnormal data through statistical methods or threshold detection;

[0043] Missing value filling: use interpolation, mean method or machine learning prediction method to fill in missing values;

[0044] Denoising: The collected operating data are treated as time series data, and de-noised using wavelet transform to extract frequency domain features related to faults. The image data is processed by CNN to extract spatial features and remove irrelevant background noise. For grid-connected data, filtering is first performed, and then wavelet transform is used for de-noising to extract frequency domain features related to faults to ensure data quality.

[0045] Data standardization: normalize data from different sources to ensure consistency in data format.

[0046] Furthermore, the data feature extraction and dimension reduction module uses wavelet transform and CNN to extract features from the preprocessed data, specifically:

[0047] The voltage data and current data in the energy storage device data are non-stationary signals. Wavelet transform is used to decompose the non-stationary signals, extract frequency domain features of different frequency bands, and reveal potential fault modes.

[0048] The image data uses CNN to extract key spatial features;

[0049] The traditional power generation equipment data is used to extract fuel consumption trends through data fitting and difference method or time series decomposition method, and power output fluctuation characteristics are extracted through wavelet transform method or rolling standard deviation method;

[0050] The grid-connected data uses wavelet transform to extract frequency domain features, analyzes the spectrum components and energy distribution in the frequency domain, and is used to determine the matching between the grid load state and the energy supply;

[0051] The feature extraction of data feature extraction and dimension reduction module also includes:

[0052] The photovoltaic power generation data extracts the characteristics of power generation stability, the characteristics of the correlation between light intensity and power generation, and the characteristics of the impact of temperature on power generation efficiency;

[0053] The environmental data extracts humidity change trend characteristics, wind speed change trend characteristics and air pressure change trend characteristics, and is obtained by determining the change of their mean and variance over time through a time series analysis method.

[0054] Furthermore, the preprocessed data features are extracted and fused to obtain a fused feature set, specifically:

[0055] The grid-connected data is integrated with photovoltaic power generation data, energy storage equipment data, and traditional power generation equipment data, and features extracted from different data sources are weighted and integrated through a multi-head attention mechanism to provide more comprehensive energy supply information;

[0056] The environmental data is combined with features extracted from power load data, energy storage equipment data, traditional power generation equipment data, and photovoltaic power generation data to provide more dimensional information for fault prediction and system scheduling.

[0057] Furthermore, the dimensionality reduction process of the high-dimensional data in the fusion feature set is specifically as follows:

[0058] The principal component analysis method is used to reduce the dimension of the high-dimensional data of the fusion feature set, retain key features, and reduce redundant data. The fusion feature set includes:

[0059] The characteristics of power generation stability, the correlation between light intensity and power generation, and the influence of temperature on power generation efficiency extracted from photovoltaic power generation data;

[0060] Voltage frequency domain characteristics and current frequency domain characteristics of energy storage equipment data extraction;

[0061] Fuel consumption trends and power output fluctuation characteristics extracted from traditional power generation equipment data;

[0062] Spatial features extracted from image data;

[0063] Frequency domain characteristics of grid stability and load variation extracted from grid-connected data;

[0064] Environmental data extracts humidity change trend characteristics, wind speed change trend characteristics and air pressure change trend characteristics;

[0065] The principal component analysis method calculates the covariance matrix of the data, finds the eigenvalues ​​and eigenvectors, selects the main eigenvectors according to the size of the eigenvalues, and thus projects the high-dimensional data into a low-dimensional space, retains the key features, and combines them with the spatial features extracted from the image data to form a unified low-dimensional feature set, thereby simplifying the subsequent analysis process.

[0066] Furthermore, the LSTM models time series data based on a low-dimensional feature set, captures long-term dependencies of system operating states, and predicts potential failures and abnormal behaviors.

[0067] Furthermore, the digital twin model is dynamically updated according to the real-time collected data, and the specific steps are as follows:

[0068] Trigger mechanism: When the data acquisition module collects image data or one or more of the operating data changes beyond the normal range, the digital twin model is automatically updated;

[0069] Incremental learning: using incremental learning methods to dynamically update digital twin model parameters, simulate the operating status of multi-source energy systems, and ensure that the virtual model can timely reflect the operating status of equipment in the actual multi-source energy system;

[0070] Real-time adjustment: The digital twin model provides real-time warning of system failures and makes adjustments based on updated data.

[0071] Furthermore, the results output by the fault detection and root cause analysis module are specifically shown as follows:

[0072] Fault warning information: through real-time monitoring and analysis, fault warnings are issued to users or management systems;

[0073] Root cause analysis report, providing a detailed analysis report listing the fault type, possible causes and recommended maintenance measures;

[0074] Data visualization displays system operating status, fault prediction and diagnosis results in chart form to facilitate user understanding and decision-making.

[0075] The present invention is applicable to multi-source energy management systems in factories or other industrial scenarios, especially for the coordinated optimization and management of solar power generation, grid power supply, energy storage equipment and traditional standby power generation equipment. By adopting multimodal data fusion and hybrid algorithms, and combining digital twin technology, the present invention can monitor and analyze the operating status of multi-source energy systems in real time, and realize instant detection, prediction and root cause analysis of faults. Digital twin technology reflects the working status of equipment in real time by constructing virtual models, while hybrid algorithms improve the accuracy and efficiency of fault diagnosis by deeply mining multimodal data. The present invention provides a comprehensive and accurate solution for the intelligent operation and maintenance of multi-source energy systems, significantly improving the reliability, operating efficiency and emergency response capabilities of source energy systems.

[0076] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0077] The present invention is a multi-source energy system fault detection and root cause analysis system, which uses a camera to collect image data of energy storage devices in the multi-source energy system in real time, and detects abnormal conditions such as cracks, damage, stains, etc. that may exist on the surface of the energy storage device. The image data provides direct visual evidence for appearance faults, and can identify external fault sources in a timely manner, especially providing strong support for the detection of external faults of energy storage devices. The image data is combined with the sensor data, and through multimodal data fusion, the accuracy of equipment fault diagnosis is enhanced, and the root cause of the fault can be more comprehensively identified and analyzed. In addition, the fusion of grid-connected data and power load data with image data further enhances the fault prediction capability of the multi-source energy system in a grid-connected environment.

