Failure mode analysis method and device of power conversion system, medium and equipment
By establishing a causal analysis model and utilizing historical data of multimodal operating parameters, the causes of failures in power conversion systems are inferred, solving the problem of inaccurate analysis in existing technologies and achieving rapid and comprehensive failure mode analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA ENERGY INVESTMENT CORP LTD
- Filing Date
- 2024-11-26
- Publication Date
- 2026-05-26
AI Technical Summary
The accuracy of failure mode analysis in existing power conversion systems is not ideal, making it difficult to find the root cause.
By establishing a causal analysis model, utilizing historical data of multimodal operating parameters, and combining it with a path search algorithm, we can infer the current faults of the power conversion system and their internal and external causes, and provide remedial measures.
It improves the accuracy and comprehensiveness of failure mode analysis for power conversion systems, enabling rapid identification and resolution of potential problems.
Smart Images
Figure CN122088024A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of energy storage technology, and more specifically, to a method, apparatus, medium, and equipment for failure mode analysis of power conversion systems. Background Technology
[0002] Electricity conversion systems include systems that convert electricity into other forms of energy, such as chemicals or heat, or into other forms of electrical energy. Examples include systems used for the electrochemical production of hydrogen, ammonia, alcohols, and methane, or systems that convert electrical energy into pressure energy, chemical energy, mechanical energy, or thermal energy for storage.
[0003] Failure Mode Analysis (FMA) is a formal, structured procedure used to analyze failure mode data of current and past processes to prevent these failure modes from recurring in the future. Currently, the accuracy of FMA in power conversion systems remains unsatisfactory, and it is sometimes difficult to pinpoint the root cause of failure. Summary of the Invention
[0004] The purpose of this disclosure is to provide a method, apparatus, medium, and device for failure mode analysis of power conversion systems, which can improve the accuracy of failure mode analysis of power conversion systems.
[0005] To achieve the above objectives, this disclosure provides a failure mode analysis method for power conversion systems, the method comprising: Based on historical data of the multimodal operating parameters of the power conversion system, a causal analysis model of the power conversion system is established. The causal analysis model includes the correspondence between faults, internal causes, external causes, and remedial measures. Determine the current fault in the power conversion system; The failure mode inference is performed in the causal analysis model using a path search algorithm to obtain the internal causes, external causes, and remedial measures corresponding to the current failure.
[0006] Optionally, the multimodal operating parameters include one or more of the following: The parameters include gas spectra, sound waves caused by equipment vibration, sound waves caused by environmental vibration, images inside the equipment, meteorological images, and parameters in text form.
[0007] Optionally, establishing a causal analysis model of the power conversion system based on historical data of its multimodal operating parameters includes: Acquire historical data of the multimodal operating parameters of the power conversion system; The operating mechanism of the power conversion system is modeled based on the historical data to obtain a mechanism model; The causal analysis model is trained based on the historical data and the mechanistic model to obtain a trained causal analysis model.
[0008] Optionally, training the causal analysis model based on the historical data and the mechanistic model to obtain a trained causal analysis model includes: The historical data is then subjected to feature engineering to obtain feature vector data; The feature vector data is corrected according to the aforementioned mechanism model to obtain the first dataset; The causal analysis model is trained based on the first dataset to obtain a trained causal analysis model.
[0009] Optionally, training the causal analysis model based on the historical data and the mechanistic model to obtain a trained causal analysis model includes: The historical data is then subjected to feature engineering to obtain feature vector data; A data-driven model is established based on the feature vector data; The data-driven model and the mechanism model are fused to obtain a trained causal analysis model.
[0010] Optionally, training the causal analysis model based on the historical data and the mechanistic model to obtain a trained causal analysis model includes: The historical data is then subjected to feature engineering to obtain feature vector data; The causal analysis model is trained based on the feature vector data to obtain a reference model; The reference model is modified based on the mechanistic model to obtain a trained causal analysis model.
[0011] Optionally, training the causal analysis model based on the historical data and the mechanistic model to obtain a trained causal analysis model includes: The historical data is then subjected to feature engineering to obtain feature vector data; Substituting the feature vector data into the mechanism model, a second dataset is obtained; The causal analysis model is trained using the second dataset to obtain a trained causal analysis model.
