Fault diagnosis method and device for hydrogen-electricity integrated energy system and storage medium

Abstract

The invention discloses a hydrogen-electricity integrated energy system fault diagnosis method, and relates to the technical field of electric power, and the method comprises the steps: carrying out the integral primary fault analysis of a hydrogen-electricity integrated energy system through building a comprehensive fault detection model, and obtaining a first analysis result to preliminarily judge the fault condition of the system; dynamically performing secondary fault analysis on the subsystem according to the first analysis result to obtain a second analysis result of the subsystem; and the first analysis result is verified based on the second analysis result to obtain a final fault diagnosis result of the hydrogen-electricity integrated energy system, so that the situation of fault misjudgment of the subsystem is avoided, the accuracy of the fault diagnosis result of the integrated energy system is improved, meanwhile, a secondary fault detection mechanism of the subsystem reduces the data analysis amount, the analysis efficiency is improved, and the fault diagnosis accuracy of the hydrogen-electricity integrated energy system is improved. The accuracy is improved, the fault diagnosis efficiency is considered, and the accuracy, efficiency and reliability of fault processing are further improved.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a hydrogen-electricity integrated energy system fault diagnosis method, device and storage medium. Background Art

[0002] As a bridge between renewable energy and traditional coal-fired power generation, the hydrogen-electric coupling system uses renewable energy to generate hydrogen for energy storage and uses fuel cells to generate hydrogen for electricity. It can effectively absorb wind and power abandonment, reduce dependence on fossil fuels to achieve energy transformation, and effectively solve the problem of unstable and volatile renewable energy. However, the entire system is composed of different modules such as photovoltaics, wind power, fuel cells, and electrolyzers. The interaction and coordination between these modules are very complex. If one of the modules fails, the performance and efficiency of the entire system will be seriously affected.

[0003] Fault diagnosis of hydrogen-electric integrated energy systems is usually applied to peripheral components such as sensor and actuator component failures. The output performance degradation caused by internal system failures such as flooding and membrane drying deserves attention, and fault-tolerant control of water failures needs further exploration. In addition, since hydrogen-electric energy conversion devices are usually coupled with wind and solar power to achieve energy complementarity and balance, it is necessary to diagnose and warn wind and solar power while conducting fault diagnosis of hydrogen-electric energy conversion devices, and use different fault detection methods for different modules to determine the fault type and find out the location and cause of the fault.

[0004] Conventional fault detection of hydrogen-electric energy integrated systems focuses on overall system analysis and lacks specific fault analysis of each module, which leads to unreliable final fault detection results and difficulty in tracing the most fundamental source of the fault, which in turn leads to problems such as unreliable fault handling.

[0005] The patent "A method, device and medium for comprehensive performance testing of a hydrogen-electric coupling system", publication number: CN117491793A, publication date: February 2, 2024, discloses: obtaining the state operation data of the hydrogen-electric coupling system under normal conditions and preprocessing the data; obtaining a state prediction model based on an LSTM network; processing the state operation data collected in real time through the state prediction model to achieve fault time dimension positioning, and obtaining the state operation data of the hydrogen-electric coupling system during the fault time period; constructing a fault diagnosis model based on the combination of PCA and Fuzzy-ART, processing the state operation data of the hydrogen-electric coupling system during the fault time period, and achieving fault location positioning; outputting fault time dimension positioning data and fault location positioning data to complete fault diagnosis. However, this invention performs fault analysis based on the overall operating status of the system, lacks local targeted fault analysis within the system, ignores the operating characteristics and change trends of each module within the system, and has the problem of unreliable fault detection due to the nonlinearity and coupling complexity of the hydrogen-electric system. Summary of the invention

[0006] The purpose of the present invention is to address the problem that the prior art lacks local targeted fault analysis within a hydrogen-electricity energy integrated system, resulting in poor fault handling effect; a hydrogen-electricity energy integrated system fault diagnosis method, device and storage medium are proposed, and an overall primary fault analysis is performed on the hydrogen-electricity energy integrated system by establishing a comprehensive fault detection model, and a first analysis result is obtained to preliminarily determine the fault condition of the system; a secondary fault analysis is dynamically performed on the subsystem according to the first analysis result to obtain a second analysis result of the subsystem; the first analysis result is verified based on the second analysis result to obtain the final fault diagnosis result of the hydrogen-electricity energy integrated system, thereby avoiding the misjudgment of subsystem faults, helping to accurately find the actual fault state, and improving the accuracy of the fault diagnosis results of the integrated energy system. At the same time, the subsystem secondary fault detection mechanism reduces the analysis data and improves the analysis efficiency, thereby achieving improved accuracy and taking into account the fault diagnosis efficiency, further improving the accuracy, efficiency and reliability of fault handling; in addition, the dynamic scheduling subsystem fault detection model mechanism reduces the number of analyses, saves computing resources, and improves the adaptability of the diagnosis method.

[0007] To solve the above technical problems, according to a first aspect of an embodiment of the present invention, a method for fault diagnosis of a hydrogen-electricity integrated energy system is provided, comprising the following steps: S1. Establish a comprehensive fault detection model for the hydrogen-electricity integrated energy system; S2. Perform a fault analysis on the real-time operation data of the hydrogen-electricity integrated energy system based on the comprehensive fault detection model to obtain a first analysis result; S3. Determine whether the fault of the hydrogen-electricity integrated energy system includes only one power generation type system fault according to the first analysis result. If so, output the fault diagnosis result; if not, execute S4; S4. Perform secondary fault analysis on the subsystems of the hydrogen-electricity integrated energy system and output a second analysis result; S5. Verify the first analysis result based on the second analysis result to determine the final fault diagnosis result of the hydrogen-electricity integrated energy system.

[0008] In this scheme, the comprehensive fault detection model is used to first conduct an overall data analysis of the hydrogen-electric integrated energy system to preliminarily determine the fault condition of the system. If the first analysis result is not a power generation type system fault of number 1, that is, there may be a fault in the hydrogen-electric coupling system or multiple subsystems, then a secondary fault analysis is further performed within each subsystem to output a second analysis result; a result verification is performed on the first analysis result based on the second analysis result. On the one hand, the verification can identify abnormal operating data caused by the correlation between subsystems, that is, a subsystem that has not failed but is affected by the failure of the associated subsystem, resulting in abnormal operating data of the system. The verification avoids the misjudgment of subsystem faults and helps to accurately find the actual fault state, thereby improving the comprehensive energy The accuracy of the fault diagnosis results of the source system is improved, thereby improving the reliability of fault handling; on the other hand, secondary fault detection is only performed on the subsystems related to the first analysis results, and only the data of the corresponding subsystems need to be analyzed, which reduces the amount of analysis data, further improves the accuracy of fault diagnosis and takes into account the efficiency of fault diagnosis; at the same time, by dynamically performing secondary fault detection of each subsystem according to the first analysis result, computing resources are saved; first, based on the comprehensive detection model, an overall analysis and evaluation of the comprehensive system is performed to achieve a preliminary screening of the comprehensive system. When the first analysis result only includes one fault condition, it is directly output as the final fault diagnosis result, without the need to secondary analyze the fault conditions of each subsystem, thereby improving the fault diagnosis efficiency and the adaptability of the method.