[0078] The present invention is a multi-source energy system fault detection and root cause analysis system, which adopts a hybrid algorithm combining wavelet transform, CNN and LSTM to solve the limitations of traditional methods in fault diagnosis. Among them, the signal is denoised and decomposed by wavelet transform, combined with the advantages of CNN in local feature extraction, and time series data is modeled by LSTM. The combination of the three effectively improves the accuracy and real-time performance of fault detection, and can cope with the complex and changeable operating environment of multi-source energy systems, especially adapting to the dynamic characteristics of multi-source energy systems. In this process, grid-connected data, such as grid load fluctuations, electricity price fluctuations, etc., can also effectively improve the accuracy of prediction and the response speed of the system.

[0079] The present invention is a multi-source energy system fault detection and root cause analysis system. The hybrid algorithm of the present invention combines wavelet transform, convolutional neural network and long short-term memory network, giving full play to their respective advantages to achieve high-precision and high-efficiency detection of multi-source energy system faults. The algorithm integrates traditional signal processing and deep learning technology, can fully mine potential information in multimodal data, and provide more comprehensive fault detection and root cause analysis capabilities. Through the frequency domain characteristic analysis of wavelet transform, CNN extracts image features, and LSTM models time series data, the three work together to significantly improve the accuracy of fault diagnosis and enhance the robustness of the system in complex and dynamic environments, so that it can maintain efficient operation under changing operating conditions and respond to faults in real time.

[0080] The present invention is a multi-source energy system fault detection and root cause analysis system. The present invention uses digital twin technology to create a virtual model of the equipment, and simulates the status and behavior of the equipment through real-time updating and optimization. The digital twin model can quickly adapt to equipment changes based on actual operating data, and perform early warning and diagnosis before or when a fault occurs. Through incremental learning and online update mechanisms, the digital twin model maintains consistency with the physical equipment and adjusts in real time, which enhances the accuracy of fault detection, significantly improves the reliability of the equipment, reduces the risk of failure, and optimizes the system operation and maintenance efficiency. Under grid-connected conditions, the digital twin model can simultaneously consider the interaction effects between energy storage equipment and the power grid, and reflect the overall operating status of the system in real time.

[0081] The present invention is a multi-source energy system fault detection and root cause analysis system. By integrating multiple data sources and combining digital twin technology, a digital twin model of the multi-source energy system is created, and the digital twin model is updated and optimized in real time to achieve dynamic monitoring of equipment status and fault detection. The present invention can provide comprehensive and accurate fault detection and root cause analysis, enhance the reliability and real-time response capability of the system, especially in complex and dynamic environments, and can reflect changes in equipment status in real time and provide accurate fault warnings. Grid-connected data provides the system with real-time feedback of the external environment, further enhancing fault detection and scheduling optimization capabilities.

[0082] The present invention is a multi-source energy system fault detection and root cause analysis system that can detect and diagnose equipment faults in real time, quickly generate fault warnings, and provide detailed root cause analysis reports. The report is presented through a graphical interface to help operation and maintenance personnel quickly identify problems and take effective measures. Through this function, the system can significantly reduce equipment downtime, optimize system operating efficiency, and improve the reliability of the overall system. The generation of fault warnings and analysis reports can provide accurate decision support, help operation and maintenance personnel take maintenance actions in a timely manner, and avoid system downtime. The introduction of grid-connected data helps to better evaluate the impact of grid load fluctuations on equipment and provide timely maintenance recommendations.

[0083] The present invention is a multi-source energy system fault detection and root cause analysis system, which has strong adaptability and scalability, can be widely used in multi-source energy systems of different scales and types, can be flexibly adjusted according to different needs, supports fault detection and analysis in different devices and environments, and supports the expansion and upgrading of future equipment, ensuring that the system has greater application potential in future development. The scalability of the system is not only reflected in supporting multiple types of energy equipment, but also in integrating power grid data to provide management and optimization solutions for larger-scale energy systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] 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 embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative work, among which:

[0085] Figure 1 is a system block diagram of the present invention;

[0086] Figure 2 It is a data processing flow chart of the feature extraction and dimension reduction module in the present invention;

[0087] Figure 3 This is a data processing flow chart of Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0088] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0089] It should be noted that the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, product or equipment that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, product or equipment.

[0090] Embodiment 1

[0091] like Figure 1 As shown, the present invention proposes a multi-source energy system fault detection and root cause analysis system, the system of the present invention includes a data acquisition module, a data preprocessing module, a feature extraction and dimension reduction module, a digital twin module, and a fault detection and root cause analysis module. The details are as follows:

[0092] Data acquisition module

[0093] Through various sensors installed in multi-source energy equipment and its environment, various operating data of the system are collected in real time, including voltage, current, temperature, pressure, humidity, wind speed, etc. The camera is used to obtain the appearance data of the energy storage equipment, such as whether there are cracks, stains, damage, etc. on the surface of the equipment, so as to further analyze the equipment status. It also includes collecting the operating data of the power grid status, traditional power generation equipment and photovoltaic power generation system to provide comprehensive energy monitoring and system management.

[0094] The sensors of this embodiment include but are not limited to voltage sensors with a measurement accuracy of ±0.5mV, current sensors with an accuracy of ±0.1mA, sensors with an accuracy of ±0.5℃, etc., which can accurately obtain multi-dimensional operating data including voltage, current, temperature, pressure, humidity, wind speed, etc. At the same time, a high-resolution 1080p clear camera is used to monitor subtle changes in the appearance of real-time energy storage equipment, such as whether there are cracks, stains, damage, etc. on the surface of the equipment, providing intuitive visual information for a comprehensive assessment of the equipment status. In addition, special power monitoring equipment is used to collect grid status data, as well as operating data of traditional power generation equipment and photovoltaic power generation systems, for comprehensive energy monitoring and system management. The specific data collected are:

[0095] Photovoltaic power generation data: power generation, light intensity, temperature, etc., used to evaluate photovoltaic power supply capacity;

[0096] Energy storage device data: voltage, current, temperature, health status SOH, number of charge and discharge cycles, etc.; the energy storage device in this embodiment is a lithium battery, and the energy storage device data can monitor the battery status in real time;

[0097] Traditional power generation equipment data: operating parameters such as fuel consumption rate, power generation, and startup frequency, used to optimize power generation efficiency;

[0098] Power load data: electricity price fluctuations, load demand data, and other sensor data to participate in fault detection and prediction;

[0099] Image data: The camera collects changes in the appearance of energy equipment, such as cracks, damage, stains, etc., to provide visual information on the status of the equipment;

[0100] Environmental data: humidity, wind speed, air pressure, etc., for fault prediction and optimized scheduling;

[0101] Grid-connected data: grid voltage, current, power, latitude, longitude, etc., used to monitor grid power supply stability and grid load status, affecting switching decisions at energy sources;

[0102] Use wireless sensor network WSN or wired communication to transmit the collected multimodal data to the data preprocessing module to ensure efficient data transmission and real-time performance. In this embodiment, the ZigBee communication protocol is used with a frequency band of 2.4 GHz, a transmission rate of 250 kbps, and a transmission power of 20 dBm for wireless sensor network data transmission.

