Intelligent detection method and system for electrolysis system failure based on explosion-proof robot dog

Abstract

The application discloses an electrolytic system fault intelligent detection system and method based on an explosion-proof robot dog, relates to the technical field of electrolytic system intelligent detection, and comprises the following steps: collecting electrolytic tank environment data and electrolytic tank operation state data in real time through the explosion-proof robot dog, adopting Kalman filtering and timestamp synchronization methods, performing filtering and denoising processing, and converting into unified timestamps; fusing the processed electrolytic tank environment data and electrolytic tank operation state data through a multivariate statistical method to obtain electrolytic tank multi-source fusion state data, and performing real-time compensation by using a linear regression model; adopting a random forest training method, training an isolation forest model by using the compensated electrolytic tank multi-source fusion state historical data to obtain a short-circuit detection model; and collecting data through the explosion-proof robot dog and processing the data by using Kalman filtering and timestamp synchronization, so that the safety of data collection is ensured, and the accuracy and time dimension consistency of the data are ensured.

Description

Technical Field

[0001] This invention relates to the field of electrolysis system technology, specifically to an intelligent fault detection method and system for electrolysis systems based on an explosion-proof robotic dog. This method and system focus on detecting short-circuit faults during electrolytic cell operation. By integrating multi-source data and employing advanced data processing and model training techniques, it achieves accurate and efficient fault detection in electrolysis systems, providing strong support for the safe and stable operation of electrolysis systems. Background Technology

[0002] In industrial production, electrolytic cells are core equipment, and their stable operation plays a crucial role in the safety and efficiency of the entire production process. Short-circuit faults, as one of the most common types of failures during electrolytic cell operation, can not only lead to a decline in cell performance and affect production efficiency, but also potentially cause serious safety accidents, resulting in personal injury and equipment damage. Therefore, accurate and timely detection of short-circuit faults in electrolytic cells is of paramount importance for ensuring the normal operation of the electrolysis system and the safety of industrial production.

[0003] With the rapid development of sensor technology, the amount of data available for electrolyzers is becoming increasingly abundant, including environmental data (such as electrolyte temperature and gas concentration) and operational status data (such as current, voltage, and electrolyte circulation flow rate). Multi-source data fusion technology is also gradually being applied to the field of short-circuit detection in electrolyzers. By integrating data from different sources, a more comprehensive understanding of the electrolyzer's operating status can be obtained, thereby improving the accuracy and reliability of short-circuit detection.

[0004] However, existing short-circuit detection technologies for electrolytic cells still have many limitations in the data processing stage. During data acquisition, due to the complex operating environment of the electrolytic cell, the acquired data often contains a large amount of noise and interference signals. Simultaneously, the data formats and time bases of different data sources may differ; for example, the sampling frequency of a temperature sensor may differ from that of a current sensor, leading to timing discrepancies in the data. Failure to accurately process these noise and timing discrepancies will severely affect the effectiveness of subsequent data fusion.

[0005] Currently, most short-circuit detection methods fail to fully consider the complexity and diversity of data during data processing, and their handling of noise and timing deviations is inadequate. Some methods merely employ simple filtering techniques to remove noise, but their effectiveness in suppressing complex noise is poor. Regarding time synchronization, there is a lack of effective methods to accurately calibrate the time base of different data sources, resulting in the fused data not accurately reflecting the actual operating status of the electrolyzer.

[0006] Furthermore, existing short-circuit detection models also have shortcomings in training and optimization. Some models are trained based on only a single or a few data sources, failing to fully utilize information from multiple data sources, resulting in limited generalization ability and detection accuracy. Moreover, the parameter adjustments and optimizations during model training are not precise enough, failing to dynamically adjust according to the actual operating conditions of the electrolyzer, further affecting the accuracy of short-circuit detection.

[0007] In summary, existing electrolytic cell short-circuit detection technologies suffer from numerous problems in data processing and model training, making it difficult to meet the safety and efficiency requirements of industrial production for electrolysis systems. Therefore, developing a method and system capable of effectively processing multi-source data and accurately detecting short-circuit faults in electrolytic cells is of significant practical importance. Summary of the Invention

[0008] This invention aims to provide an intelligent fault detection method and system for electrolysis systems based on an explosion-proof robot dog. By integrating multi-source data and applying advanced data processing and model training technologies, it effectively solves the problems of imperfect data processing and insufficient model training in existing technologies, and achieves accurate and efficient detection of faults in electrolysis systems, providing a strong guarantee for the safe and stable operation of electrolysis systems.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0010] In a first aspect, the present invention provides an intelligent fault detection method for an electrolytic system based on an explosion-proof robot dog. The method includes: real-time acquisition of electrolytic cell environmental data and electrolytic cell operating status data using an explosion-proof robot dog; filtering and denoising processing using Kalman filtering and timestamp synchronization methods, and converting the data into a unified timestamp; fusing the processed electrolytic cell environmental data and electrolytic cell operating status data using multivariate statistical methods to obtain multi-source fused state data of the electrolytic cell, and performing real-time compensation using a linear regression model; training an isolated forest model using the compensated historical multi-source fused state data of the electrolytic cell using a random forest training method to obtain a short-circuit detection model; inputting the compensated multi-source fused state data of the electrolytic cell into the short-circuit detection model, and outputting the electrolytic cell short-circuit detection result.