[0009] Preferably, S1 comprises the following sub-steps: Extract key data features based on key operating data of hydrogen-electricity integrated energy system; synchronously, collecting environmental parameters, and extracting interaction features between the environmental parameters and the key data features; The key data features and the interactive features are integrated as input parameters of a deep learning model to perform model training and obtain a basic comprehensive fault detection model; The hyperparameters of the network are adjusted through a training process to optimize the basic comprehensive fault detection model to obtain the comprehensive fault detection model.

[0010] Preferably, the key data features extracted based on the key operating data of the hydrogen-electricity integrated energy system include: Collect historical normal operation data and historical fault data of the hydrogen-electricity integrated energy system; Performing data preprocessing on the historical normal operation data and the historical fault data, including at least removing noise data and outlier processing, to obtain target operation parameters; Feature selection is performed based on the target operating parameters to extract key data features of each subsystem in the hydrogen-electricity integrated energy system, including at least the first key data feature of the photovoltaic power generation system, the second key data feature of the wind power generation system, and the third key data feature of the hydrogen-electricity coupling system.

[0011] Preferably, S4 includes the following sub-steps: Performing fault analysis on the photovoltaic power generation system based on changes in electrical characteristics of the photovoltaic array to obtain a first fault diagnosis result of the photovoltaic power generation system; Perform fault analysis on the wind power generation system based on the fuzzy sliding mode observer to obtain the second fault diagnosis result of the wind power generation system; analyze the real-time operation data of the hydrogen-electricity coupling system based on the hydrogen-electricity coupling fault detection model to obtain the third fault diagnosis result of the hydrogen-electricity coupling system; The first fault diagnosis result, the second fault diagnosis result and the third fault diagnosis result are used as the second analysis result.

[0012] Preferably, the performing fault analysis on the photovoltaic power generation system based on the change of the electrical characteristics of the photovoltaic array to obtain a first fault diagnosis result of the photovoltaic power generation system includes: Based on the first key data feature, the current change state of each parallel branch of the photovoltaic array is extracted to determine the fault state of the photovoltaic power generation system, and the fault type is determined by the change of the branch current value, wherein: If the current output of the photovoltaic power generation system changes, and the branch current flow direction changes, the photovoltaic power generation system will fail; if the branch current value is less than the threshold, the photovoltaic power generation system will have a short circuit fault; if the branch current value meets the threshold range, the photovoltaic power generation system will have an open circuit fault; The voltages of the faulty branches are compared according to Kirchhoff's current theorem to determine the branch fault point and obtain the first fault diagnosis result.

[0013] Preferably, the performing fault analysis on the wind power generation system based on the fuzzy sliding mode observer to obtain the second fault diagnosis result of the wind power generation system includes: According to the operating characteristics of the wind power generation system, the second key data characteristics are divided into different fuzzy subsets, and the fuzzy rules are determined by combining expert knowledge and system operating data; Establishing a fuzzy TS system sliding mode observer based on the fuzzy rules; The fault state of the wind power generation system is detected according to the error between the state vector estimated by the sliding mode observer of the fuzzy TS system and the state vector actually measured, and the second fault diagnosis result is obtained.

[0014] Preferably, the analyzing the real-time operation data of the hydrogen-electricity coupling system based on the hydrogen-electricity coupling fault detection model to obtain the third fault diagnosis result of the hydrogen-electricity coupling system includes: Extracting fault features of the hydrogen-electricity coupling system based on the third key data features, training a radial basis function neural network based on the fault features, and acquiring a basic fault model; Adding a second radial basis function neural network to the basic fault model and using the second radial basis function neural network as a fault classifier; Based on the operating parameters of the fuel cell system input to the hydrogen-electric coupling system and the residual signal output by the basic fault model, a four-dimensional vector is generated, and the four-dimensional vector is normalized and used as the input of the fault classifier for training to construct the hydrogen-electric coupling fault detection model; The real-time operation data of the hydrogen-electricity coupling system is detected based on the hydrogen-electricity coupling fault detection model, and the third fault diagnosis result including the fault type is output.

[0015] Preferably, S5 comprises the following sub-steps: Compare the first analysis result with the fault information in the second analysis result to determine whether the fault type and fault range are consistent; If the fault types in the first analysis result are more than those in the second analysis result, or the fault range in the first analysis result is larger than that in the second analysis result, then an early warning is issued for the fault types or fault ranges of the different parts; If the number of fault types in the first analysis result is less than the number of fault types in the second analysis result, or the fault range in the first analysis result is smaller than the fault range in the second analysis result, the second analysis result will be used as the final fault diagnosis result of the hydrogen-electric integrated energy system.

[0016] In this solution, by integrating the secondary analysis results of the subsystem and the primary analysis results of the comprehensive fault detection model, the information at different levels is comprehensively considered to improve the reliability and accuracy of fault diagnosis. By verifying the results, the limitations of a single model are avoided, the credibility of the fault diagnosis results is enhanced, and a more scientific basis is provided for system maintenance and optimization.

[0017] According to a second aspect of an embodiment of the present invention, there is provided a fault diagnosis device for a hydrogen-electricity integrated energy system, comprising: a data acquisition module, for collecting various operating data of the hydrogen-electricity integrated energy system, and extracting key data features and interaction features between the data; The first fault diagnosis module is used to construct a hydrogen-electricity integrated fault detection model, perform a fault analysis on the hydrogen-electricity integrated energy system, and obtain a first analysis result; A second fault diagnosis module is used to perform a secondary fault analysis on the subsystem of the hydrogen-electricity integrated energy system and output a second analysis result; A fault verification module is used to determine whether the fault of the hydrogen-electricity integrated energy system only includes a power generation type system fault based on the first analysis result, and to verify the first analysis result based on the second analysis result to determine the final fault diagnosis result of the hydrogen-electricity integrated energy system.

[0018] A computer storage medium having non-volatile program codes executable by a processor, wherein a hydrogen-electricity integrated energy system fault diagnosis program is stored on the computer storage medium, and when the hydrogen-electricity integrated energy system fault diagnosis program is executed by the processor, a hydrogen-electricity integrated energy system fault diagnosis method is implemented.