[0103] Data preprocessing module

[0104] Preprocess the collected multimodal data, including outlier detection, missing value filling, noise reduction and data normalization, to ensure the quality and availability of the data and provide a reliable data basis for subsequent analysis.

[0105] Outlier detection: For each piece of collected operating data and image data, outlier detection is first performed, and obvious abnormal data is removed through statistical methods or threshold detection. For example, for the voltage data of energy storage equipment, the normal range is determined based on historical data or equipment specifications. If the voltage value at a certain moment exceeds the normal range, it is determined as an abnormal value and removed.

[0106] Missing value filling: For missing values ​​in the data set, a variety of flexible and effective filling methods are used. For time series data, if the data at a certain moment is missing, linear interpolation, spline interpolation and other methods can be used to fill it according to the changing trend of the data before and after the moment to maintain the continuity of the data. At the same time, for some complex data, machine learning prediction methods can be used, such as prediction and filling based on support vector machine SVM regression, neural network and other algorithms. When using machine learning methods, historical data similar to the time period of missing data is carefully selected as the training set, the data is strictly cleaned and normalized, and the model is optimized through evaluation indicators such as root mean square error RMSE and determination coefficient R² to ensure the accuracy and reliability of the filled value.

[0107] Noise reduction: The collected operating data are treated as time series data, and wavelet transform is used for noise reduction to effectively extract the frequency domain features related to the fault. For image data, the powerful image processing capabilities of CNN are used to extract key spatial features, remove irrelevant background noise, and highlight the key information in the image. For grid-connected data, an efficient filtering algorithm is first used for preliminary filtering to remove interference such as high-frequency noise, and then wavelet transform is used for further noise reduction to ensure that accurate frequency domain features are extracted, thereby improving data quality and providing clear and reliable data signals for subsequent analysis.

[0108] In the embodiment, the wavelet function Daubechies is selected, and the Db4 wavelet in the wavelet series has good time-frequency resolution according to the frequency characteristics of the signal and the fault characteristic frequency. The number of decomposition layers is determined according to the complexity and noise level of the signal. The empirical formula determines that it is 5 layers. The soft threshold method is used for noise reduction processing, so as to accurately extract the frequency domain characteristics of different frequency bands and effectively reveal potential fault modes.

[0109] For image data, the powerful image processing capabilities of CNN are used to extract key spatial features, remove irrelevant background noise, and highlight the key information in the image. The CNN structure includes 3 convolutional layers, with a convolution kernel size of 3×3 and a step size of 1, 2 pooling layers, and 2 fully connected layers. The number of neurons is 128 and 64 respectively. The activation function uses the ReLu function to extract spatial features in the image, remove background noise, and identify damage or abnormalities in the appearance of energy storage equipment, such as subtle cracks and stains.

[0110] For grid-connected data, we first use an efficient filtering algorithm, such as the Butterworth filter, with a filter order of 4 and a cutoff frequency of 50 Hz, to perform preliminary filtering to remove interference such as high-frequency noise, and then use wavelet transform to further reduce noise to ensure that accurate frequency domain features are extracted, thereby improving data quality and providing clear and reliable data signals for subsequent analysis.

[0111] Data standardization: To ensure the format consistency and comparability of data from different sources, advanced data cleaning and normalization algorithms are used, such as the Z-score standardization method, to convert the data collected by each sensor into a standard normal distribution with a mean of 0 and a standard deviation of 1, eliminating the differences in dimensions and orders of magnitude between data, creating favorable conditions for subsequent data fusion and analysis.

[0112] Data feature extraction and dimension reduction module

[0113] This module uses the wavelet transform + CNN + LSTM hybrid algorithm to conduct in-depth data analysis and fault diagnosis.

[0114] Wavelet transform is a commonly used signal processing method that can effectively extract local features from signals, and is particularly suitable for the analysis of non-stationary signals. Wavelet transform plays an important role in fault detection, especially in multi-source energy systems, where it can process time series signals such as battery voltage and current, remove noise, and enhance the appearance of fault features. Therefore, wavelet transform is often combined with other deep learning algorithms such as CNN and LSTM to enhance the accuracy of fault detection and prediction.

[0115] The combination of wavelet transform, CNN and LSTM can effectively process and analyze the time series signals and image data of multi-source energy systems. This hybrid algorithm not only improves the feature extraction capability, but also enables efficient fault prediction and root cause analysis. The wavelet transform is used to denoise and decompose the signal, extract the local features of the signal, and then use CNN for feature learning. Finally, LSTM is used to model and predict the time series data, thereby greatly improving the accuracy and real-time performance of fault detection.

[0116] Wavelet transform: used to process time series data from sensors, extract frequency domain features, and perform noise reduction to obtain characteristic information of faults.

[0117] Specifically, wavelet transform processes time series data from sensors, and decomposes non-stationary signals by carefully selecting appropriate wavelet functions, such as Db4 wavelet in the Daubechies wavelet series, which has good time-frequency resolution based on the frequency characteristics of the signal and the fault characteristic frequency. Determine the reasonable number of decomposition layers, such as the specific number of layers determined by experiments or empirical formulas based on the complexity and noise level of the signal, and use soft threshold or hard threshold methods for processing, so as to accurately extract frequency domain features of different frequency bands and effectively reveal potential fault modes. The soft threshold formula is:

[0118] ,

[0119] in, For the Frequency band The wavelet coefficients of the position, Used to determine the position of the wavelet function in the time domain or space domain, different The values ​​represent different translation positions of the wavelet function in time or space. , is the threshold value, When the wavelet coefficient Will be based on Shrinkage treatment; when When the wavelet coefficient Directly set to 0, is the symbolic function, .

[0120] By deeply analyzing the energy distribution of different frequency bands, such as calculating the energy of each frequency band ,in, is the wavelet coefficient at position k in the jth frequency band, For the The energy in the frequency band can accurately determine whether there is a short circuit or plate aging fault inside the battery, such as a sudden increase in low-frequency band energy or a gradual decrease in high-frequency band energy.