[0012] This disclosure also provides a failure mode analysis device for a power conversion system, the device comprising: A module is established to build a causal analysis model of the power conversion system based on historical data of the multimodal operating parameters of the power conversion system. The causal analysis model includes the correspondence between faults, internal causes, external causes, and remedial measures. The determination module is used to determine the current fault of the power conversion system; The inference module is used to perform failure mode inference in the causal analysis model using a path search algorithm to obtain the internal causes, external causes, and remedial measures corresponding to the current failure.
[0013] This disclosure also provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method provided in this disclosure.
[0014] This disclosure also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory; the processor is configured to execute the computer program in the memory to implement the steps of the method provided in this disclosure.
[0015] Through the above technical solution, a causal analysis model is established in the power conversion system using historical data of multimodal operating parameters. This model is then used to infer the internal and external causes of current faults in the power conversion system, as well as remedial measures, for failure mode analysis (FMEA). On one hand, the establishment of a causal analysis model for the power conversion system enables rapid and comprehensive failure mode analysis. On the other hand, the fusion of multimodal operating parameters allows for the construction of a causal analysis model from different angles and levels, resulting in a more accurate model and thus more precise FMEA analysis.
[0016] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description
[0017] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a failure mode analysis method for a power conversion system provided in an exemplary embodiment.
[0018] Figure 2 This is a schematic diagram of a causal analysis model provided in an exemplary embodiment.
[0019] Figure 3 This is a schematic diagram illustrating the search path results using a causal analysis model, provided in an exemplary embodiment.
[0020] Figure 4This is a schematic diagram of a failure mode analysis method for a power conversion system provided in an exemplary embodiment.
[0021] Figure 5 This is a schematic diagram illustrating the establishment of a causal analysis model provided in an exemplary embodiment.
[0022] Figure 6 This is a schematic diagram of establishing a causal analysis model provided by another exemplary embodiment.
[0023] Figure 7 This is a schematic diagram of establishing a causal analysis model provided in yet another exemplary embodiment.
[0024] Figure 8 This is a schematic diagram of establishing a causal analysis model provided in yet another exemplary embodiment.
[0025] Figure 9 This is a flowchart of a failure mode analysis method for a power conversion system provided in another exemplary embodiment.
[0026] Figure 10 This is a user interface diagram of a power conversion system failure mode analysis provided in an exemplary embodiment.
[0027] Figure 11 This is a structural block diagram of a failure mode analysis apparatus provided in an exemplary embodiment.
[0028] Figure 12 This is a structural block diagram of an electronic device provided in an exemplary embodiment. Detailed Implementation
[0029] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.
[0030] Figure 1 This is a flowchart of a failure mode analysis method for a power conversion system provided in an exemplary embodiment. For example... Figure 1 As shown, the method includes steps S11 to S13: In S11, a causal analysis model of the power conversion system is established based on historical data of the multi-modal operating parameters of the power conversion system. The causal analysis model includes the correspondence between faults, internal causes, external causes, and remedial measures.
[0031] In S12, the current fault of the power conversion system is determined.
[0032] In S13, the path search algorithm is used to infer failure modes in the causal analysis model to obtain the internal causes, external causes, and remedial measures corresponding to the current failure.
[0033] Power conversion systems can be, for example, click-to-hydrogen systems. Causal analysis models can take various forms, including fault trees, symbolic directed graphs, multi-level flow models, dynamic Bayesian networks, and knowledge graphs. Faults, internal causes, external causes, and remedial measures in a causal analysis model can be interconnected through paths within the model.
[0034] Figure 2 This is a schematic diagram of a causal analysis model provided in an exemplary embodiment. For example... Figure 2 As shown, the causal analysis model is represented as a symbolic directed graph of the causal analysis model of the electro-hydrogen production system. In this symbolic directed graph, the first column of nodes (a1~a6) represents external causes, where a1 represents weather changes, a2 represents power output fluctuations, a3 represents ambient temperature changes, a4 represents high external pressure, a5 represents high ambient temperature, and a6 represents high ambient humidity.
[0035] The second column of nodes (b1~b6) represents faults, where b1 indicates catalyst temperature rise, b2 indicates hydrogen leakage, b3 indicates electrolyzer breakdown, b4 indicates large temperature difference between the inlet and outlet of the alkali solution in the fuel cell stack, b5 indicates high hydrogen content in oxygen, and b6 indicates insufficient hydrogen purity.