[0019] Beneficial effects of the present invention: 1. Real-time fault monitoring of the hydrogen-electricity integrated energy system is achieved through a comprehensive fault detection model, which can quickly determine whether the system has a fault and take into account the influence of environmental factors, thereby improving the accuracy of fault judgment; 2. The first analysis result can be used to make a preliminary judgment on the fault of the hydrogen-electric integrated energy system, which can narrow the scope of troubleshooting to a specific subsystem or system type, thereby making more effective use of resources, and allowing targeted selection of subsequent fault diagnosis paths to improve the accuracy and pertinence of fault diagnosis; at the same time, it helps to quickly locate the approximate scope of the fault, avoid unnecessary fault analysis steps, and improve the efficiency of the diagnostic process; 3. Through secondary fault analysis of each subsystem, more detailed fault information is output, which improves the accuracy and detail of fault diagnosis, and finds faults that may be missed or difficult to detect by the comprehensive fault detection model from the perspective of the subsystem. In addition, the secondary analysis of different subsystems can use the data and characteristics unique to the subsystem to more accurately locate the source of the fault, thereby better guiding the subsequent maintenance work. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0021] Figure 1 The present invention is a flow chart of a method for fault diagnosis of a hydrogen-electricity integrated energy system according to an embodiment.

[0022] Figure 2 A schematic diagram of a photovoltaic array simulation structure according to an embodiment.

[0023] Figure 3A schematic diagram of a photovoltaic array open-circuit fault current change in an embodiment.

[0024] Figure 4 A schematic diagram of a photovoltaic array detection voltage and output voltage according to an embodiment.

[0025] Figure 5 The present invention is a flow chart of an active fault-tolerant control applied to a wind power generation system when a fault occurs according to an embodiment.

[0026] Figure 6 The present invention is a schematic diagram of fault tolerance control results of a high-speed shaft speed sensor of a wind power generation system according to an embodiment.

[0027] Figure 7 The present invention is a schematic diagram of fault-tolerant results of wind energy utilization coefficient of a wind power generation system under fault-tolerant control in an embodiment.

[0028] Figure 8 The present invention is a tip speed ratio fault tolerance result of a wind power generation system under fault tolerance control in an embodiment.

[0029] Fig. 9 The figure is a block diagram of an active fault-tolerant control structure of a hydrogen-electricity coupling system according to an embodiment.

[0030] Fig.10 A schematic diagram of a module of a hydrogen-electricity integrated energy system fault diagnosis device according to an embodiment. DETAILED DESCRIPTION

[0031] In order to make the objectives, technical solutions 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 implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] Example 1: Figure 1 As shown, a fault diagnosis method for a hydrogen-electricity integrated energy system includes steps S1-S5, wherein: S1, establishing a comprehensive fault detection model for a hydrogen-electricity integrated energy system, wherein: Key data features are extracted based on the key operating data of the hydrogen-electric integrated energy system; synchronously, environmental parameters are collected, and the interaction features between the environmental parameters and the key data features are extracted; the key data features and the interaction features are fused as input parameters of the deep learning model, and the model is trained to obtain a basic comprehensive fault detection model; the hyperparameters of the network are adjusted through the training process to optimize the basic comprehensive fault detection model and obtain a comprehensive fault detection model.

[0033] Specifically, the interaction characteristics between environmental parameters and key data features mainly include: the correlation coefficient that affects the operating characteristics of the hydrogen-electric integrated energy system under environmental conditions; for example, the change in the correlation coefficient between light intensity and the output power of the photovoltaic power generation system under different temperatures and humidities, etc. By screening out the features that have an important impact on fault diagnosis and reducing the data dimension, it helps to improve the efficiency of the model.

[0034] Specifically, key data features are extracted based on the key operating data of the hydrogen-electric integrated energy system, including: Collect historical normal operation data and historical fault data of the hydrogen-electricity integrated energy system; perform data preprocessing on the historical normal operation data and historical fault data, including at least removing noise data and outlier processing, to obtain target operation parameters; perform feature selection based on the target operation parameters, and extract key data features of each subsystem in the hydrogen-electricity integrated energy system, including at least the first key data feature of the photovoltaic power generation system, the second key data feature of the wind power generation system, and the third key data feature of the hydrogen-electricity coupling system.

[0035] Furthermore, the first key data feature includes at least the output voltage and current of the photovoltaic power generation system in normal operating state and fault state; the second key data feature includes at least the relationship curve characteristics of power coefficient and wind speed, the vibration spectrum characteristics of the generator, the output voltage and current, and the fault output information of the sensor; the third key data feature includes at least the hydrogen pressure, flow, electrolyzer voltage and current, and fuel cell output voltage and current in normal operating state and fault state; the environmental parameters include at least temperature and humidity.

[0036] In this embodiment, by collecting historical normal operation data and historical fault data from different subsystems of the hydrogen-electric integrated energy system, the key information of the photovoltaic power generation system, the wind power generation system and the hydrogen-electric coupling system are covered, providing a comprehensive data basis for subsequent model training; by preprocessing these data, removing noise and outliers, ensuring the quality and reliability of the data, the model training is not disturbed by abnormal data. By extracting the key data features of each subsystem, its operating status and fault characteristics can be reflected from the perspective of different subsystems, providing a specific basis for subsequent fault diagnosis. By collecting environmental parameters and extracting interactive features with key data features, the impact of environmental factors on system operation is considered, the dimension of fault diagnosis is enriched, and the model can perform fault diagnosis under the consideration of environmental influences; at the same time, considering the interaction between environmental factors and key data features, it can effectively avoid misjudgment caused by environmental changes (such as shading affecting photovoltaic power generation systems). When environmental factors such as light intensity, temperature and humidity change, the model can judge whether the change in system status is a normal fluctuation caused by environmental factors or the occurrence of a fault based on its relationship with key data features. The basic comprehensive fault detection model is optimized by adjusting the network hyperparameters to improve the performance and generalization ability of the model so that it can better adapt to different operating conditions and fault situations.

[0037] S2. Perform a fault analysis on the real-time operation data of the hydrogen-electricity integrated energy system based on the comprehensive fault detection model to obtain a first analysis result.

[0038] Specifically, S2 includes the following steps: Collect data of the hydrogen-electric integrated energy system in real time, including at least the output voltage and current of the photovoltaic power generation system, the wind speed, power, and vibration information of the wind power generation system, the hydrogen pressure, flow, voltage, current, etc. of the hydrogen-electric coupling system, and environmental parameters such as temperature and humidity; extract the same key data features and interactive features as those used in training the model from the above real-time data, and input them into the trained integrated fault detection model; output a first analysis result through the integrated fault detection model according to the input real-time features, and the first analysis result can be a numerical value representing the fault probability, a classification code of the fault type, or a binary judgment of whether there is a fault or not.