[0121] In the present invention, the voltage data and current data in the energy storage device data are non-stationary signals. Wavelet transform is used to decompose the non-stationary signals, extract the frequency domain characteristics of different frequency bands, and reveal potential fault modes. By analyzing the energy distribution in different frequency bands, it can be determined whether there are faults such as short circuits or plate aging inside the battery.

[0122] Traditional power generation equipment data uses data fitting and difference method or time series decomposition method to extract fuel consumption trends, and uses wavelet transform method or rolling standard deviation method to extract power output fluctuation characteristics;

[0123] Fuel consumption trend feature extraction:

[0124] Data fitting and difference method: first collect the time series data of fuel consumption of traditional power generation equipment, perform polynomial fitting according to the time series, and then obtain the changing trend of fuel consumption by calculating the derivative or difference of the fitting function.

[0125] Time series decomposition method: decompose fuel consumption data into trend components, seasonal components, cyclical components and random components, focusing on the trend components after decomposition, such as using the Holt-Winters method to decompose the time series to separate the long-term trend of fuel consumption. The Holt-Winters method contains three smoothing equations: the horizontal equation ,

[0126] Trend equation ,

[0127] Seasonal equation ,

[0128] in, For in time The level value of For in time The level value of For in time The trend value of For in time The seasonal factor, For in time The actual observed value of , , is the smoothing parameter, For in time The seasonal factor, For seasonal cycles, the best decomposition results are obtained by optimizing these parameters, e.g. , , .

[0129] Power output fluctuation feature extraction:

[0130] Wavelet transform method: Use wavelet transform on the power output of power generation equipment, which may be non-stationary, to decompose it into sub-signals of different scales or frequencies, and analyze the energy changes and fluctuations of each sub-signal. In this embodiment, the feature extraction step selects Moret wavelet, and the wavelet coefficient formula is:

[0131] ,

[0132] in, is the wavelet coefficient, is the power output signal, which is about time The function of is the scale parameter used to scale the width of the wavelet function. is the translation parameter, which is used to move the wavelet function on the time axis. is the wavelet function, Represents the wavelet function after scale and translation transformation, by analyzing the modulus square of the wavelet coefficients Describe the fluctuation characteristics of power output over time and scale.

[0133] Rolling standard deviation method: Set a rolling window, such as a window size of 10, and calculate the standard deviation of the power output data in the window. As the window rolls, a series of standard deviation values ​​are obtained to describe the change in the fluctuation degree of power output over time.

[0134] The photovoltaic power generation data extracts the characteristics of power generation stability, the characteristics of the correlation between light intensity and power generation, and the characteristics of the impact of temperature on power generation efficiency;

[0135] Power generation stability feature extraction:

[0136] First, select a time period to examine the stability of the generated power. During this time period, record the power value of photovoltaic power generation at each moment. Then calculate the average level of these power values: ,in, is the average power value, For the The power generation at a given moment, is the number of data points in the time period. For example, if 240 data points are collected in one day, then , and then find the deviation of the power value from the average value , the formula is: By comprehensively calculating these deviations, such as calculating the variance or standard deviation, we can get a value that reflects the fluctuation of the power value. This value is the stability characteristic of the power generation. If this value is small, it means that the power generation is relatively stable during this period; otherwise, it means that the power fluctuation is large.

[0137] For example, by observing the power generation in the past day, calculating the average power of the day, and then looking at the difference between the power at each moment and the average power, and summing up these differences, we can know the stability of the power generation on that day, and then understand the reliability of the power output of the photovoltaic power generation system during that period. This is critical for grid access and energy scheduling. Stable power output helps maintain the balance and stable operation of the grid and avoid shocks to the grid due to large power fluctuations.

[0138] Feature extraction of correlation between light intensity and power generation:

[0139] Select a time range and record the light intensity at each moment during this period and power generation Calculate the average light intensity and the average power generation ,in, is the average light intensity, The average power generation capacity is is the number of data points. 10080 data points were collected in one week. , and then analyze the deviation of light intensity from the average value and the deviation of the generated power from the average value , and study the correlation between these two sets of deviations, so as to obtain a value that can reflect the close relationship between light intensity and power generation, that is, the correlation characteristic.

[0140] Calculate the correlation coefficient, such as using the Pearson correlation coefficient: ,

[0141] in, is the Pearson correlation coefficient, For the The deviation of the light intensity from the average value at each moment, For the The deviation of the power generated at any moment from the average value, is the number of data points;

[0142] If it is found that the power generation increases with the increase of light intensity, and the correlation coefficient is close to 1, it means that the two are highly correlated. On the contrary, if the power generation does not change significantly when the light intensity changes, and the correlation coefficient is close to 0, the correlation is weak. Understanding this correlation helps to predict the change of power generation in advance when the trend of light intensity change is known, so as to optimize energy configuration and improve energy utilization efficiency. For example, when the light intensity is about to increase, the energy storage equipment can be prepared in advance to receive more electricity.

[0143] For example, record relevant data for a week, calculate the average values ​​of light intensity and power generation, and then look at the corresponding relationship when the two deviate from the average value. If it is found that the power generation increases with the increase of light intensity, and this relationship is more obvious, it means that the two are highly correlated; conversely, if the power generation does not change significantly when the light intensity changes, the correlation is weak. Understanding this correlation can help predict the change in power generation in advance when the trend of light intensity changes is known, thereby optimizing energy configuration and improving energy utilization efficiency. For example, when the light intensity is about to increase, prepare the energy storage equipment in advance to receive more electricity.

[0144] Feature extraction of the impact of temperature on power generation efficiency:

[0145] Continuously track the temperature of photovoltaic power generation equipment and the power generation at the same time, and accumulate data for a certain period of time. Based on temperature, try to find the intrinsic relationship between it and power generation, and build a model that can describe the relationship between the two. Through the relevant parameters in this model, measure the impact of temperature on power generation efficiency.

[0146] For example, we can continuously collect temperature and power generation data within a month, observe how the power generation changes when the temperature changes, and build a suitable mathematical model to describe this change relationship. A certain coefficient in the model can be used as a characteristic of the impact of temperature on power generation efficiency. If this coefficient shows that the power generation decreases significantly when the temperature rises, we need to pay attention to the heat dissipation of the equipment, because this may lead to reduced power generation efficiency and affect the performance and power generation of the entire photovoltaic power generation system; conversely, if the coefficient is abnormal, we also need to further analyze whether the equipment is in the appropriate operating temperature range to ensure normal operation of the equipment and efficient power generation.