[0011] As a preferred embodiment of the intelligent fault detection method for electrolysis systems based on an explosion-proof robotic dog described in this invention, the specific steps for real-time acquisition of electrolytic cell environmental data and electrolytic cell operating status data by the explosion-proof robotic dog are as follows:

[0012] An explosion-proof robotic dog collects real-time electrolyte temperature and gas concentration values ​​to obtain electrolytic cell environmental data, and also collects real-time current, voltage, and electrolyte circulation flow rate values ​​to obtain electrolytic cell operating status data.

[0013] As a preferred embodiment of the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog described in this invention, the method employs Kalman filtering and timestamp synchronization to perform filtering and noise reduction processing and convert the data into a unified timestamp. The specific steps are as follows:

[0014] The electrolytic cell environmental data and electrolytic cell operating status data are combined into a state vector, and the Kalman filter matrix and noise parameters are defined by analyzing the characteristics of the electrolytic cell.

[0015] Based on the Kalman filter matrix, noise parameters, and state vector, the filtered and denoised electrolytic cell environmental data and electrolytic cell operating status data are obtained through Kalman filter calculation.

[0016] Extract the timestamp information of the electrolytic cell environment data and the electrolytic cell operating status data, calculate the timestamp deviation, and use the timestamp of the electrolytic cell environment data as a reference to align the timestamp of the electrolytic cell operating status data using linear interpolation.

[0017] As a preferred embodiment of the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog described in this invention, the step of fusing processed electrolytic cell environmental data and electrolytic cell operating status data using a multivariate statistical method to obtain multi-source fused state data of the electrolytic cell is as follows:

[0018] The electrolytic cell environmental data and electrolytic cell operating status data after Kalman filtering and timestamp synchronization are standardized using the Z-score method, and key features for short circuit detection are extracted.

[0019] Principal component analysis fusion method is used to extract key information and reduce dimensionality, and then fuse it with standardized electrolytic cell environmental data and electrolytic cell operating status data to obtain short circuit detection related information. The data is then sorted and normalized by feature scaling method to obtain multi-source fused state data of electrolytic cells.

[0020] As a preferred embodiment of the intelligent fault detection method for electrolysis systems based on an explosion-proof robotic dog described in this invention, the real-time compensation using a linear regression model comprises the following steps.

[0021] The historical data of the multi-source fusion state of the electrolyzer was divided into training set, test set and test set. The linear regression model was trained by ordinary least squares linear regression training method.

[0022] Input the multi-source fused state data of the electrolytic cell into the trained linear regression model and output the compensation amount of the multi-source fused state data of the electrolytic cell;

[0023] Based on the compensation amount of multi-source fusion state data of the electrolyzer, and combined with the operation and control principle of the electrolyzer, a compensation strategy is formulated, and the multi-source fusion state data of the electrolyzer is executed for compensation.

[0024] As a preferred embodiment of the intelligent fault detection method for electrolytic systems based on an explosion-proof robot dog described in this invention, the method employs a random forest training approach, utilizing compensated multi-source fusion state historical data of the electrolytic cell to train an isolated forest model, thereby obtaining a short-circuit detection model. The specific steps are as follows:

[0025] The compensated historical data of the multi-source fusion state of the electrolyzer is divided into training set, validation set and test set. The training set is used to train an isolated forest model to detect short circuit anomalies in a random forest manner.

[0026] The performance of the isolated forest model was evaluated using the validation set, and the parameters were adjusted to optimize the isolated forest model. The optimized model was then evaluated using the test set to obtain the short-circuit detection model.

[0027] As a preferred embodiment of the intelligent fault detection method for electrolytic systems based on an explosion-proof robot dog described in this invention, the specific steps for inputting the compensated multi-source fusion state data of the electrolytic cell into the short-circuit detection model and outputting the short-circuit detection result of the electrolytic cell are as follows.

[0028] The compensated multi-source fused state data of the electrolytic cell is input into the trained short-circuit detection model;

[0029] After receiving the compensated multi-source fusion state data of the electrolytic cell, the short-circuit detection model extracts key short-circuit features and performs normalization processing. It then uses its short-circuit detection and judgment capabilities to analyze the data and output the short-circuit detection results of the electrolytic cell.

[0030] Secondly, this invention provides an intelligent fault detection system for electrolytic systems based on an explosion-proof robotic dog, comprising a data acquisition module, a data fusion module, a model training module, and a result output module. The data acquisition module is used to collect real-time environmental data and operating status data of the electrolytic cell through the sensors of the explosion-proof robotic dog, and performs filtering and noise reduction processing using Kalman filtering and timestamp synchronization methods, converting the data into a unified timestamp. The data fusion module is used to fuse the processed environmental data and operating status data of the electrolytic cell using multivariate statistical methods to obtain multi-source fused state data of the electrolytic cell, and performs real-time compensation using a linear regression model. The model training module is used to train an isolated forest model using the compensated historical multi-source fused state data of the electrolytic cell using a random forest training method, obtaining a short-circuit detection model. The result output module is used to input the compensated multi-source fused state data of the electrolytic cell into the short-circuit detection model and output the short-circuit detection result of the electrolytic cell.