[0039] In this embodiment, real-time fault monitoring of the hydrogen-electricity integrated energy system is achieved through a comprehensive fault detection model, which can quickly determine whether the system has a fault and take into account the influence of environmental factors, thereby improving the accuracy of fault judgment. Providing preliminary information for subsequent fault judgment and processing steps helps to improve the overall efficiency of fault diagnosis. Among them, through the integration of data at the subsystem level, it is possible to avoid relying solely on the information of a certain subsystem and reduce misdiagnosis or missed diagnosis due to incomplete local information. For example, when an abnormality occurs in the system, the comprehensive fault detection model can determine from a global perspective whether it is due to a single subsystem failure or a problem with the collaborative work between multiple subsystems.

[0040] S3. Determine whether the fault of the hydrogen-electricity integrated energy system includes only one power generation type system fault according to the first analysis result. If so, output the fault diagnosis result; if not, execute S4.

[0041] In this embodiment, a logical judgment is made on the first analysis result to distinguish whether the fault mainly occurs in a subsystem of a certain power generation type, such as a photovoltaic power generation system or a wind power generation system. Through the preliminary judgment of the fault, the scope of troubleshooting can be narrowed to a specific subsystem or system type, so that resources can be used more effectively, and the subsequent fault diagnosis path can be selected in a targeted manner to improve the accuracy and pertinence of fault diagnosis. At the same time, it helps to quickly locate the approximate scope of the fault, avoid unnecessary fault analysis steps, and improve the efficiency of the diagnostic process. For example, if it is judged to be a power generation type system fault, the fault location step can be directly entered to save time and computing resources; if not, a more detailed subsystem level analysis is further performed to avoid missing problems in other subsystems. In addition, since environmental factors can affect the performance and efficiency of the power generation system, secondary fault analysis of each subsystem, such as a photovoltaic power generation system and a wind power generation system, can more accurately obtain fault information and fault points.

[0042] S4. Perform secondary fault analysis on the subsystems of the hydrogen-electricity integrated energy system and output a second analysis result, wherein: based on the change in the electrical characteristics of the photovoltaic array, perform fault analysis on the photovoltaic power generation system to obtain a first fault diagnosis result of the photovoltaic power generation system; based on the fuzzy sliding mode observer, perform fault analysis on the wind power generation system to obtain a second fault diagnosis result of the wind power generation system; based on the hydrogen-electricity coupling fault detection model, analyze the real-time operation data of the hydrogen-electricity coupling system to obtain a third fault diagnosis result of the hydrogen-electricity coupling system; and use the first fault diagnosis result, the second fault diagnosis result and the third fault diagnosis result as the second analysis result.

[0043] In this embodiment, when the first analysis result shows that the fault may not only be a power generation type system fault, or more in-depth information is required, a detailed fault analysis is performed on each subsystem. Special fault detection models and methods are used for photovoltaic power generation systems, wind power generation systems, and hydrogen-electric coupling systems, respectively, so as to facilitate in-depth mining of fault information within the subsystems. By performing secondary fault analysis on each subsystem and outputting more detailed fault information, the accuracy and detail of fault diagnosis are improved, and faults that may be missed or difficult to detect by the comprehensive fault detection model are discovered from the perspective of the subsystem. In addition, the secondary analysis of different subsystems can use the data and characteristics unique to the subsystem to more accurately locate the source of the fault, thereby better guiding subsequent maintenance work.

[0044] Specifically, performing fault analysis on the photovoltaic power generation system based on the change of the electrical characteristics of the photovoltaic array to obtain a first fault diagnosis result of the photovoltaic power generation system includes: Based on the first key data feature, the current change state of each parallel branch of the photovoltaic array is extracted to judge the fault state of the photovoltaic power generation system, and the fault type is determined by the change of the branch current value, wherein: if the current output of the photovoltaic power generation system changes, and the branch current flow direction changes, the photovoltaic power generation system fails; if the branch current value is less than the threshold, a short circuit fault occurs in the photovoltaic power generation system; if the branch current value meets the threshold range, an open circuit fault occurs in the photovoltaic power generation system; according to Kirchhoff's current theorem, the voltages of each faulty branch are compared to determine the branch fault point and obtain the first fault diagnosis result.

[0045] In this embodiment, the photovoltaic array of the photovoltaic power generation system is composed of M*N photovoltaic modules, that is, M branches are connected in parallel, and each branch is composed of N photovoltaic modules in series. The size of M and N plays a decisive role in the size of the photovoltaic array capacity. In order to determine whether a fault occurs, it can be determined by the current sensor on each branch, and the specific location of the fault in the branch can be determined by the voltage sensor on each branch. In order to make the electrical matching of the photovoltaic array normal, the number of photovoltaic modules on each branch is the same. Under normal working conditions, the current and voltage output by each branch should also be the same. Therefore, the type and location of the fault can be determined based on the changes in voltage and current.

[0046] Specifically, when the photovoltaic array can work stably and efficiently at the maximum power point, if a short-circuit fault occurs, the reduction in the number of photovoltaic cells serving as the current source will lead to a reduction in the output current, and the current will flow from the high-voltage branch to the low-voltage branch, and the current in the short-circuited branch will also be temporarily negative; if an open-circuit fault occurs, the output current of the faulty branch will be close to zero, but because the open-circuit voltage of other components in the branch will greatly exceed the operating voltage of other branches when an open-circuit fault occurs, the component with the open-circuit fault will be subjected to a reverse voltage, causing the diode to turn on, and the entire branch starts to work again, but the output current will be reduced accordingly.

[0047] Furthermore, according to Kirchhoff's current theorem, the output current of the normally operating photovoltaic module in the faulty branch will be reduced due to the series connection with the faulty module, and the increase in the current flowing through the diode and the resistor will increase the output voltage of the normal photovoltaic module. Therefore, when analyzing the detected branch voltage, when N is an even number, the number of detected modules is N / 2. If its output voltage is greater than U / 2 (where U is the output voltage of the photovoltaic array), the fault is not in these N / 2 modules. If the voltage is less than U / 2, the fault point is among these N / 2 modules. When N is an odd number, the number of detected modules is (N+1) / 2. Similarly, the fault point can be accurately found by comparing the voltages.