[0147] The environmental data is extracted by extracting humidity change trend characteristics, wind speed change trend characteristics and air pressure change trend characteristics, and the change of their mean and variance over time is determined by a time series analysis method;

[0148] Humidity change trend feature extraction:

[0149] Record the environmental humidity data at regular intervals, such as once every hour. Calculate the average humidity data over a longer period of time, such as a week or a month, and then observe the change of each humidity data compared to the average value, whether it is higher or lower than the average value, and the trend of this difference over time, whether it continues to rise, fall, or fluctuate. You can also further analyze the discreteness of the humidity data, that is, the size of the fluctuation.

[0150] For example, record the humidity every hour for one month in summer, calculate the average humidity of the month, and check the difference between the humidity and the average value at different times of the day. If it is found that the humidity is continuously higher than the average with small fluctuations on certain days, it may indicate that rainy weather is about to come, which is a reminder for moisture-proof protection of outdoor energy equipment. When the humidity fluctuates greatly, it may increase the risk of equipment getting damp and affect the electrical performance and life of the equipment. Therefore, it is necessary to take corresponding measures in advance according to the characteristics of the humidity change trend, such as strengthening equipment sealing or adding dehumidification equipment.

[0151] Wind speed change trend feature extraction:

[0152] Collect wind speed data at fixed time intervals. Calculate the average wind speed over a certain period of time, such as a day or a week, determine the maximum and minimum wind speeds during this period and the times they occur, and understand the extreme wind speeds. At the same time, analyze the distribution of wind speed data around the average value, that is, the degree of fluctuation.

[0153] For example, in an area with abundant wind resources, wind speed data is continuously recorded for a week, the average wind speed is calculated, and the time and value of the maximum and minimum wind speed are found. If the average wind speed is high and the fluctuation is small, it may mean relatively stable power generation conditions for wind power generation equipment, but if the maximum wind speed exceeds the equipment's tolerance limit, wind protection measures need to be taken to protect the equipment; and when the wind speed fluctuates greatly, it may affect the stability of wind power generation, and energy storage equipment and other means are needed to balance the power generation output and ensure the stability of energy supply. At the same time, it is also necessary to reinforce transmission lines to prevent line damage caused by strong winds.

[0154] Air pressure change trend feature extraction:

[0155] Record air pressure data regularly, observe the trajectory of air pressure data changes over time, and determine whether the air pressure is rising, falling, or remaining relatively stable. Calculate how fast the air pressure changes over a period of time, and the distribution of air pressure in different value ranges.

[0156] For example, if the air pressure is monitored for a long time near energy facilities in mountainous areas and it is found that the air pressure continues to drop, it may indicate that the weather is about to change, such as rain or storms, which may have an impact on solar power generation equipment because weather changes may cause changes in light intensity, which in turn affects power generation efficiency. Understanding the characteristics of air pressure changes can help predict the impact of weather on the operation of energy equipment in advance, so as to adjust the equipment operation strategy or take protective measures in time. For example, when air pressure changes indicate bad weather, the equipment can be reinforced or some outdoor maintenance work can be suspended to ensure the safe and stable operation of the energy system.

[0157] It is also possible to extract frequency features under extreme environmental conditions:

[0158] According to the operating requirements of energy equipment and local environmental characteristics, set the judgment criteria or thresholds for extreme environmental conditions such as high temperature, high humidity, and strong wind. For each recorded temperature, humidity, and wind speed data, compare it with the corresponding threshold. If it exceeds the threshold, record this type of extreme event. In a specific observation period, such as a quarter, count the number of occurrences of various extreme events, and then calculate the frequency of these extreme events.

[0159] For example, in the energy system in coastal areas, the high temperature threshold is set at 35°C, the high humidity threshold is set at 80%, and the strong wind threshold is set at level 10. Within a quarter, determine whether these extreme conditions are met each time data is recorded. If statistics show that the frequency of high temperature events is high, it may be necessary to optimize the heat dissipation of the equipment or increase cooling measures to prevent the equipment from being damaged by high temperature; when the frequency of high humidity is high, it is necessary to strengthen the moisture and rust prevention of the equipment; when the frequency of strong wind is high, it is necessary to strengthen the windproof design and reinforcement measures of the equipment to ensure the normal operation of energy equipment in extreme environments, reduce equipment failures and downtime caused by extreme environments, and improve the reliability and stability of the energy system.

[0160] CNN: Used to analyze spatial features in image data and identify damage or abnormalities in the appearance of energy storage devices, such as cracks, stains, etc.

[0161] Specifically, CNN uses its powerful convolutional layer and pooling layer structure to perform in-depth analysis of image data and extract key spatial features. By properly setting the convolution kernel size and selecting the appropriate activation function, it can effectively identify damage or abnormalities on the appearance of energy storage equipment, such as subtle cracks and stains, and provide an intuitive visual basis for equipment status assessment.

[0162] In this embodiment, the CNN structure includes 3 convolutional layers, the convolution kernel size is 3×3, the step size is 1, 2 pooling layers and 2 fully connected layers, the number of neurons in the fully connected layers are 128 and 64 respectively, and the activation function adopts the ReLu function.

[0163] Multimodal data fusion and dimensionality reduction: The features extracted from various operating data and the spatial features extracted by CNN are combined, and a unified low-dimensional feature set is formed after dimensionality reduction through PCA for further modeling and prediction by LSTM.

[0164] The frequency domain features extracted by wavelet transform and the spatial features extracted by CNN are combined. Specifically, in the process of data fusion, the frequency domain features extracted by wavelet transform use a multi-head attention mechanism to assign reasonable weights to different features according to the importance and relevance of the data. The number of attention heads is determined based on the number of key performance indicators of energy storage equipment in multi-source energy systems, such as energy storage capacity, charging and discharging efficiency, etc., the complexity of the interaction with other energy sources, and the accuracy requirements of the system for state monitoring of energy storage equipment. The weight of each attention head is determined by calculating indicators such as the dynamic correlation coefficient information entropy between features, and different initial weights are assigned according to the importance of energy storage equipment in the system. The fused feature set uses the PCA dimensionality reduction method to calculate the covariance matrix of the data, solve the eigenvalues ​​and eigenvectors, select the main eigenvectors according to the size of the eigenvalues, and project the high-dimensional data into a low-dimensional space to form a unified low-dimensional feature set, providing optimized data input for LSTM modeling.