[0036] The third column of nodes (c1~c8) indicates internal causes. Among them, c1 indicates that the catalyst near the heating rod is burned out, c2 indicates that the insulating material is corroded by impurities, c3 indicates that the nickel plating of carbon steel is uneven, c4 indicates that the nickel mesh is broken, c5 indicates that the alkaline solution flow rate is slowed down, c6 indicates that the bipolar plate is severely electrochemically corroded, c7 indicates that the diaphragm is broken or degreased, and c8 indicates that the alkaline solution concentration is diluted.
[0037] The fourth column of nodes (d1~d7) represents remedial measures (suggestions). Among them, d1 indicates inspection, d2 indicates strict control of replenishing water and alkali, d3 indicates blue spot test, d4 indicates risk point control, d5 indicates replacing the electrolyte with electroplating, d6 indicates shutdown inspection, and d7 indicates timely replenishment of alkali.
[0038] Nodes in a symbolic directed graph can be connected by directed line segments. Internal cause nodes, external cause nodes, and remedial action nodes can have failure probability scores, ranging from 0 to 1. These nodes can also be displayed in different colors based on their failure probability scores.
[0039] Figure 3 This is a schematic diagram illustrating the search path results using a causal analysis model, as provided in an exemplary embodiment. Figure 3As shown, the current detected fault is the large temperature difference at the alkaline outlet of the fuel cell stack at node b4. Failure mode inference is performed in the causal analysis model using a path search algorithm. The inference results are shown by thick lines, indicating that the current fault is caused by both internal and external factors. The external factor is the change in ambient temperature at node a3, and the internal factor is the slowdown in the flow rate of the alkaline solution at node c5. The causal analysis model suggests controlling the risk point at node d4.
[0040] Path search algorithms are used to find multiple possible causal transmission paths of failure modes and to find the most direct and accurate transmission path. They analyze and quantify the optimal transmission path from the failure point to the root cause point, including transmission path analysis that calculates the vector contribution of each path from the stimulus source to the response, path ranking algorithms that use the probability of each different path as a feature and train and predict using a logistic regression classifier, and knowledge graph reasoning, etc.
[0041] Among them, path search algorithms are algorithms used to find the best path from the starting point to the destination in a map or grid. Examples include Dijkstra's algorithm, A* algorithm, D* algorithm, GBFS (Greedy Best-First Search), and Bellman-Ford algorithm.
[0042] Through the above technical solution, a causal analysis model is established in the power conversion system using historical data of multimodal operating parameters. This model is then used to infer the internal and external causes of current faults in the power conversion system, as well as remedial measures, for failure mode analysis (FMEA). On one hand, the establishment of a causal analysis model for the power conversion system enables rapid and comprehensive failure mode analysis. On the other hand, the fusion of multimodal operating parameters allows for the construction of a causal analysis model from different angles and levels, resulting in a more accurate model and thus more precise FMEA analysis.
[0043] In another embodiment, the multimodal operating parameters may include electrolytic cell power, electrolytic cell heat generation, battery voltage, reversible voltage of the electrolysis process, current density, device operating temperature, circulating pump power, water pump power, device operating pressure, fiber optic sensing data of gas in the gas-liquid separator, high-speed camera images of bubbles in the alkali tank or gas-liquid separator outlet pipe, input and output current of the electrolytic cell, input voltage of the electrolytic cell, ambient temperature, humidity and ambient air pressure of the power conversion system, pressure inside the gas-liquid separator, flow rate and pH value of liquid in the pipeline, and gas concentration at the outlet of the gas-liquid separator.
[0044] Furthermore, multimodal operating parameters may include one or more of the following: gas spectra, sound waves caused by equipment vibration, sound waves caused by environmental vibration, images within the equipment, meteorological images, and parameters in text form. For example, in an electrolytic hydrogen production system, the gas spectrum may include the spectrum of the gas in the gas-liquid separator, equipment vibration may include the vibration of the electrolyzer casing, images within the equipment may include images of the proton exchange membrane inside the electrolyzer, and parameters in text form may include text information such as operating status and operating time.
[0045] When multimodal operating parameters include sound waves caused by environmental vibrations, such as power conversion systems at sea or in extreme environments, causal analysis models can be accurately established, thereby enabling accurate failure mode analysis. When multimodal operating parameters include meteorological images, environmental factors such as wind, rain, and sunlight can be determined based on these images. This allows for the consideration of weather conditions during failure mode analysis when the power conversion system is outdoors, making the failure mode analysis more accurate.