[0031] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in the first aspect of the present invention.

[0032] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in the first aspect of the present invention.

[0033] The beneficial effects of this invention are as follows: By training the model using a random forest method combined with compensated multi-source fused data, complex and diverse data features can be effectively mined. Combined with an isolated forest model, the detection accuracy of short-circuit anomalies is improved. Furthermore, the explosion-proof robot dog collects data and uses Kalman filtering and timestamp synchronization processing, ensuring both the safety of data acquisition and the accuracy and consistency of the data over time. The combination of these two methods not only improves the accuracy and reliability of short-circuit detection but also provides a safe and high-quality data foundation for the entire detection process, ensuring the safe and stable operation of the electrolytic cell. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart of the intelligent fault detection method for an electrolysis system based on an explosion-proof robot dog in Example 1.

[0036] Figure 2 This is a flowchart of data preprocessing in Example 1.

[0037] Figure 3 This is a flowchart of model training and short-circuit detection in Example 1.

[0038] Figure 4 This is a schematic diagram of the intelligent fault detection system for the electrolysis system based on an explosion-proof robot dog in Example 1. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0042] Example 1, referring to Figures 1-4 This is the first embodiment of the present invention, which provides an intelligent fault detection method for an electrolysis system based on an explosion-proof robot dog, including the following steps:

[0043] S1. Real-time collection of electrolytic cell environmental data and electrolytic cell operating status data via an explosion-proof robot dog.

[0044] S1.1 The explosion-proof robot dog collects the electrolyte temperature and gas concentration values ​​in real time to obtain the electrolytic cell environmental data, and collects the current value, voltage value and electrolyte circulation flow rate value in real time to obtain the electrolytic cell operating status data.

[0045] It should be noted that the explosion-proof robot dog achieves comprehensive monitoring of the electrolytic cell environment and operating status by incorporating high-precision sensors and a real-time data transmission system. For environmental monitoring, the robot dog uses an explosion-proof temperature sensor to collect real-time electrolyte temperature values, and simultaneously acquires gas concentration data around the electrolytic cell using an explosion-proof gas detector, ensuring environmental safety and obtaining electrolytic cell environmental data. For operating status monitoring, the robot dog integrates a current transformer and a voltage probe to capture real-time current and voltage fluctuations in the electrolytic cell, and accurately measures the electrolyte circulation flow rate using an electromagnetic flowmeter, forming complete operating parameters and obtaining electrolytic cell operating status data.

[0046] S2. Kalman filtering and timestamp synchronization methods are used to perform filtering and noise reduction and convert the data into a unified timestamp.

[0047] S2.1 Combine the electrolytic cell environmental data and electrolytic cell operating status data into a state vector, and define the Kalman filter matrix and noise parameters by analyzing the characteristics of the electrolytic cell.

[0048] It's important to understand the working principle and characteristics of the electrolyzer first. For example, considering the order of influencing factors, certain environmental factors may affect the operation of the electrolyzer first, so data related to these factors can be prioritized. Alternatively, based on the importance and relevance of the data, parameters closely related to the core operation of the electrolyzer should be prioritized to facilitate faster capture of key information during subsequent analysis. For instance, temperature data reflecting the overall operating environment of the electrolyzer might be prioritized, followed by environmental data such as air pressure, and then state data directly affecting the operation of the electrolyzer, such as current and voltage. The collected environmental and operational state data are then arranged in a specific order to form an ordered dataset, which constitutes the state vector.

[0049] Then, the characteristics of the electrolyzer were analyzed in depth from the perspective of data statistical analysis. First, the basic statistics of each parameter in the state vector were calculated, such as the mean, median, and standard deviation. The mean and median can indicate the average operating level; deviations from the ideal range suggest environmental anomalies, while a large standard deviation indicates unstable operation, possibly due to poor electrode contact. Next, the data distribution characteristics, such as skewness and kurtosis, were calculated. Abnormal skewness or peaks suggest special operating conditions. The correlation coefficients between different parameters were calculated; anomalies in these coefficients indicate changes in the interaction of internal factors. Considering the effects of the internal chemical reactions on factors such as temperature and gas concentration, and the relationships between these factors and current and voltage, the elements of the Kalman filter matrix were determined. Simultaneously, based on the sources of various noises during electrolyzer operation, such as sensor noise and environmental interference, corresponding noise parameters were defined.

[0050] S2.2 Based on the Kalman filter matrix, noise parameters, and state vector, the filtered and denoised electrolytic cell environmental data and electrolytic cell operating status data are obtained through Kalman filter calculation.

[0051] It should be noted that, through Kalman filtering calculations, the expressions for the filtered and denoised electrolytic cell environmental data and electrolytic cell operating status data are as follows:

[0052]

[0053] in, These are the filtered and denoised environmental and operational status data of the electrolytic cell at time k. k Z is the Kalman gain matrix at time k. k H is the observation vector at time k. k It is the observation matrix at time k;

[0054] It should be noted that the state vector includes both the electrolyzer's environmental data and its operational status data. The Kalman filter matrix describes the changing patterns of the electrolyzer's state; it is constructed based on the correlation between the electrolyzer's operating characteristics and the environmental and operational status data. The noise parameters reflect the statistical characteristics of noise in the environmental and operational status data, and are determined based on the electrolyzer's working environment and sensor characteristics.