[0165] The number of attention heads is determined based on the number of key performance indicators of energy storage equipment in the multi-source energy system, the complexity of the interaction with other energy sources, and the accuracy requirements of the system for monitoring the status of energy storage equipment. Specifically:

[0166] Energy storage devices have multiple key performance indicators in multi-source energy systems, such as energy storage capacity, charge and discharge efficiency, remaining power, health status SOH, etc. Each key performance indicator has an important impact on the operating status and fault detection of the system, so an attention head needs to be assigned to each indicator. For example, when considering the two key performance indicators of energy storage capacity and charge and discharge efficiency, at least 2 attention heads are required. If the system further considers indicators such as remaining power and health status, it may increase to 4 or more attention heads depending on actual needs and computing resources to ensure that it can fully focus on the impact of different aspects of the performance of energy storage equipment on system operation;

[0167] In a multi-source energy system, there is a complex interactive relationship between energy storage devices and energy sources such as photovoltaic power generation, traditional power generation equipment, and power grids. If the energy exchange between energy storage devices and other energy sources is frequent, for example, when photovoltaic power generation is unstable, the energy storage devices need to charge and discharge frequently to balance energy supply and demand. In this case, the interactive relationship is complex. At this time, more attention heads are needed to capture the characteristics of the interaction data between different energy sources. Assuming that in a simple multi-source energy system, the energy storage device only has a small amount of energy exchange with the power grid, only 2-3 attention heads may be needed to process the data interaction related to the power grid; in a complex system, the energy storage device has a close energy scheduling and collaborative working relationship with photovoltaic power generation, traditional power generation equipment and the power grid, then 4-6 attention heads may be needed to comprehensively analyze the interaction characteristics between the energy sources in order to accurately judge the operating status and potential failures of the system;

[0168] If the system has high requirements for the accuracy of energy storage device status monitoring, for example, when accurately evaluating the health status of energy storage equipment to prevent failures in advance, it is necessary to analyze various data features more carefully. At this time, in order to improve the monitoring accuracy, the number of attention heads will be increased. Taking the monitoring of the health status of energy storage equipment as an example, if you only need to have a general understanding of its health status, 2-3 attention heads may be enough to preliminarily analyze the relevant data; but if you want to accurately judge the tiny hidden faults inside the battery, such as the early stages of plate aging, 4-8 attention heads may be required. By paying attention to and analyzing more data features, the accuracy of energy storage device status monitoring is improved, so as to more accurately predict and diagnose faults. In the present invention, taking the above factors into consideration, according to the actual operation requirements of the multi-source energy system and the accuracy requirements for energy storage device status monitoring, the number of attention heads is determined to be 4, which can achieve a good balance between computing resources and monitoring accuracy, and meet the system's requirements for fault detection and root cause analysis.

[0169] LSTM: Use LSTM to model time series data, learn the timing characteristics of data, capture the long-term dependencies of system operation status, and perform fault prediction and abnormal behavior identification. LSTM uses its unique gating structure to reasonably set the number of hidden layers and learning rate, and combines external information such as grid voltage and load fluctuations to effectively improve the accuracy and real-time performance of fault prediction, identify potential faults and abnormal behaviors in advance, and provide reliable protection for the stable operation of the system.

[0170] The LSTM model structure of this embodiment includes an input layer, a hidden layer, and an output layer. The model performance is optimized by carefully adjusting the number of hidden layers, the learning rate, and other parameters, such as a learning rate of 0.001. During the training process, the model is iteratively trained using the loss function mean square error MSE and the Adam optimization algorithm, so that it can accurately learn the time series patterns in the data, predict potential faults and abnormal behaviors in advance, and provide reliable protection for the stable operation of the system.

[0171] like Figure 2 As shown in the figure, the data is processed by the data feature extraction and dimension reduction modules.

[0172] 1. Multimodal data input: The input data is divided into two parts: time series data, such as voltage and current from sensors, and image data, such as the appearance of the device from the camera.

[0173] 2. Wavelet transform: extract frequency domain features from time series data and perform noise reduction.

[0174] 3. CNN: Extracts spatial features from image data to identify damage to the device appearance, such as cracks, stains, etc.

[0175] 4. Multimodal data fusion: Combine the frequency domain features extracted by wavelet transform with the spatial features extracted by CNN to perform data dimensionality reduction and obtain data of low-dimensional feature sets.

[0176] 5. LSTM modeling: Based on the low-dimensional feature set data after dimensionality reduction, LSTM is used to model time series data and predict faults and abnormal behaviors.

[0177] 6. Result output: Finally, the fault prediction and abnormal diagnosis results are output through the hybrid algorithm.

[0178] Digital Twin Module

[0179] The digital twin model is a virtual model of a multi-source energy system built based on initial data, establishing a one-to-one correspondence between the physical system and the virtual model, and is used to monitor the operating status of the multi-source energy system in real time. The digital twin model can be dynamically updated based on real-time collected data, such as changes in voltage and current, and detection results of appearance damage, etc., to generate a corresponding virtual model to maintain consistency with the physical equipment. The digital twin model will integrate data from energy storage equipment, photovoltaic power generation, traditional power generation equipment and power grids to ensure that the virtual model accurately reflects the coordination and interaction of all energy sources.

[0180] Updates to the digital twin model

[0181] Trigger mechanism: When the image data collected by the data acquisition module shows obvious changes in the appearance of energy equipment, such as cracks, damage, etc., or one or more parameters in the operating data, such as voltage, current, temperature, etc., change beyond the pre-set normal range, the system automatically triggers the update process of the digital twin model to ensure that the virtual model can respond to the state changes of the physical equipment in a timely manner. Taking lithium batteries as an example, when the collected voltage value is less than 2.5V or greater than 4.2V, it is judged to be beyond the normal range, triggering the update of the digital twin model.

[0182] Incremental learning: Use incremental learning methods, such as gradient-based online learning algorithms, to dynamically optimize the parameters of the digital twin model. By introducing newly collected data in real time, the model can quickly adapt to changes in the operating status of the multi-source energy system, ensuring that the virtual model is always highly consistent with the operating status of the equipment in the actual system, providing an accurate simulation environment for fault detection and prediction.