[0046] The spectrum of a gas can be detected by a spectrometer and used to analyze the gas content; the sound waves caused by vibration can be detected by an acoustic sensor and used to analyze the vibration amplitude; the image can be detected by an image sensor and used to analyze the anomalies of the measured object.
[0047] This embodiment adds parameters such as text, image, sound, and spectrum to the conventional parameters, thereby enabling more comprehensive integration and analysis of data sources or different types of data.
[0048] In another embodiment, a causal analysis model of the power conversion system is established based on historical data of the multimodal operating parameters of the power conversion system, including: Acquire historical data of multimodal operating parameters of the power conversion system; The operating mechanism of the power conversion system is modeled based on historical data to obtain a mechanism model; The causal analysis model is trained based on historical data and mechanistic models to obtain a well-trained causal analysis model.
[0049] The mechanistic model obtained by modeling the operating mechanism of the power conversion system based on historical data can include external environment-operating characteristic curves, hydrogen production and temperature operating characteristic curves, and electrolyzer voltage and efficiency characteristic curves. The operating results of a hydrogen electrolyzer are usually the result of the combined effects of external operating parameters and internal intrinsic parameters. The mathematical model of the electrolyzer can include the following models.
[0050] (1) Mathematical model of PEM electrolyzer: The output voltage of the electrolytic cell can be expressed as:
[0051] Open circuit voltage of electrolysis system The formula is:
[0052] in, This is the basic voltage for the electrolysis reaction under thermal equilibrium conditions. The gas constant is... It is Faraday's constant. The activity of water (characterizing the internal parameters of the electrolyzer). These are the hydrogen and oxygen pressures, respectively (characterizing the external parameters of the electrolyzer). The current reaction temperature of the electrolyzer (external characteristics of the electrolyzer) is expressed in °C.
[0053] Activation voltage of electrolysis system The formula is:
[0054] in, This represents the current density (external characteristics of the electrolytic cell). , These represent the charge transfer coefficients of the anode and cathode of the electrolytic cell, respectively. , These represent the exchange current densities of the anode and cathode, respectively (characterizing the internal parameters of the electrolytic cell). This represents the inverse hyperbolic sine function.
[0055] Ohmic voltage of electrolysis system The formula is:
[0056] in, , , , These are the equivalent internal resistances of the electrode layer, gas diffusion layer, catalyst layer, and exchange membrane, respectively. For the thickness of the exchange membrane, For the resistivity of the exchange membrane, The water content of the exchange membrane (characterizing the internal parameters of the electrolyzer). The area of the electrolytic cell chamber (characterizing the internal parameters of the electrolytic cell), in square meters (m²). 2 ).
[0057] (2) Mathematical model of alkaline electrolyzer: The terminal voltage of an alkaline electrolyzer includes the reverse voltage, polarization voltage, and ohmic loss voltage. The UI equation at any temperature is:
[0058] in, , These are the terminal voltage and reverse voltage of the electrolysis chamber, respectively, in volts (V). The ohmic resistance parameter of the electrolyte; This represents the DC current of the electrolytic cell (external characteristic of the electrolytic cell), and the unit is ampere (A). , The overvoltage coefficient (characterizing the internal parameters of the electrolytic cell); This represents the current in the electrolytic cell.
[0059] The reversible voltage under standard conditions (25℃, 1.01325 bar) is 1.229V. The value is 2, representing the number of electrons transferred in each reaction. This is the Faraday constant, with a value of 96485 C / mol; Standard atmospheric pressure; This represents the current pressure of the electrolytic cell (an external characteristic of the electrolytic cell), expressed in bar.
[0060] Training utilizes multimodal data and mechanistic models in a data + knowledge combination approach to identify causal relationships between features and outcomes. Training methods can include similarity-based methods, statistical experiments, quasi-experiments, counterfactual methods, uplift models, causal forests, Markov chains, Shapley values, cluster analysis, information theory, state-space structures, data-driven approaches, knowledge and data fusion, deep learning, graph neural networks, machine learning, and reinforcement learning.
[0061] In this embodiment, the causal analysis model trained by combining data and knowledge has high accuracy.