[0055] In the calculation process, the state vector is used as the initial input, and the state at the next moment is predicted by combining the Kalman filter matrix with the prediction update rule of the state vector. Simultaneously, the covariance of the prediction error is calculated using noise parameters to assess the uncertainty of the prediction. The actual measured state vector is compared with the predicted value, and the weight of the predicted and measured values ​​is weighed using the Kalman gain. The Kalman gain, calculated based on the prediction error covariance and the measurement noise covariance, determines the degree of influence of the measured value on the final result when updating the state. By continuously repeating these two steps of prediction and update, the state estimation data in the state vector is gradually corrected, thereby achieving filtering and denoising of the electrolytic cell environmental data and operating status data, ultimately obtaining the filtered and denoised electrolytic cell environmental data and electrolytic cell operating status data.

[0056] S2.3 Extract the timestamp information of the electrolytic cell environment data and the electrolytic cell operating status data, calculate the timestamp deviation, and use the timestamp of the electrolytic cell environment data as a reference to align the timestamp of the electrolytic cell operating status data using linear interpolation.

[0057] It should be noted that the expression for aligning the timestamps of electrolytic cell operating status data using linear interpolation is as follows:

[0058]

[0059] in, This is the value obtained by aligning the electrolytic cell operating status data at timestamp r, where α is the weighting coefficient of the linear interpolation, and p e This represents the value corresponding to the timestamp e in the electrolytic cell environmental data;

[0060] It should be noted that during the data acquisition phase, it is essential to ensure that the acquisition equipment can accurately record the time information corresponding to each data point, thereby obtaining the timestamps for both the electrolytic cell environmental data and the operating status data. After acquisition, these two sets of timestamps are compared one by one, and the timestamp deviation between them is calculated.

[0061] Since the electrolytic cell environmental data timestamps are to be used as the benchmark, the next step is to determine the environmental data timestamp sequence. For each timestamp in the operating status data, the two corresponding time reference points in the environmental data timestamp sequence are found. Then, based on the principle of linear interpolation, the alignment value of the operating status data on the environmental data time reference points is calculated according to the proportional relationship between these two reference points and the corresponding operating status data values. Specifically, the operating status data values ​​of the preceding and following reference points are weighted according to the time interval ratio, with the weight determined by the distance between the current time and the preceding and following reference point times. This completes the alignment operation of the electrolytic cell operating status data timestamps based on the environmental data timestamps.

[0062] S3. The processed electrolytic cell environmental data and electrolytic cell operating status data are fused using multivariate statistical methods to obtain multi-source fused state data of the electrolytic cell.

[0063] S3.1 The electrolytic cell environmental data and electrolytic cell operating status data after Kalman filtering and timestamp synchronization are standardized using the Z-score method, and key features for short circuit detection are extracted.

[0064] It should be noted that the expression for standardization using the Z-score method is:

[0065]

[0066] in, These are the standardized values, where v is the original value, μ is the mean of the dataset, and σ is the standard deviation of the dataset.

[0067] It should be noted that the purpose of standardization is to give the electrolytic cell environmental data and electrolytic cell operating status data, after Kalman filtering and timestamp synchronization, better statistical characteristics. For the processed environmental and operating status data, the mean and standard deviation are calculated separately. To calculate the mean, all values ​​in the electrolytic cell environmental and operating status data are added together and then divided by the number of values. The standard deviation is obtained by taking the square root of the average of the squared differences between each data value and the mean. Then, the Z-score method is applied: for each point in the electrolytic cell environmental and operating status data, the mean is subtracted from the data value, and then the result is divided by the standard deviation. This transforms the electrolytic cell environmental and operating status data into a standard normal distribution with a mean of 0 and a standard deviation of 1, thus achieving standardization.

[0068] After standardization, based on professional knowledge and past experience in electrolytic cell short-circuit detection, the characteristics related to short circuits in the standardized electrolytic cell environmental data and electrolytic cell operating status data were analyzed. This included the numerical changes of certain data features before and after a short circuit, and the correlation changes between different features. From the standardized electrolytic cell environmental data and electrolytic cell operating status data, features that are indicative of short-circuit detection were extracted.

[0069] S3.2. Using principal component analysis fusion method, the main information is extracted and dimensionality is reduced, and then fused with the standardized electrolytic cell environmental data and electrolytic cell operating status data to obtain short circuit detection related information. The data is then sorted and normalized using feature scaling method to obtain multi-source fused state data of the electrolytic cell.

[0070] It should be noted that principal component analysis (PCA) fusion is first used to process the key features of short-circuit detection. This method analyzes the covariance matrix and other parameters of the key features of short-circuit detection to identify the main components, i.e., extract the main information. In this process, based on the inherent structure and characteristics of the key features of short-circuit detection, redundant information is removed, achieving dimensionality reduction and making the key features of short-circuit detection more concise and efficient.