[0183] Real-time adjustment: Combine newly collected data to optimize the fault detection algorithm online to improve its adaptability and real-time performance in dynamic environments. Based on the updated data, the digital twin model immediately issues real-time warnings for system faults and adjusts the model accordingly based on the warning information. At the same time, the fault detection algorithm is optimized online in combination with new data to improve the model's adaptability and response speed to dynamic environmental changes, ensuring that the system can promptly detect and handle potential faults under complex and changing operating conditions. For example, when it is found that the grid voltage fluctuates greatly, the digital twin model promptly adjusts the simulation of the energy storage device's charging and discharging strategy, predicts the optimal working state of the energy storage device under the current grid conditions, and feeds back relevant information to the fault detection and root cause analysis module to assist it in making more accurate fault judgments and analyses.

[0184] Fault detection and root cause analysis module

[0185] Real-time and all-round detection of various types of equipment in the multi-source energy system is carried out. The collected multi-modal data is deeply analyzed through hybrid algorithms, and combined with the digital twin model, root cause analysis is performed to locate the equipment and cause of the fault. Multi-modal data includes multiple data sources such as energy storage equipment data, grid-connected data, power load data, photovoltaic power generation data, traditional power generation equipment data, image data and environmental data, which can comprehensively identify and analyze potential sources of faults.

[0186] For example, if abnormal parameters such as voltage, current, and temperature are detected, the system will locate possible causes of failure, such as equipment overheating, battery damage, and load fluctuations, through data processing and model analysis, combined with the simulation and prediction results of the digital twin model on the equipment's operating status.

[0187] The results of fault diagnosis and root cause analysis are displayed through a visual interface to provide decision support for operation and maintenance personnel. This module will comprehensively display data from different data sources, including real-time operation data of energy storage equipment, power output of photovoltaic power generation, working status of traditional power generation equipment, fluctuation of power grid and changing trend of environmental data, and provide accurate fault warning information to provide comprehensive decision-making basis for operation and maintenance personnel.

[0188] Through real-time monitoring and analysis, the results of fault diagnosis and root cause analysis are used to issue fault warnings to users or management systems, and combined with front-end displays such as HTML5, JavaScript, etc., the present invention is able to generate detailed fault diagnosis reports, intuitive root cause analysis charts, and forward-looking trend prediction charts in real time, helping operation and maintenance personnel to quickly and accurately understand the system status, and take effective troubleshooting measures or preventive maintenance strategies in a timely manner to ensure the stable and efficient operation of the multi-source energy system.

[0189] Here are some examples:

[0190] In a multi-source energy system, multiple lithium battery modules provide power support through the charging and discharging process. During the operation of the multi-source energy system, a battery module overheats, and the temperature sensor detects abnormal temperature data, such as Figure 3 As shown, the system of the present invention performs fault diagnosis and prediction through the following steps:

[0191] (1) Temperature anomaly detection

[0192] First, the temperature signal is analyzed in different frequency bands through wavelet transform, the high-frequency part of the signal is extracted, and potential fault characteristics are identified. Wavelet transform helps to remove noise and extract weak anomalies in battery temperature changes, so as to detect problems earlier. The signal data is processed together with other related data from the power grid and traditional power generation equipment, such as load fluctuations and grid voltage, to provide the system with more comprehensive operating status information.

[0193] (2) Image data analysis

[0194] At the same time, the convolutional neural network analyzes the battery surface image data to extract the damage features of the battery appearance, such as cracks, stains, damage, etc., to assist in determining the cause of the failure. The image data is combined with the sensor data to help the system comprehensively evaluate the health status of the battery, especially the possible failures caused by the battery surface.

[0195] (3) Time series data prediction

[0196] The long short-term memory network is used to model the time series data of the temperature signal to capture the long-term dependency of the battery module temperature fluctuation. LSTM can identify temperature trends and sudden anomalies, providing accurate early warning signals for subsequent fault prediction. At this time, power load data, such as electricity price fluctuations, and grid-connected data, such as grid load fluctuations, will jointly participate in time series prediction to optimize the accuracy of the prediction.

[0197] (4) Grid connection and energy coordination

[0198] The joint processing of grid status and battery energy storage system data helps improve the system's adaptability to external load fluctuations. The real-time load changes and electricity price fluctuations of the grid provide important external environmental data for fault prediction, making the system more real-time and accurate in the fault diagnosis process.

[0199] (5) Digital twin model update

[0200] As temperature and image data are updated in real time, the digital twin model receives real-time data from sensors in a timely manner and automatically triggers the model update of the battery module. The digital twin model reflects the state changes of the battery under abnormal conditions by simulating the performance changes caused by battery overheating. The digital twin model ensures that the virtual model is always consistent with the physical device, so as to monitor the operating status of the device in real time and simulate the interaction between the battery and the grid when connected to the grid.

[0201] (6) Root cause analysis

[0202] With the support of digital twin models and hybrid algorithms, further root cause analysis was conducted. Through comprehensive analysis of multiple data sources, including those from the power grid, energy storage equipment, and load data, it was finally confirmed that the root cause of the fault was an internal short circuit in the battery module. This fault not only caused abnormal temperature, but also affected the discharge capacity of the battery module and the overall stability of the system.

[0203] (7) Fault warning and maintenance suggestions

[0204] Fault information is reported through a visual interface, and maintenance personnel can view the fault type, possible cause, and recommended maintenance measures of the battery module. The system helps maintenance personnel take quick action to ensure stable system operation by intelligently recommending maintenance steps, and coordinates the management of other energy sources, such as grid load and photovoltaic power generation, to optimize energy scheduling.

[0205] The above is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be thought of by a person skilled in the art within the technical scope disclosed by the present invention without creative work should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope defined in the claims.