[0048] As a specific example analysis, Figure 2 As shown, this embodiment builds a simulation model of a small photovoltaic array fault diagnosis system including an MPPT algorithm controller. The simulation model uses photovoltaic components as the smallest unit, and a 2*4 SP structure photovoltaic array formed by series-parallel connection is the overall system, where Ai represents the current sensor of the i-th branch, and Vij represents the j-th voltage sensor in the i-th branch. In the photovoltaic array simulation system, the current and voltage sensors are replaced by oscilloscopes and the corresponding branch currents and voltages are detected. In this photovoltaic array, the number of current sensors and voltage sensors in each branch can be derived according to the mathematical relationships in the previous sections, that is, one current sensor and two voltage sensors, and the output voltage of the photovoltaic array can be measured. The fault diagnosis process is as follows: First, compare and observe the current value of each branch to determine whether a fault has occurred. If a fault has occurred, determine the branch where the fault has occurred and the type of fault; then compare the voltage measurement value of the faulty branch with half of the output voltage in turn, and finally determine the location of the faulty PV module. When an open circuit fault occurs in the PV array, it is equivalent to an infinite resistance connected in series with the faulty branch, and the current will be close to 0. However, since there is an MPPT algorithm controller in the system, in the event of an open circuit fault, the open circuit voltage of other components in the branch will greatly exceed the operating voltage of other branches, causing the component with the open circuit fault to be subjected to a reverse voltage, causing the diode to turn on, and the entire branch starts to work again, but the output current will be reduced accordingly. Open circuit fault current detection is as follows: Figure 3 As shown in the figure, it can be seen that the current detected by the current sensor in branch one is close to 0, and when the photovoltaic array is working stably, the output current is reduced to a certain extent relative to the output current of branch two. Therefore, it can be determined that branch one is a fault branch, and the fault type is an open circuit fault.

[0049] At this time, the faulty component is accurately located, such as Figure 4 As shown in the figure, after the open circuit fault occurs in the photovoltaic array, the photovoltaic power generation system can stably operate at the maximum power point, and U11=29.58V, U12=14.21V, U=43.79V, U11>U / 2, U12

[0050] In this embodiment, through simulation analysis of open circuit fault conditions, it can be found that the location and type of the fault can be quickly determined online and in real time in a relatively short period of time, which verifies the feasibility and effectiveness of the fault diagnosis method, provides certain technical support for the timely elimination of faults, and reduces a series of labor costs and time costs caused by unclear fault location and type.

[0051] Specifically, the fault analysis of the wind power generation system is performed based on the fuzzy sliding mode observer to obtain the second fault diagnosis result of the wind power generation system, including: According to the operating characteristics of the wind power generation system, the second key data features are divided into different fuzzy subsets, and the fuzzy rules are determined by combining expert knowledge and system operating data; a fuzzy TS system sliding mode observer is established based on the fuzzy rules; The fault state of the wind power generation system is detected according to the error between the state vector estimated by the sliding mode observer of the fuzzy TS system and the state vector actually measured, and the second fault diagnosis result is obtained.

[0052] ​In this embodiment, when establishing a fuzzy TS system sliding mode observer, a fuzzy TS model is first constructed. According to the operating characteristics of the wind power generation system, the second key data features such as wind speed, pitch angle, etc. are divided into different fuzzy subsets. For example, the wind speed is divided into three fuzzy subsets of low wind speed, medium wind speed and high wind speed. Each subset can be described by a membership function. Fuzzy rules are determined according to expert knowledge and system operation data. For example, if the wind speed is low wind speed and the pitch angle is a small angle, then the output power of the wind power generation system is proportional to the square of the wind speed; for each linear sub-model corresponding to the fuzzy rule, its parameters are determined by a system identification method, and the least squares method, the maximum likelihood estimation method, etc. can be used. For example, for a linear sub-model describing the relationship between output power and wind speed, the output power data at different wind speeds are collected, and the coefficients of the linear model are fitted using the least squares method. These coefficients include the proportional coefficient of the relationship between power and wind speed, the offset, etc.

[0053] Specifically, the nonlinear wind power generation system is approximated by multiple linear sub-models through the fuzzy TS model, and the general form of the fuzzy rules is expressed as follows: R i :if and ...and Then x(k+1)=A i x(k)+B i u(k)(1) Among them, R i represents the i-th fuzzy rule; z 1 ,z 2 ,...,z p As premise variables such as wind speed and pitch angle; is a fuzzy subset; x(k) is the system state vector at time t, such as [i sd ,i sq ,i rd ,i rq ,ω r ] T , where i sd ,i sq are the d-axis and q-axis components of the stator current, i rd ,i rq are the d-axis and q-axis components of the rotor current, ω r is the synchronous speed, T is the transposed matrix; u(k) is the input vector representing the voltage, such as [u rd ,u rq ] T , where u rd ,u rq are the d-axis and q-axis components of the rotor voltage; A i and B iis the system matrix and input matrix of the i-th linear sub-model; For the i-th linear sub-model, construct the sliding mode observer as follows: In the formula, is the state estimation vector, y(k) is the system output vector, such as [i sd ,i sq ] T , C is the output matrix, L i is the observer gain matrix, which is used to ensure the stability of the observer. i is the sliding mode gain matrix, σ(k) is the sliding mode switching function, usually: The state vector estimated by the sliding mode observer The error between the actual measured state vector x(k) and the actual measured state vector x(k) is used to detect the fault and define the fault index: Set the fault threshold J th , when J(k)>J th When the wind power generation system fails, it is determined that the wind power generation system fails.

[0054] Furthermore, the fault type can be determined based on the error variation patterns of different state variables and the physical characteristics of the system. For example, a stator winding short circuit fault may be manifested as a stator current error e sd and e sq Increases rapidly, at this time: In some examples, when a sensor failure occurs in a wind power generation system, the wind power generation system equation is: In the formula, f s is the sensor failure, F is the fault allocation matrix, indicating which sensor providing the measurement is prone to failure, y f is the fault output of the sensor, x(t) is the system state vector at time t, u(t) is the input vector representing the voltage, C i is the output matrix of the ith linear sub-model, is the state estimation vector of the system at time t, y(t) is the output vector of the system at time t, and h(z) is the sliding mode switching function; Based on the above wind power generation system equation, the TS fuzzy sliding mode observer can be designed as follows: In the formula, In order to ensure the sliding mode motion, a sliding mode control strategy is designed, namely: Where ρ is the correction factor (gain), which is determined by the fault amplitude, L i ,L n are the unknown linear feedback matrices and nonlinear feedback matrices of the robust sliding mode observer.

[0055] Then the state estimation error is: In the formula, e=[e x e zf ] T ; Furthermore, by reconstructing the sensor fault, the sensor output signal is corrected, and the controller input is replaced by the corrected sensor output signal, thereby ensuring that the controller outputs accurate control of the wind power generation system and realizing fault-tolerant control of the wind power generation system under sensor failure. The details are as follows: In order to reconstruct the sensor failure, an equivalent output control is introduced, which represents the average characteristics of the sliding mode strategy and the necessary action to maintain the sliding surface motion; when the system slides, Then the error equation formula (11) can be expressed as: The discontinuous switching term, i.e. the sliding mode control strategy v, is used in the design process of ρ , S fuzzy sliding mode observer, that is, formula (8) and (9) This makes the observer deviation equation produce a sliding mode motion, maintaining the estimated error trajectory in the error space, ensuring that the system is robust to external disturbances. However, this will inevitably produce jitter, bring high-frequency interference to the system, and increase the mechanical stress of the system.