[0062] Figure 4 This is a schematic diagram of a failure mode analysis method for a power conversion system provided in an exemplary embodiment. The power conversion system includes an electro-hydrogen production system, which includes multiple electro-hydrogen coupling modules connected in series. These electro-hydrogen coupling modules include one or more systems such as electro-hydrogen storage, electro-hydrogen ammonia, and electro-hydrogen alcohol. Figure 4The diagram shows one of the electro-hydrogen coupling modules. The electro-hydrogen production system includes an AC power supply, a DC power supply, an AC / DC converter, electrolyzers 1 to n (including alkaline hydrogen production electrolyzers and PEM hydrogen production electrolyzers), a hydrogen production power supply, a cooling water subsystem, a gas-liquid separation subsystem, a drying subsystem, and multi-modal safety parameter detection equipment. The electrolyzers include one or more of alkaline electrolysis, proton exchange membrane electrolysis, solid oxide electrolysis, and anion exchange membrane electrolysis. Alkaline electro-hydrogen production and PEM electro-hydrogen production are combined in different capacities and quantities, and after hydrogen drying, they are coupled together via a gas pressure regulator before being sent to a hydrogen purification unit. The hydrogen production power supply subsystem consists of several power electronic converter modules connected in parallel. The converter module topology can be a DC / DC converter with DC input or an AC / DC converter with AC input. Each converter is connected to the corresponding alkaline hydrogen production electrolyzer and PEM electrolyzer. The cooling water subsystem controls the flow rate of the circulating pump to ensure that the temperature of the mixed hydrogen production electrolyzer operates within the normal range. The oxygen and hydrogen gas-liquid separation subsystem is used to separate the hydrogen and oxygen generated in the electrolyzer from the water vapor and electrolyte. At the same time, it controls the outlet pressure of hydrogen and oxygen, and controls the oxygen content in hydrogen and the hydrogen content in oxygen to ensure normal chemical reactions and system safety in the electrolyzer. The drying subsystem is mainly used to dry the hydrogen and reduce its humidity. The purification subsystem is mainly used to further purify the generated hydrogen to meet the production requirements.
[0063] In power conversion systems, data is collected using multi-source sensors such as acoustic vibration, infrared, and spectral sensors. Data acquisition and detection are first performed using these multi-source sensors. Sensors used for detection include current sensors, voltage sensors, temperature sensors, pressure sensors, flow sensors, pH sensors, humidity sensors, gas concentration sensors, vibration sensors, image sensors, and acoustic sensors.
[0064] Preprocessing for integrating and analyzing data from different data sources or of different types (such as text, images, sound, sensor data, etc.), including operations such as cleaning, imbalance handling, decomposition, transformation, time synchronization, spatial processing, semantic fusion, and interactive fusion.
[0065] Simultaneously, modeling the process mechanism (external environment-operating characteristic curves and hydrogen production curves) based on historical data allows the fused data and knowledge features to be input into a multimodal causal analysis model for training, outputting a causal relationship graph. Next, the results of fault detection, fault diagnosis, and fault location are used as the starting point of the causal relationship graph. Failure mode inference is performed using path search or path propagation algorithms to identify the root cause of the failure mode. Finally, fault-tolerant control is implemented based on the root cause, ensuring continued stable system operation. Alternatively, other remedial measures (such as manual maintenance or direct shutdown) can be taken based on the root cause.
[0066] Figure 5 This is a schematic diagram illustrating the establishment of a causal analysis model provided in an exemplary embodiment. Figure 5 In this embodiment, the causal analysis model is trained based on historical data and mechanistic models to obtain a trained causal analysis model, including: Historical data is processed using feature engineering to obtain feature vector data; the feature vector data is then corrected based on the mechanistic model to obtain the first dataset; the causal analysis model is trained based on the first dataset to obtain the trained causal analysis model.
[0067] Feature engineering is used to extract the most useful features for model prediction from historical data. This can include methods such as statistical analysis, feature correlation analysis, feature importance analysis, time series analysis, or frequency analysis to highlight key information and reduce unnecessary data dimensionality.
[0068] By correcting the feature vector data according to the mechanistic model, feature vectors that conform to the mechanistic model can be selected and corrected. These selected data can be used as training datasets to train the causal analysis model, which can reduce model errors caused by errors in the training data.