Claims

1. A multi-source energy system fault detection and root cause analysis system, characterized in that: It includes data acquisition module, data preprocessing module, data feature extraction and dimension reduction module, digital twin module and fault detection and root cause analysis module, among which: The data acquisition module acquires various operation data and image data of the multi-source energy system in real time and uploads them to the data preprocessing module; The data preprocessing module preprocesses the various operating data and image data collected by the data acquisition module, including outlier detection, missing value filling, noise reduction and data standardization; The data feature extraction and dimensionality reduction module extracts features and reduces the dimensionality of various pre-processed operation data and image data, and inputs the pre-processed operation data and image data into a hybrid algorithm combining wavelet transform, CNN and LSTM, wherein wavelet transform and CNN extract features of the pre-processed data, fuse the extracted features to obtain a fused feature set, perform dimensionality reduction processing on the high-dimensional data in the fused feature set, and combine LSTM for time series modeling to detect abnormal patterns and potential faults in multi-source energy systems; The digital twin module builds a virtual model of the multi-source energy system based on the initial data, namely the digital twin model. The digital twin model is dynamically updated according to the real-time collected data, maintains consistency with the physical equipment in the multi-source energy system, reflects the status of the physical equipment, and provides a more accurate basis for fault detection and root cause analysis; In the fault detection and root cause analysis module, a hybrid algorithm analyzes the real-time collected data, detects abnormal patterns and potential faults in the multi-source energy system, compares the prediction results of the real-time data input digital twin model, determines whether there is a deviation, and further confirms the occurrence of the fault; combines the feature extraction capabilities of the digital twin model and the hybrid algorithm to locate the key causes of the fault, simulate the system operation status, analyze the potential causes of the fault, and output the analysis results.

2. A multi-source energy system fault detection and root cause analysis system according to claim 1, characterized in that: The various operating data of the multi-source energy system include photovoltaic power generation data, energy storage equipment data, traditional power generation equipment data, power load data, environmental data and grid-connected data. The various operating data of the multi-source energy system are acquired through various sensors installed in the multi-source energy system; the image data of the multi-source energy system, i.e., the appearance changes of energy equipment, is collected through a camera and transmitted to a data preprocessing module in real time.

3. A multi-source energy system fault detection and root cause analysis system according to claim 2, characterized in that: The data preprocessing module preprocesses the various operation data and image data collected by the data collection module, specifically: Outlier detection, removing obviously abnormal data through statistical methods or threshold detection; Missing value filling: use interpolation, mean method or machine learning prediction method to fill in missing values; Denoising: The collected operating data are used as time series data, and the wavelet transform is used to perform denoising to extract the frequency domain features related to the fault; the image data is processed by CNN to extract the spatial features and remove the irrelevant background noise; For grid-connected data, filtering is first performed, and then wavelet transform is used for noise reduction to extract frequency domain features related to faults to ensure data quality; Data standardization: normalize data from different sources to ensure consistency in data format.

4. A multi-source energy system fault detection and root cause analysis system according to claim 3, characterized in that: The data feature extraction and dimension reduction module uses wavelet transform and CNN to extract features from the preprocessed data, specifically: The voltage data and current data in the energy storage device data are non-stationary signals. Wavelet transform is used to decompose the non-stationary signals, extract frequency domain features of different frequency bands, and reveal potential fault modes. The image data uses CNN to extract key spatial features; The traditional power generation equipment data is used to extract fuel consumption trends through data fitting and difference method or time series decomposition method, and power output fluctuation characteristics are extracted through wavelet transform method or rolling standard deviation method; The grid-connected data uses wavelet transform to extract frequency domain features, analyzes the spectrum components and energy distribution in the frequency domain, and is used to determine the matching between the grid load state and the energy supply; The feature extraction of data feature extraction and dimension reduction module also includes: The photovoltaic power generation data extracts the characteristics of power generation stability, the characteristics of the correlation between light intensity and power generation, and the characteristics of the impact of temperature on power generation efficiency; The environmental data extracts humidity change trend characteristics, wind speed change trend characteristics and air pressure change trend characteristics, and is obtained by determining the change of their mean and variance over time through a time series analysis method.

5. A multi-source energy system fault detection and root cause analysis system according to claim 4, characterized in that: The preprocessed data features are extracted and fused to obtain a fused feature set, specifically: The grid-connected data is integrated with photovoltaic power generation data, energy storage equipment data, and traditional power generation equipment data, and features extracted from different data sources are weighted and integrated through a multi-head attention mechanism to provide more comprehensive energy supply information; The environmental data is combined with features extracted from power load data, energy storage equipment data, traditional power generation equipment data, and photovoltaic power generation data to provide more dimensional information for fault prediction and system scheduling.

6. A multi-source energy system fault detection and root cause analysis system according to claim 5, characterized in that: The dimensionality reduction process of the high-dimensional data in the fusion feature set is specifically as follows: The principal component analysis method is used to reduce the dimension of the high-dimensional data in the fusion feature set, retain key features, and reduce redundant data. The fusion feature set includes: The characteristics of power generation stability, the correlation between light intensity and power generation, and the influence of temperature on power generation efficiency extracted from photovoltaic power generation data; Voltage frequency domain characteristics and current frequency domain characteristics of energy storage equipment data extraction; Fuel consumption trends and power output fluctuation characteristics extracted from traditional power generation equipment data; Spatial features extracted from image data; Frequency domain characteristics of grid stability and load variation extracted from grid-connected data; Environmental data extracts humidity change trend characteristics, wind speed change trend characteristics and air pressure change trend characteristics; The principal component analysis method calculates the covariance matrix of the data, finds the eigenvalues ​​and eigenvectors, selects the main eigenvectors according to the size of the eigenvalues, and thus projects the high-dimensional data into a low-dimensional space, retains the key features, and combines them with the spatial features extracted from the image data to form a unified low-dimensional feature set, thereby simplifying the subsequent analysis process.

7. A multi-source energy system fault detection and root cause analysis system according to claim 6, characterized in that: The LSTM models time series data based on a low-dimensional feature set, captures the long-term dependencies of system operating states, and predicts potential failures and abnormal behaviors.

8. A multi-source energy system fault detection and root cause analysis system according to claim 1, characterized in that: The digital twin model is dynamically updated according to the real-time collected data, and the specific steps are as follows: Trigger mechanism: When the data acquisition module collects image data or one or more of the operating data changes beyond the normal range, the digital twin model is automatically updated; Incremental learning: using incremental learning methods to dynamically update digital twin model parameters, simulate the operating status of multi-source energy systems, and ensure that the virtual model can timely reflect the operating status of equipment in the actual multi-source energy system; Real-time adjustment: The digital twin model provides real-time warning of system failures and makes adjustments based on updated data.

9. A multi-source energy system fault detection and root cause analysis system according to claim 1, characterized in that: The results output by the fault detection and root cause analysis module are specifically shown as follows: Fault warning information: through real-time monitoring and analysis, fault warnings are issued to users or management systems; Root cause analysis report, providing a detailed analysis report listing the fault type, possible causes and recommended maintenance measures; Data visualization displays system operating status, fault prediction and diagnosis results in chart form to facilitate user understanding and decision-making.

Images