[0056] In order to reduce the impact of chattering on the system, the continuous function is used to control the equivalent output v ρeq The approximation method performed is as follows: In the formula, the parameter σ is a scalar with a small value. As long as the parameter is small enough, the equivalent control described by the continuous function can infinitely approach the control strategy, that is, formula (10); Through equations (12), (13) and (14), the sensor fault reconstruction value can be expressed as follows: The sensor fault reconstruction error is: The designed fuzzy system sliding mode observer is used to reconstruct the high-speed shaft speed sensor fault and the electromagnetic torque sensor fault to obtain the sensor fault reconstruction value. Reconstructing values ​​using sensor failures Output signal y for sensor failure f For correction, the system outputs the correction signal as: y r =y f -Ff s (17) Then use the output correction signal y r As the input of the controller, it ensures that the controller outputs accurate control quantity u to the system, thereby realizing active fault-tolerant control of the wind power generation system under sensor failure. At this time, the controller input is: u c =y f -Ff s (18) In the formula, u c =[Ω h Γ G ] T is the input of the sliding mode controller.

[0057] Furthermore, if Figure 5 As shown in the figure, the controller takes the sensor output signal as input. Since the wind power generation system adopts closed-loop feedback control, when the output sensor fails, the input signal and of the controller change, resulting in abnormal control signal u fed back to the system, which in turn affects the maximum wind energy capture effect of the wind power generation system. The output residual of the wind power generation system is obtained through the TS fuzzy sliding mode observer. Based on this signal, the fault reconstruction system is designed. The equivalent output injection theory is used to estimate the sensor fault signal at the output end. The obtained sensor fault reconstruction value compensates for the fault output, corrects the sensor output signal, and uses the corrected output sensor signal as the input of the controller to achieve the active fault-tolerant control goal of the wind power generation system sensor fault, so that the performance of the faulty system is restored to a level similar to that of the intact system without faults.

[0058] It is understandable that the simulation test of fault-tolerant control of wind power generation system under sensor failure is as follows: Make the following two assumptions: (1) The wind power generation system only considers the effect of sensor failure, and does not involve external unknown disturbances and uncertainties; (2) For the convenience of research and simulation experiments, the sensor failure of deviation is considered. Take the high-speed shaft speed sensor failure as an example for simulation verification. The mathematical expression of the fault is shown as follows: (1) Sensor deviation fault is indicated as follows: When the sensor at the output end of the wind power generation system fails, the equivalent output injection method obtains the sensor fault reconstruction value The output sensor signal is corrected to replace the input signal of the controller, providing accurate input to the system controller, ensuring the quality of the closed-loop feedback signal, and realizing active fault-tolerant control of the system when the sensor fails. Taking the high-speed shaft speed sensor as an example, the fault f s1 , we can get the state variable high-speed shaft speed Ω after fault tolerance h , wind energy utilization coefficient Cp, blade tip speed ratio λ and the waveforms of various parameters when intact and without faults, such as Figure 6 , Figure 7 and Figure 8 As shown. Figure 6 It can be seen that the high-speed shaft speed can approximately follow the actual fault-free state after fault tolerance; according to Figure 7 and Figure 8 , it can be seen that the performance parameters wind energy utilization coefficient Cp and blade tip speed ratio λ can be maintained in a range close to the actual values.

[0059] In this embodiment, by comparing the following degree of state variables before and after the fault-tolerant control of the wind power generation system, the recovery of system performance indicators, and the related error analysis of performance indicators, it can be seen that the designed active fault-tolerant control strategy can ensure that the system can achieve real-time fault-tolerant control regardless of whether it is a deviation fault or a drift fault, which reduces the impact of sensor faults on system performance and improves the efficiency of wind energy utilization under faults. Although the effect of fault-tolerant control will be affected by the accuracy of reconstructed faults, resulting in the system performance indicators after fault tolerance failing to reach the actual ideal indicators, there will always be a certain error, but it is still maintained within the acceptable range of the system.

[0060] Specifically, the real-time operation data of the hydrogen-electricity coupling system is analyzed based on the hydrogen-electricity coupling fault detection model to obtain the third fault diagnosis result of the hydrogen-electricity coupling system, including: The fault characteristics of the hydrogen-electric coupling system are extracted based on the third key data characteristics, and the radial basis function neural network is trained based on the fault characteristics to obtain the basic fault model; Add a second radial basis function neural network to the basic fault model and use the second radial basis function neural network as a fault classifier; Based on the operating parameters of the fuel cell system input to the hydrogen-electric coupling system and the residual signal output by the basic fault model, a four-dimensional vector is generated, and the four-dimensional vector is normalized and used as the input of the fault classifier for training to build a hydrogen-electric coupling fault detection model; The real-time operation data of the hydrogen-electricity coupling system is detected based on the hydrogen-electricity coupling fault detection model, and a third fault diagnosis result including the fault type is output.

[0061] It is understandable that, considering the strong nonlinearity of the hydrogen-electric coupling system, if the entire system and the impact of water faults are to be fully analyzed, the powerful approximation ability of neural networks for nonlinearity can be used to introduce radial basis function (RBF) neural networks into the fault diagnosis and fault-tolerant control of flooding and membrane drying, and adopt water fault diagnosis based on RBF neural networks. In this embodiment, the RBF neural network does not need to adjust all parameters on the network at the same time for different inputs, so it has the characteristics of fast convergence and small amount of calculation. By using RBF neural networks to construct a fault detection model for the hydrogen-electric coupling system, the model efficiency can be improved, thereby improving the fault diagnosis efficiency of the hydrogen-electric coupling system. At the same time, since the RBF neural network has a small amount of calculation, the accuracy of the calculation results is guaranteed, and the problem of uncertainty in data quality affecting the calculation accuracy is avoided.

[0062] Furthermore, the RBF neural network uses radial basis functions as basis functions to transform the input vector space into the hidden layer space, making the linearly inseparable problem linearly separable. The steps for fault detection based on the RBF neural network training model are as follows: Based on the two inputs of the PEMFC system (PEMFC system: proton exchange membrane fuel cell system, is an efficient, clean and sustainable new energy technology, mainly used to directly convert the chemical energy of hydrogen and oxygen into electrical energy), the gas pressure and stack current, and the delayed values ​​of the two outputs of the RBF model: the pressure difference on both sides of the membrane and the stack voltage, the combination constitutes the 10 inputs of the RBF neural network model; Three states of the hydrogen-electric coupling fault detection model are defined: healthy, dry membrane and flooded, where dry membrane and flooded membrane are fault states; A basic fault model is developed based on the third key data feature of the hydrogen-electric coupling system and the above 10 inputs, and an error result of basic fault detection is output; On the basis of the neural network model, fault isolation is achieved by adding another RBF network as a fault classifier. The error result of fault detection and the gas pressure and stack current are simultaneously used as the input of the classifier, and the three system states are used as outputs to obtain the fault type. Among them, since the error vector obtained from the fault detection part is caused by the fault, the RBF classifier uses this feature to classify these faults.