[0071] The standardized electrolytic cell environmental data and electrolytic cell operating status data possess good statistical properties, and when fused with the key features for short-circuit detection obtained from principal component analysis, they can fully combine their respective advantages. During the fusion process, data is integrated according to specific rules and algorithms to obtain a dataset containing rich short-circuit detection information. A feature scaling method is then applied to the fused data. This method adjusts the scale to ensure a more reasonable proportional relationship between different features. After processing and standardization, the final multi-source fused state data of the electrolytic cell is obtained.

[0072] S4. Real-time compensation using a linear regression model.

[0073] S4.1 Divide the historical data of the multi-source fusion state of the electrolyzer into a training set, a test set, and a test set. Train the linear regression model using the ordinary least squares linear regression training method.

[0074] It should be noted that the historical data of the multi-source fusion state of the electrolyzers first needs to be reasonably divided. Based on the characteristics and analysis requirements of the historical data of the multi-source fusion state of the electrolyzers, it is divided into three parts with the training set accounting for 70%, the validation set and the test set each accounting for 15%. The division process should ensure randomness and representativeness, so that each part can reflect the characteristics of the overall historical data of the multi-source fusion state of the electrolyzers to a certain extent.

[0075] After dividing the historical data of the multi-source fusion state of the electrolyzer, the linear regression model was trained using the ordinary least squares (OLS) linear regression method. The goal of OLS is to find an optimal set of regression coefficients that minimizes the sum of squared errors between the predicted and observed values. Mean squared error (MSE) is used to measure the fit of the linear regression model by averaging the squared differences between the predicted and actual values. During training, the linear regression model continuously adjusts the regression coefficients to minimize the MSE.

[0076] It should be noted that the expression for the mean squared error in the ordinary least squares linear regression training method is defined as follows:

[0077]

[0078] Where MSE(β) is the mean squared error, n is the number of samples in the training dataset, y is the true response variable vector of the training dataset, R is the feature matrix of the training dataset, β is the regression coefficient vector, and T is the matrix transpose operation.

[0079] The training data is input into the linear regression model. Based on the principle of least squares, the model continuously adjusts the regression coefficients, calculates the difference between predicted and actual values, and optimizes the model. In each iteration, the mean squared error (MSE) is used to evaluate the model's fit, specifically by calculating the sum of squares of the differences between predicted and actual values ​​and then averaging them. The linear regression model continuously adjusts the regression coefficients based on this evaluation until it finds the coefficient combination that minimizes the MSE, thus completing the training of the linear regression model.

[0080] S4.2 Input the multi-source fusion state data of the electrolytic cell into the trained linear regression model and output the compensation amount of the multi-source fusion state data of the electrolytic cell.

[0081] It should be noted that when it is necessary to detect the short circuit of the electrolytic cell, the real-time acquired and preprocessed multi-source fusion state data of the electrolytic cell is used as input to the trained linear regression model. This data, which carries various state information of the electrolytic cell at the current moment, is the basis for the prediction of the linear regression model.

[0082] After receiving data containing various state information of the electrolyzer at the current moment, the linear regression model will perform calculations according to the rules learned during training. It will match and calculate the input multi-source fusion state data of the electrolyzer with feature weights and regression coefficients, analyze the contribution of each feature to the prediction result and perform weighted summation, and assign greater weight to key features to highlight their influence. After calculation and analysis, the linear regression model will set the compensation amount of the multi-source fusion state data of the electrolyzer based on the input multi-source fusion state data of the electrolyzer.

[0083] S4.3 Based on the compensation amount of multi-source fusion state data of the electrolyzer, and combined with the operation and control principle of the electrolyzer, a compensation strategy is formulated, and the multi-source fusion state data of the electrolyzer is executed for compensation.

[0084] It should be noted that this study provides a comprehensive and in-depth analysis of the operation and control principles of electrolytic cells, detailing the complex chemical reaction processes, physical properties, and interrelationships among various operating parameters within the cell. For example, it examines how changes in current and voltage affect the cell's temperature and gas generation rate, and how these factors collectively influence the overall performance and stability of the electrolytic cell. This thorough understanding of these principles is crucial.

[0085] Next, based on the compensation amount of the multi-source fusion state data of the electrolyzer, the underlying information is analyzed in depth. The compensation amount reflects the degree and direction of deviation of the current multi-source fusion state data from the ideal or normal state. Through detailed study of the compensation amount, the specific circumstances under which the current state of the electrolyzer deviates from the normal range can be determined. For example, if the compensation amount shows a large deviation in temperature-related data, it may mean that the heat distribution inside the electrolyzer is uneven or that the heating equipment is malfunctioning, requiring targeted analysis and evaluation.

[0086] Based on an understanding of the electrolyzer's operating principle and the analysis of compensation parameters, a targeted compensation strategy is formulated. When formulating this strategy, multiple factors need to be considered comprehensively. For example, if a key parameter is found to be outside the normal range, it is necessary to determine, based on the principle, how to adjust other relevant key parameters to correct this deviation. This might involve adjusting the current and voltage to control the reaction rate, or adjusting the operating parameters of the cooling equipment to maintain a stable temperature environment.