[0069] Figure 6 This is a schematic diagram of establishing a causal analysis model provided by another exemplary embodiment. Figure 6 In this embodiment, the causal analysis model is trained based on historical data and mechanistic models to obtain a trained causal analysis model, including: Historical data is processed through feature engineering to obtain feature vector data; a data-driven model is built based on the feature vector data; the data-driven model and the mechanistic model are then fused to obtain a trained causal analysis model.
[0070] In this embodiment, a data-driven model is established, and a causal analysis model is obtained by fusing the data-driven model and the mechanism model, which can also yield a causal analysis model with high accuracy.
[0071] Figure 7 This is a schematic diagram of establishing a causal analysis model provided in yet another exemplary embodiment. Figure 7 In this embodiment, the causal analysis model is trained based on historical data and mechanistic models to obtain a trained causal analysis model, including: Historical data is processed using feature engineering to obtain feature vector data; the causal analysis model is trained based on the feature vector data to obtain a reference model; the reference model is then corrected based on the mechanistic model to obtain the trained causal analysis model.
[0072] In this embodiment, a causal analysis model is first trained based on feature vector data, and the mechanistic model then plays a corrective role, resulting in the final trained causal analysis model.
[0073] Figure 8 This is a schematic diagram of establishing a causal analysis model provided in yet another exemplary embodiment. Figure 8 In this embodiment, the causal analysis model is trained based on historical data and mechanistic models to obtain a trained causal analysis model, including: Historical data is processed using feature engineering to obtain feature vector data; the feature vector data is then substituted into the mechanistic model to obtain the second dataset; the causal analysis model is trained using the second dataset to obtain the trained causal analysis model.
[0074] In this embodiment, the feature vector data is linked with the mechanism model, which is a simple method with fast modeling speed.
[0075] Figure 9 This is a flowchart of a failure mode analysis method for a power conversion system provided in another exemplary embodiment. (e.g.) Figure 9 As shown, the failure mode analysis method may include the following steps: 1. Acquire historical data of multimodal operating parameters of the power conversion system. Various sensors can be used to detect operating parameters at different locations within the power conversion system.
[0076] 2. Preprocess the multimodal data. This includes aligning the multimodal data in time and space.
[0077] 3. Perform cross-modal fusion of multimodal data consisting of visual, textual, vibrational, and structured data.
[0078] 4. Train a causal analysis model for multimodal failure modes. The network parameters are updated using the worst-case gradient descent method. If the performance on the validation set improves, the parameters are updated. If there is no significant improvement over several consecutive cycles, training is stopped. The final causal analysis model for multimodal failure modes can be represented as a causal relationship graph.
[0079] 5. The interactive coupling of internal and external uncertainties can affect the stable operation of the system. At the moment when internal and external uncertainties occur, the system fault should be judged, the type of fault should be diagnosed, and the location of the fault should be determined. The fault information should be used as the starting point of the causal directed graph for reasoning.
[0080] Fault detection, diagnosis, and localization algorithms can include machine learning models, deep learning models, generative adversarial models, autoregressive integrated moving models, self-attention models, ensemble learning models, LSTM, Autoencoders, and Transformers, etc., and can be trained using processed data. Simultaneously, during training, techniques such as cross-validation, grid search, and Bayesian optimization are used to optimize the model's hyperparameters to improve its accuracy and generalization ability. The output of this module is the time, location, and information of the fault occurrence under the current input data.
[0081] 6. The path search algorithm identifies nodes where the failure mode may propagate, the path inference algorithm pinpoints the root cause of the failure, and fault-tolerant control is implemented based on the root cause. For PEM electrolyzers, active adjustments can be made, while for alkaline electrolyzers, passive adjustments can be made. The causal analysis model can run on environments including GPUs, TPUs, FPGAs, cloud computing, and embedded chips.
[0082] This method enables timely judgment and response to failure states in power conversion systems, and infers failure mode propagation paths based on causal analysis models to identify the causes of system failures. It can leverage the characteristics of various modal information under the start-up and shutdown states and steady-state operation of electrolyzers, using multimodal fusion and knowledge graph algorithms to infer causal relationships between multiple failure mode phenomena. Finally, it searches for the most accurate propagation path to find the most accurate propagation path, providing technical support for renewable energy hybrid hydrogen production systems, breaking through the fault diagnosis technology of hybrid hydrogen production, and laying the foundation for the transformation from new power systems to new energy systems.