[0063] Specifically, during the training process of the fault classifier, the pressure difference on both sides of the membrane, the stack voltage, the gas pressure and the stack current are formed into a 4-dimensional vector, which is normalized and used as the input of the newly added neural network to construct a water fault classifier, which is used as a hydrogen-electric coupling fault detection model.

[0064] In some examples, after the fault is diagnosed, the fault-tolerant controller is used to mitigate the impact of flooding and membrane dry failure on the system. In order to overcome the nonlinear characteristics of the proton exchange membrane fuel cell system and the internal uncertainty when a fault occurs, this application proposes a model reference adaptive fault-tolerant control based on RBF neural network.

[0065] like Fig. 9 As shown in the figure, the purpose of the model reference adaptive fault-tolerant control based on RBF neural network is to ensure that the hydrogen-electric coupling system can still achieve stable output in the event of a fault. When a fault occurs, due to the existence of the PEMFC reference model, the active fault-tolerant controller based on the neural network can give full play to the potential of the system dynamic characteristics, so they can achieve fast and accurate control performance, so that the output of the system with the fault can track the output of the PEMFC reference model. The controller takes the input of the PEMFC reference model, the output of the actual hydrogen-electric coupling system, and the residual with the output of the PEMFC reference model as the input of the neural network, and the output of the neural network as the control input of the actual system, thereby achieving a good approximation and tracking ability of the PEMFC reference model in a healthy state, and ensuring that the system has satisfactory dynamic performance in the event of a water fault.

[0066] Furthermore, the PEMFC reference model is an ideal model established based on the fuel cell system theory, from which the reference model output y m (t) and input r(t), the input of the RBF neural network is the input r(t) of the reference model, and the feedback error e c (t) and the output y(t) of the controlled system. Input x(t) = [r(t), y(t), e c (t)] T ,ω(t)=[ω 1 ,ω 2 ,…ω m ] T is the weight of RBF, h(x,t)=[h 1 ,h 2 ,…h m ] T is the radial basis vector of RBF, where h j is the Gaussian activation function, which is: Among them, c j and b j are the center vector and width of the Gaussian basis function respectively.

[0067] The output of the RBF network is: This can describe the control objectives of the model reference adaptive fault-tolerant control. The entire fault-tolerant control system consists of a reference model in a healthy state, a neural network controller, and an output control law. The purpose is to ensure that when a water fault occurs in the system, the output voltage trajectory can still maintain an ideal trajectory, and ultimately obtain the static and dynamic performance of the control system.

[0068] S5. Verify the first analysis result based on the second analysis result to determine the final fault diagnosis result of the hydrogen-electricity integrated energy system, wherein: The fault information in the first analysis result is compared with that in the second analysis result to determine whether the fault type and fault range are consistent; if the fault types in the first analysis result are more than those in the second analysis result, or the fault range in the first analysis result is larger than that in the second analysis result, then an early warning is issued for the fault types or fault ranges of the different parts; if the fault types in the first analysis result are less than those in the second analysis result, or the fault range in the first analysis result is smaller than that in the second analysis result, then the second analysis result is used as the final fault diagnosis result of the hydrogen-electricity integrated energy system.

[0069] As an implementation method, first, check whether the fault type in the first analysis result is consistent or consistent with the fault type in the second analysis result. For example, if the first analysis result is an electrical fault, it is necessary to check in the second analysis result whether the secondary analysis of the photovoltaic power generation system detects a short circuit or open circuit in the photovoltaic cell, whether the wind power generation system has a generator winding short circuit, and whether the hydrogen-electric coupling system has an electrical connection problem with the electrolyzer or fuel cell. Secondly, for the fault range that may be involved in the first analysis result, such as the entire system or a subsystem, check whether the second analysis result supports the range. If the first analysis result shows that the output state of the entire system is abnormal, it is necessary to use the second analysis result to specifically check the fault conditions of each subsystem of the integrated system to avoid a decrease in the overall output efficiency of the integrated system or other data abnormalities due to a failure in a subsystem, and to avoid misjudgment of faults caused by the associated subsystems due to the collateral impact.

[0070] As another optional implementation, in some cases where the fault diagnosis information is redundant, different weights can be assigned to the first analysis result and the second analysis result according to the reliability, accuracy and importance of the comprehensive fault detection model and each subsystem fault detection model. The comprehensive fault detection model has a certain authority in system-level fault judgment because it takes into account the overall characteristics and environmental factors of the system; while each subsystem fault detection model is more professional in subsystem internal fault detection. For example, the weight w can be assigned to the first analysis result. 1 , assign weight w to the second analysis result 2 , and w 1 +w 2=1, the weight can be determined by historical data verification and expert experience. For example, if the accuracy of the comprehensive fault detection model in the previous system-level fault diagnosis is 80%, a higher w can be appropriately assigned. 1 .

[0071] Furthermore, the first analysis result and the second analysis result are fused according to the assigned weights to calculate the fused failure probability, which can be P=w 1 ×P 1 +w 2 ×P 2 , so as to obtain more accurate and reliable final fault diagnosis results and provide comprehensive support for fault diagnosis of hydrogen-electric integrated energy systems.

[0072] In this embodiment, by integrating the secondary analysis results of the subsystem and the primary analysis results of the comprehensive fault detection model, the information at different levels is comprehensively considered to improve the reliability and accuracy of fault diagnosis. By verifying the results, the limitations of a single model are avoided, the credibility of the fault diagnosis results is enhanced, and a more scientific basis is provided for system maintenance and optimization.

[0073] Embodiment 2, as Fig.10 As shown, a hydrogen-electricity integrated energy system fault diagnosis device comprises: The data acquisition module is used to collect various operating data of the hydrogen-electric integrated energy system and extract key data features and interaction features between data; The first fault diagnosis module is used to construct a hydrogen-electricity integrated fault detection model, perform a fault analysis on the hydrogen-electricity integrated energy system, and obtain a first analysis result; A second fault diagnosis module is used to perform a secondary fault analysis on the subsystem of the hydrogen-electricity integrated energy system and output a second analysis result; The fault verification module is used to determine whether the fault of the hydrogen-electricity integrated energy system only includes a power generation type system fault based on the first analysis result, and to verify the first analysis result based on the second analysis result to determine the final fault diagnosis result of the hydrogen-electricity integrated energy system.

[0074] Furthermore, the present embodiment provides a computer storage medium having a non-volatile program code executable by a processor, wherein a hydrogen-electricity integrated energy system fault diagnosis program is stored on the computer storage medium, and a hydrogen-electricity integrated energy system fault diagnosis method is implemented when the hydrogen-electricity integrated energy system fault diagnosis program is executed by the processor.