[0087] During the compensation phase, following a pre-defined compensation strategy, precise control equipment and actuators are used to finely adjust various operating parameters in the multi-source fusion state data of the electrolyzer. This requires high-precision sensors and controllers capable of real-time monitoring and adjustment to ensure the accuracy and timeliness of the compensation process. For example, adjusting the current controller changes the current in the electrolyzer, or adjusting the valve opening controls the coolant flow rate, thereby achieving effective compensation for the electrolyzer's operating state. During the compensation process, the compensation effect needs continuous monitoring and evaluation, and the compensation strategy should be adjusted promptly based on feedback information until the multi-source fusion state data of the electrolyzer returns to a reasonable range.

[0088] S5. Using the random forest training method, the isolated forest model is trained with the compensated multi-source fusion state historical data of the electrolytic cell to obtain the short circuit detection model.

[0089] S5.1 Divide the compensated electrolytic cell multi-source fusion state historical data into training set, validation set and test set, and use the training set to train an isolated forest model to detect short circuit anomalies in a random forest manner.

[0090] It should be noted that the compensated historical data of the multi-source fusion state of the electrolyzers needs to be partitioned. Based on the characteristics and analytical requirements of the compensated historical data of the multi-source fusion state of the electrolyzers, it is divided into three parts with the training set accounting for 70%, the validation set and the test set each accounting for 15%. The partitioning must ensure randomness and representativeness, so that each subset of the compensated historical data of the multi-source fusion state of the electrolyzers can reflect the characteristics of the overall compensated historical data of the multi-source fusion state of the electrolyzers to a certain extent.

[0091] Using the training set data, the Isolation Forest model was trained using a random forest approach. Specifically, the compensated historical data from the multi-source fusion state of the electrolyzer was preprocessed first. This included randomly sampling a subset of samples with replacement from the training set to construct each decision tree, selecting features highly correlated with short-circuit anomalies, and handling missing and outlier values. Next, a random forest was constructed, with each decision tree trained independently. This was done by randomly selecting splitting features and node splitting points, and finally, the results of all decision trees were combined for judgment. Then, the processed data was mapped to the Isolation Forest feature space, and isolated trees were constructed by randomly selecting features and splitting values. The degree of anomaly was judged based on the path length of data points in the trees to train the Isolation Forest. Finally, the model performance was evaluated through cross-validation, and hyperparameters such as the number of decision trees and isolated tree parameters were adjusted based on the results to optimize the Isolation Forest model's ability to identify short-circuit anomalies.

[0092] S5.2. Evaluate the performance of the isolated forest model using the validation set, adjust the parameters to optimize the isolated forest model, and use the test set to perform the final performance evaluation of the optimized model to obtain the short-circuit detection model.

[0093] It's important to note that by inputting validation set data into the trained Isolation Forest model and comparing its predictions with actual results, relevant evaluation metrics, such as accuracy and recall, are calculated. Based on these metrics, the shortcomings of the Isolation Forest model are analyzed, and its parameters are adjusted accordingly. For example, parameters such as the number of decision trees and tree depth in the random forest are adjusted to continuously optimize the Isolation Forest model's structure, enabling it to more accurately detect short-circuit anomalies. After multiple adjustments and optimizations, the optimized Isolation Forest model is finally evaluated using a test set. The test set data is entirely new and unseen by the Isolation Forest model; the evaluation results on the test set accurately reflect its generalization ability and practical application effectiveness. Through continuous optimization and evaluation, a high-performance short-circuit detection model is obtained.

[0094] S6. Input the compensated multi-source fusion state data of the electrolytic cell into the short-circuit detection model and output the short-circuit detection result of the electrolytic cell.

[0095] S6.1 Input the compensated multi-source fusion state data of the electrolytic cell into the trained short-circuit detection model.

[0096] It is important to note that after completing the compensation processing of the multi-source fusion state data of the electrolytic cells, it is crucial to ensure the accuracy and completeness of the compensated multi-source fusion state data. This involves strict control over the compensation process, ensuring that each compensated multi-source fusion state data point has undergone appropriate correction and adjustment. Next, the trained short-circuit detection model must be confirmed to be operational, with its parameters and structure optimized and adjusted to possess good detection capabilities. In practice, the compensated multi-source fusion state data of the electrolytic cells is organized and arranged according to the format and order required by the short-circuit detection model through a specially designed transmission channel and interface. The compensated multi-source fusion state data is then systematically input into the short-circuit detection model, enabling it to analyze and process the input compensated multi-source fusion state data based on learned patterns and rules, thereby achieving the detection of short circuits in the electrolytic cells.

[0097] S6.2 After receiving the compensated multi-source fusion state data of the electrolytic cell, the short-circuit detection model extracts key short-circuit features and performs normalization processing. It then uses its short-circuit detection judgment capability to analyze the data and outputs the short-circuit detection results of the electrolytic cell.

[0098] It should be noted that after the compensated multi-source fused state data of the electrolytic cell is input into the trained short-circuit detection model, the feature extraction mechanism must be activated first. This mechanism, based on the knowledge learned by the short-circuit detection model during training, filters out key features related to short circuits from the received compensated multi-source fused state data of the electrolytic cell. The extraction of these key features involves multiple aspects, such as analyzing abnormal temperature change trends from temperature data, including rapid temperature increases or continuous abnormal fluctuations in a short period of time; detecting changes in current amplitude, frequency, and current waveform distortion from current data; and focusing on voltage stability from voltage data, such as sudden voltage drops or voltage fluctuations exceeding the normal range. By comprehensively analyzing these different types of data, the short-circuit detection model can comprehensively and accurately extract key features related to short circuits. To eliminate the influence of differences in the dimensions between different features on the detection results, the short-circuit detection model uses a normalization algorithm to map the feature data to a specific interval.