[0083] When performing failure mode analysis on a power conversion system according to the method provided in this disclosure, a user interface diagram can be output on a display screen. Figure 10 This is a user interface diagram of a power conversion system failure mode analysis provided in an exemplary embodiment. (See diagram below.) Figure 10 As shown in the user interface diagram of the electro-hydrogen production system, the "Multimodal Data Monitoring" section displays historical data obtained from multimodal data monitoring (including data in the form of line charts, bar charts, images, and videos). The "System-wide Fault Detection and Location" section displays the detected fault locations (circled). The "Failure Mode and Cause Analysis" section displays the symbolic directed graph of the causal analysis model and marks the inferred path results. The "Log Report" section displays the detected faults, causal analysis results, and suggestions. Additionally, the interface includes time, settings, and playback buttons. Users can intuitively understand the process and results of failure mode analysis and interact with it through this interface diagram.
[0084] Figure 11This is a structural block diagram of a power conversion system failure mode analysis device provided in an exemplary embodiment. (See diagram below.) Figure 11 As shown, the failure mode analysis device 1100 includes: The module 1101 is used to establish a causal analysis model of the power conversion system based on historical data of multimodal operating parameters of the power conversion system. The causal analysis model includes the correspondence between faults, internal causes, external causes, and remedial measures.
[0085] The determination module 1102 is used to determine the current faults in the power conversion system.
[0086] The inference module 1103 is used to perform failure mode inference in the causal analysis model using a path search algorithm to obtain the internal causes, external causes, and remedial measures corresponding to the current failure.
[0087] Optionally, the multimodal operating parameters include one or more of the following: The spectrum of the gas in the gas-liquid separator, the sound waves caused by the vibration of the electrolytic cell casing, the operating parameters and the images of the proton exchange membrane inside the electrolytic cell.
[0088] Optionally, the establishment module 1101 includes an acquisition submodule, a modeling submodule, and a training submodule.
[0089] The acquisition submodule is used to acquire historical data of multimodal operating parameters of the power conversion system.
[0090] The modeling submodule is used to model the operating mechanism of the power conversion system based on historical data, and obtain the mechanism model.
[0091] The training submodule is used to train the causal analysis model based on historical data and mechanistic models to obtain a trained causal analysis model.
[0092] Optionally, the training submodule is used for: Historical data is processed using feature engineering to obtain feature vector data; The feature vector data is corrected based on the mechanism model to obtain the first dataset; The causal analysis model is trained based on the first dataset to obtain the trained causal analysis model.
[0093] Optionally, the training submodule is used for: Historical data is processed using feature engineering to obtain feature vector data; Establish a data-driven model based on feature vector data; By fusing data-driven models and mechanistic models, a well-trained causal analysis model is obtained.
[0094] Optionally, the training submodule is used for: Historical data is processed using feature engineering to obtain feature vector data; The causal analysis model is trained based on the feature vector data to obtain a reference model; The reference model is modified based on the mechanistic model to obtain a trained causal analysis model.
[0095] Optionally, the training submodule is used for: Historical data is processed using feature engineering to obtain feature vector data; The second dataset is obtained by substituting the feature vector data into the mechanism model; The causal analysis model is trained using the second dataset to obtain a trained causal analysis model.
[0096] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.
[0097] Through the above technical solution, a causal analysis model is established in the power conversion system using historical data of multimodal operating parameters. This model is then used to infer the internal and external causes of current faults in the power conversion system, as well as remedial measures, for failure mode analysis (FMEA). On one hand, the establishment of a causal analysis model for the power conversion system enables rapid and comprehensive failure mode analysis. On the other hand, the fusion of multimodal operating parameters allows for the construction of a causal analysis model from different angles and levels, resulting in a more accurate model and thus more precise FMEA analysis.
[0098] This disclosure also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory; the processor is used to execute the computer program in the memory to implement the steps of the failure mode analysis method provided in this disclosure.
[0099] Figure 12 This is a block diagram illustrating an electronic device 1200 according to an exemplary embodiment. For example... Figure 12 As shown, the electronic device 1200 may include: a processor 1201 and a memory 1202. The electronic device 1200 may also include one or more of a multimedia component 1203, an input / output (I / O) interface 1204, and a communication component 1205.