[0075] The above specific embodiments are preferred embodiments of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the present specific embodiments. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A method for fault diagnosis of a hydrogen-electricity integrated energy system, characterized in that: The steps include: S1. Establish a comprehensive fault detection model for the hydrogen-electricity integrated energy system; S2. Perform a fault analysis on the real-time operation data of the hydrogen-electricity integrated energy system based on the comprehensive fault detection model to obtain a first analysis result; S3. Determine whether the fault of the hydrogen-electricity integrated energy system includes only one power generation type system fault according to the first analysis result. If so, output the fault diagnosis result; if not, execute S4; S4. Perform secondary fault analysis on the subsystems of the hydrogen-electricity integrated energy system and output a second analysis result; S5. Verify the first analysis result based on the second analysis result to determine the final fault diagnosis result of the hydrogen-electricity integrated energy system.

2. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 1, characterized in that: The S1 comprises the following sub-steps: Extract key data features based on key operating data of hydrogen-electricity integrated energy system; synchronously, collecting environmental parameters, and extracting interaction features between the environmental parameters and the key data features; The key data features and the interactive features are integrated as input parameters of a deep learning model to perform model training and obtain a basic comprehensive fault detection model; The hyperparameters of the network are adjusted through a training process to optimize the basic comprehensive fault detection model to obtain the comprehensive fault detection model.

3. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 2, characterized in that: The key data features extracted from the key operating data of the hydrogen-electricity integrated energy system include: Collect historical normal operation data and historical fault data of the hydrogen-electricity integrated energy system; Performing data preprocessing on the historical normal operation data and the historical fault data, including at least removing noise data and outlier processing, to obtain target operation parameters; Feature selection is performed based on the target operating parameters to extract key data features of each subsystem in the hydrogen-electricity integrated energy system, including at least the first key data feature of the photovoltaic power generation system, the second key data feature of the wind power generation system, and the third key data feature of the hydrogen-electricity coupling system.

4. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 3, characterized in that: The S4 comprises the following sub-steps: Performing fault analysis on the photovoltaic power generation system based on changes in electrical characteristics of the photovoltaic array to obtain a first fault diagnosis result of the photovoltaic power generation system; Perform fault analysis on the wind power generation system based on the fuzzy sliding mode observer to obtain a second fault diagnosis result of the wind power generation system; Analyze the real-time operation data of the hydrogen-electricity coupling system based on the hydrogen-electricity coupling fault detection model to obtain the third fault diagnosis result of the hydrogen-electricity coupling system; The first fault diagnosis result, the second fault diagnosis result and the third fault diagnosis result are used as the second analysis result.

5. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 4, characterized in that: The method of performing fault analysis on the photovoltaic power generation system based on the change of the electrical characteristics of the photovoltaic array to obtain a first fault diagnosis result of the photovoltaic power generation system includes: Based on the first key data feature, the current change state of each parallel branch of the photovoltaic array is extracted to determine the fault state of the photovoltaic power generation system, and the fault type is determined by the change of the branch current value, wherein: If the current output of the photovoltaic power generation system changes, and the branch current flow direction changes, the photovoltaic power generation system will fail; if the branch current value is less than the threshold, the photovoltaic power generation system will have a short circuit fault; if the branch current value meets the threshold range, the photovoltaic power generation system will have an open circuit fault; The voltages of the faulty branches are compared according to Kirchhoff's current theorem to determine the branch fault point and obtain the first fault diagnosis result.

6. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 4, characterized in that: The method of performing fault analysis on the wind power generation system based on the fuzzy sliding mode observer to obtain a second fault diagnosis result of the wind power generation system includes: dividing the second key data features into different fuzzy subsets according to the operating characteristics of the wind power generation system, and determining fuzzy rules in combination with expert knowledge and system operating data; Establishing a fuzzy TS system sliding mode observer based on the fuzzy rules; The fault state of the wind power generation system is detected according to the error between the state vector estimated by the sliding mode observer of the fuzzy TS system and the state vector actually measured, and the second fault diagnosis result is obtained.

7. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 4, characterized in that: The analyzing the real-time operation data of the hydrogen-electricity coupling system based on the hydrogen-electricity coupling fault detection model to obtain the third fault diagnosis result of the hydrogen-electricity coupling system includes: Extracting fault features of the hydrogen-electricity coupling system based on the third key data features, training a radial basis function neural network based on the fault features, and acquiring a basic fault model; Adding a second radial basis function neural network to the basic fault model and using the second radial basis function neural network as a fault classifier; Based on the operating parameters of the fuel cell system input to the hydrogen-electric coupling system and the residual signal output by the basic fault model, a four-dimensional vector is generated, and the four-dimensional vector is normalized and used as the input of the fault classifier for training to construct the hydrogen-electric coupling fault detection model; The real-time operation data of the hydrogen-electricity coupling system is detected based on the hydrogen-electricity coupling fault detection model, and the third fault diagnosis result including the fault type is output.

8. A hydrogen-electricity integrated energy system fault diagnosis method according to claim 4, characterized in that: The S5 comprises the following sub-steps: Compare the first analysis result with the fault information in the second analysis result to determine whether the fault type and fault range are consistent; If the fault types in the first analysis result are more than those in the second analysis result, or the fault range in the first analysis result is larger than that in the second analysis result, then an early warning is issued for the fault types or fault ranges of the different parts; If the number of fault types in the first analysis result is less than the number of fault types in the second analysis result, or the fault range in the first analysis result is smaller than the fault range in the second analysis result, the second analysis result will be used as the final fault diagnosis result of the hydrogen-electric integrated energy system.

9. A hydrogen-electricity integrated energy system fault diagnosis device, applicable to a hydrogen-electricity integrated energy system fault diagnosis method as described in any one of claims 1 to 8, characterized in that: include: The data acquisition module is used to collect various operating data of the hydrogen-electric integrated energy system and extract key data features and interaction features between data; The first fault diagnosis module is used to construct a hydrogen-electricity integrated fault detection model, perform a fault analysis on the hydrogen-electricity integrated energy system, and obtain a first analysis result; A second fault diagnosis module is used to perform a secondary fault analysis on the subsystem of the hydrogen-electricity integrated energy system and output a second analysis result; A fault verification module is used to determine whether the fault of the hydrogen-electricity integrated energy system only includes a power generation type system fault based on the first analysis result, and to verify the first analysis result based on the second analysis result to determine the final fault diagnosis result of the hydrogen-electricity integrated energy system.

10. A computer storage medium having a processor-executable non-volatile program code, characterized in that: The computer storage medium stores a hydrogen-electricity integrated energy system fault diagnosis program, and when the hydrogen-electricity integrated energy system fault diagnosis program is executed by the processor, a hydrogen-electricity integrated energy system fault diagnosis method described in any one of claims 1 to 8 is implemented.

Images