[0099] Finally, the short-circuit detection model is input. Based on its established ability to identify short-circuit anomalies, a large amount of historical data with clear short-circuit annotations is used during training. By analyzing the changes and statistical patterns of various features in this data when short circuits occur, statistical methods or settings based on expert experience and actual operating conditions are used to determine the short-circuit threshold corresponding to each feature. When normalized feature data is received, the short-circuit detection model compares and matches this data with the preset short-circuit thresholds one by one, analyzing whether each feature exceeds the normal range and whether it exhibits a short-circuit-related abnormal pattern. Taking into account the analysis results of all features, the model evaluates the data based on its built-in logical operation mechanism. If the abnormality of multiple key features reaches or exceeds the set comprehensive judgment threshold, or if the abnormality of certain key features shows a specific correlation pattern with other features, conforming to the typical feature combination of short-circuit faults, the short-circuit detection model will determine that the electrolytic cell has a short circuit; otherwise, it will determine that it is normal and output the short-circuit detection result of the electrolytic cell.

[0100] This embodiment also provides an intelligent fault detection system for electrolysis systems based on an explosion-proof robot dog, including: a data acquisition module, a data fusion module, a model training module, and a result output module;

[0101] The data acquisition module is used to collect real-time environmental data and operating status data of the electrolytic cell through the sensors of the explosion-proof robot dog. It uses Kalman filtering and timestamp synchronization methods to perform filtering and noise reduction processing and convert the data into a unified timestamp.

[0102] The data fusion module is used to fuse the processed electrolytic cell environmental data and electrolytic cell operating status data through multivariate statistical methods to obtain multi-source fused state data of the electrolytic cell, and to perform real-time compensation using a linear regression model;

[0103] The model training module is used to train the isolated forest model using the random forest training method and the compensated multi-source fusion state historical data of the electrolyzer to obtain the short circuit detection model.

[0104] The result output module is used to input the compensated multi-source fusion state data of the electrolyzer into the short-circuit detection model and output the short-circuit detection results of the electrolyzer.

[0105] This embodiment also provides a computer device applicable to the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as proposed in the above embodiment.

[0106] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0107] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the intelligent fault detection method for an electrolytic system based on an explosion-proof robot dog, as proposed in the above embodiment. The storage medium 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 Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0108] In summary, this invention utilizes a random forest training method combined with compensated multi-source fused data for model training, effectively uncovering complex and diverse data features. The isolated forest model further enhances the detection accuracy of short-circuit anomalies. Furthermore, the explosion-proof robotic dog collects data and employs Kalman filtering and timestamp synchronization processing, ensuring both the safety of data acquisition and the accuracy and consistency of the data over time. This combination not only improves the accuracy and reliability of short-circuit detection but also provides a safe and high-quality data foundation for the entire detection process, guaranteeing the safe and stable operation of the electrolytic cell.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent fault detection in an electrolysis system based on an explosion-proof robotic dog, characterized in that: include, An explosion-proof robot dog collects real-time environmental data and operating status data of the electrolytic cell. Kalman filtering and timestamp synchronization methods are used to filter and denoise the data and convert it into a unified timestamp. Multivariate statistical methods were used to fuse the processed electrolytic cell environmental data and electrolytic cell operating status data to obtain multi-source fused state data of the electrolytic cell. A linear regression model was then used for real-time compensation. The specific steps are as follows. The process of obtaining multi-source fused state data of the electrolyzer includes standardizing the electrolyzer environment data and electrolyzer operating status data after Kalman filtering and timestamp synchronization, and extracting key features for short-circuit detection. Principal component analysis was performed on the key features of short circuit detection to extract the main information and reduce the dimensionality. The key features of short circuit detection after principal component analysis were then fused with the standardized electrolytic cell environmental data and electrolytic cell operating status data. The fused data was then scaled and normalized to obtain multi-source fused state data of the electrolytic cell. The historical data of the multi-source fusion state of the electrolyzer was divided into training set, validation set and test set, and the linear regression model was trained by ordinary least squares linear regression training method. Input the multi-source fused state data of the electrolytic cell into the trained linear regression model and output the compensation amount of the multi-source fused state data of the electrolytic cell; Based on the compensation amount of multi-source fusion state data of the electrolyzer, and combined with the operation and control principle of the electrolyzer, a compensation strategy is formulated, and the multi-source fusion state data of the electrolyzer is executed for compensation. The compensation process includes analyzing the operational deviation reflected by the compensation amount of the multi-source fusion state data of the electrolyzer, combining the internal chemical reaction process, physical characteristics and the mutual influence between operating parameters, identifying the specific circumstances of the current operating state of the electrolyzer deviating from the normal range, and formulating a compensation strategy based on the analysis results. The current controller is adjusted according to the compensation strategy to change the current in the electrolytic cell, thereby adjusting the reaction rate and operating status of the electrolytic cell. The changes in the electrolyzer's operating parameters during the compensation process are monitored in real time by sensors and controllers to evaluate the compensation effect. The compensation strategy is then adjusted based on the feedback information until the multi-source fusion state data of the electrolyzer is restored to a reasonable range. A random forest training method was used to train the isolated forest model using compensated multi-source fused state historical data from the electrolyzer, resulting in a short-circuit detection model. The specific steps are as follows. The compensated historical data of the multi-source fusion state of the electrolyzer is divided into training set, validation set and test set. The training set is used to train an isolated forest model to detect short circuit anomalies in a random forest manner. The method of training an isolated forest model in a random forest manner includes: randomly selecting a subset of samples with replacement from the training set, constructing a decision tree based on each subset of samples, selecting features that are highly correlated with short-circuit anomalies, and processing missing and outlier values. Decision trees are constructed by randomly selecting splitting features and node splitting points. The judgment results of each decision tree are combined, and the processed data is mapped to the isolated forest feature space. In the feature space of the isolated forest, features and partition values ​​are randomly selected to construct an isolated tree, and the degree of anomaly is judged based on the path length of data points in the isolated tree. The performance of the isolated forest model was evaluated using the validation set, and the parameters were adjusted to optimize the isolated forest model. The optimized model was then evaluated using the test set to obtain the short-circuit detection model. The compensated multi-source fusion state data of the electrolyzer is input into the short-circuit detection model, and the short-circuit detection results of the electrolyzer are output.