[0100] The processor 1201 controls the overall operation of the electronic device 1200 to complete all or part of the steps in the aforementioned failure mode analysis method. The memory 1202 stores various types of data to support the operation of the electronic device 1200. This data may include, for example, instructions for any application or method operating on the electronic device 1200, and application-related data such as contact data, sent and received messages, pictures, audio, video, etc. The memory 1202 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Multimedia component 1203 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 1202 or transmitted via communication component 1205. The audio component also includes at least one speaker for outputting audio signals. I / O interface 1204 provides an interface between processor 1201 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 1205 is used for wired or wireless communication between the electronic device 1200 and other devices. Wireless communication may include Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination of these. Therefore, the corresponding communication component 1205 may include a Wi-Fi module, a Bluetooth module, or an NFC module.
[0101] In an exemplary embodiment, the electronic device 1200 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the failure mode analysis method described above.
[0102] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the failure mode analysis method described above. For example, the computer-readable storage medium may be the memory 1202 including program instructions described above, which may be executed by the processor 1201 of the electronic device 1200 to complete the failure mode analysis method described above.
[0103] In another exemplary embodiment, a computer program product is also provided, which includes a computer program executable by a processor, which, when executed by the processor, implements the steps of the failure mode analysis method described above.
[0104] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.
[0105] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.
[0106] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.
Claims
1. A power conversion system failure mode analysis method, characterized by, The method comprises: establishing a causal analysis model of the power conversion system according to historical data of multi-modal operation parameters of the power conversion system, the causal analysis model comprising a corresponding relationship between a fault, an internal cause, an external cause and a remedial measure; determining a current fault of the power conversion system; performing failure mode inference in the causal analysis model by using a path search algorithm to obtain an internal cause, an external cause and a remedial measure corresponding to the current fault.
2. The method of claim 1, wherein, The multi-modal operation parameters comprise one or more of the following: a spectrum of a gas, a sound wave caused by equipment vibration, a sound wave caused by environmental vibration, an image in equipment, a meteorological image and a parameter in text form.
3. The method of claim 1, wherein, The establishing a causal analysis model of the power conversion system according to historical data of multi-modal operation parameters of the power conversion system comprises: obtaining historical data of multi-modal operation parameters of the power conversion system; modeling an operation mechanism of the power conversion system according to the historical data to obtain a mechanism model; training the causal analysis model according to the historical data and the mechanism model to obtain a trained causal analysis model.
4. The method of claim 3, wherein, The training the causal analysis model according to the historical data and the mechanism model to obtain a trained causal analysis model comprises: performing feature engineering processing on the historical data to obtain feature vector data; correcting the feature vector data according to the mechanism model to obtain a first data set; training the causal analysis model according to the first data set to obtain a trained causal analysis model.
5. The method of claim 3, wherein, The training the causal analysis model according to the historical data and the mechanism model to obtain a trained causal analysis model comprises: performing feature engineering processing on the historical data to obtain feature vector data; establishing a data-driven model according to the feature vector data; fusing the data-driven model and the mechanism model to obtain a trained causal analysis model.
6. The method of claim 3, wherein, The training the causal analysis model according to the historical data and the mechanism model to obtain a trained causal analysis model comprises: performing feature engineering processing on the historical data to obtain feature vector data; training the causal analysis model according to the feature vector data to obtain a reference model; correcting the reference model according to the mechanism model to obtain a trained causal analysis model.
7. The method of claim 3, wherein, The training the causal analysis model according to the historical data and the mechanism model to obtain a trained causal analysis model comprises: performing feature engineering processing on the historical data to obtain feature vector data; substituting the feature vector data into the mechanism model to obtain a second data set; training the causal analysis model according to the second data set to obtain a trained causal analysis model.
8. An electric power conversion system failure mode analysis apparatus characterized by comprising: The device comprises: an establishing module configured to establish a causal analysis model of the power conversion system according to historical data of multi-modal operation parameters of the power conversion system, the causal analysis model comprising a corresponding relationship between a fault, an internal cause, an external cause and a remedial measure; determining module, configured to determine a current fault of the power conversion system; inference module, configured to perform failure mode inference in the causal analysis model by using a path search algorithm, to obtain an internal cause, an external cause and a remedial measure corresponding to the current fault.
9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method in any one of claims 1-7.
10. An electronic device, comprising: comprising: a memory having a computer program stored thereon; a processor configured to execute the computer program in the memory to implement the steps of the method in any one of claims 1-7.