2. The intelligent fault detection method for electrolysis systems based on explosion-proof robot dogs as described in claim 1, characterized in that: The specific steps for using an explosion-proof robotic dog to collect real-time environmental data and operational status data of the electrolytic cell are as follows. An explosion-proof robotic dog collects real-time electrolyte temperature and gas concentration values ​​to obtain electrolytic cell environmental data, and also collects real-time current, voltage, and electrolyte circulation flow rates to obtain electrolytic cell operating status data.

3. The intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in claim 2, characterized in that: The method employs Kalman filtering and timestamp synchronization to perform filtering and noise reduction processing and convert the data into a unified timestamp. The specific steps are as follows. The electrolytic cell environmental data and electrolytic cell operating status data are combined into a state vector, and the Kalman filter matrix and noise parameters are defined by analyzing the characteristics of the electrolytic cell. Based on the Kalman filter matrix, noise parameters, and state vector, the filtered and denoised electrolytic cell environmental data and electrolytic cell operating status data are obtained through Kalman filter calculation. Extract the timestamp information of the electrolytic cell environment data and the electrolytic cell operating status data, calculate the timestamp deviation, and use the timestamp of the electrolytic cell environment data as a reference to align the timestamp of the electrolytic cell operating status data using linear interpolation.

4. The intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in claim 3, characterized in that: The process involves fusing the processed electrolytic cell environmental data and electrolytic cell operating status data using multivariate statistical methods to obtain multi-source fused state data of the electrolytic cell. The specific steps are as follows. The electrolytic cell environmental data and electrolytic cell operating status data after Kalman filtering and timestamp synchronization are standardized using the Z-score method, and key features for short circuit detection are extracted. Principal component analysis fusion method is used to extract key information and reduce dimensionality, and then fuse it with standardized electrolytic cell environmental data and electrolytic cell operating status data to obtain short circuit detection related information. The data is then sorted and normalized by feature scaling method to obtain multi-source fused state data of electrolytic cells.

5. The intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in claim 1, characterized in that: The steps for inputting the compensated multi-source fused state data of the electrolyzer into the short-circuit detection model and outputting the short-circuit detection result of the electrolyzer are as follows: The compensated multi-source fused state data of the electrolytic cell is input into the trained short-circuit detection model; After receiving the compensated multi-source fusion state data of the electrolytic cell, the short-circuit detection model extracts key short-circuit features and performs normalization processing. It then uses its short-circuit detection and judgment capabilities to analyze the data and output the short-circuit detection results of the electrolytic cell.

6. An intelligent fault detection system for an electrolysis system based on an explosion-proof robot dog, based on the intelligent fault detection method for an electrolysis system based on an explosion-proof robot dog as described in any one of claims 1 to 5, characterized in that: It includes a data acquisition module, a data fusion module, a model training module, and a result output module; The data acquisition module is used to collect real-time environmental data and operating status data of the electrolytic cell through the sensors of the explosion-proof robot dog. It uses Kalman filtering and timestamp synchronization methods to perform filtering and noise reduction processing and convert the data into a unified timestamp. The data fusion module is used to fuse the processed electrolytic cell environmental data and electrolytic cell operating status data through multivariate statistical methods to obtain multi-source fused state data of the electrolytic cell, and to perform real-time compensation using a linear regression model; The model training module is used to train the isolated forest model using the random forest training method and the compensated multi-source fusion state historical data of the electrolyzer to obtain the short circuit detection model. The result output module is used to input the compensated multi-source fusion state data of the electrolyzer into the short-circuit detection model and output the short-circuit detection results of the electrolyzer.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the intelligent fault detection method for electrolysis systems based on an explosion-proof robot dog as described in any one of claims 1 to 5.

Images