Diesel Generator Set Fault Prediction Method, Model, and Storage Medium

Abstract

The present invention discloses a method, model, and storage medium for fault prediction of a diesel generator set. The method includes obtaining operation data and processing it to obtain a data sample set; mapping the data sample set to obtain an extended data sample set; calculating based on the extended data sample set to obtain Euclidean distances; extracting local features according to the Euclidean distances and related functions; outputting based on the local features to obtain a first category and a first possibility; converting according to the first category and the first possibility to obtain a second fault category and a second fault occurrence possibility; outputting according to the second fault category and the second fault occurrence possibility to obtain a predicted value; analyzing the predicted value and an expected value to obtain a first parameter; performing optimization calculation according to the first parameter to obtain a second parameter; optimizing the prediction model according to the second parameter and then outputting to obtain a fault prediction result. This method can obtain optimal parameters through model iteration and update, improving the accuracy of fault prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of fault prediction, and particularly to a fault prediction method, model, and storage medium for diesel generator sets. Background Art

[0002] Due to their good adaptability and economy, diesel generator sets have been widely used in various industrial fields such as communication and industry. They can provide reliable power support in case of unstable or unavailable mains power, ensuring the normal operation of enterprises. However, various faults are inevitable during the operation of diesel generator sets, such as lubrication system problems, turbocharger failures, and fuel system abnormalities. These faults not only affect the production process of enterprises but may even cause significant economic losses. Therefore, fault prediction for diesel generator sets is of great significance, which can help maintenance personnel take preventive measures in advance, reduce losses caused by faults, and extend the service life of equipment.

[0003] In an existing technology, the specific implementation includes: sensors are distributed throughout the key parts of the unit to collect multi-dimensional operation data such as temperature and pressure. After aggregating into a data set, machine learning (such as decision tree algorithms constructing a tree-like structure based on data features to judge fault branches) and deep learning (neural networks calculating and transforming through input and hidden layers to obtain results at the output layer, CNN processing vibration image data to extract features, and RNN and its variants capturing the trends of time series data) algorithms are used. The model is trained with a large amount of historical data and parameters are adjusted to identify fault patterns. The trained model receives newly collected data, analyzes and judges, and then outputs the fault type and possibility, providing a reference for maintenance personnel to prevent faults.

[0004] However, the prediction results obtained in the existing technology are only the types of faults, and the output results are not further iteratively optimized, resulting in poor accuracy of the obtained prediction results. Summary of the Invention

[0005] The present invention provides a fault prediction method, model, and storage medium for diesel generator sets to solve the problem that the prediction results obtained in the existing technology are only the types of faults, and the output results are not further iteratively optimized, resulting in poor accuracy of the obtained prediction results.

[0006] In a first aspect, to solve the above technical problems, the present invention provides a fault prediction method for diesel generator sets, including:

[0007] Obtaining the operation data of the diesel generator set;

[0008] Performing normalization processing on the operation data of the diesel generator set to obtain a sample set of the operation data of the diesel generator set;

[0009] Perform high-dimensional mapping on the diesel generator set operation data sample set to obtain an extended operation data sample set;

[0010] Calculate the Euclidean distance based on the extended operation data sample set;

[0011] Extract features based on the Euclidean distance and the pre-stored fault location occurrence function to obtain local features;

[0012] Output through a model based on the local features to obtain the first fault category and the first fault occurrence probability;

[0013] Perform conversion based on the first fault category and the first fault occurrence probability to obtain the second fault category and the second fault occurrence probability;

[0014] Perform set output based on the second fault category and the second fault occurrence probability to obtain a predicted output value;

[0015] Perform error analysis based on the predicted output value and the pre-stored expected output value to obtain the first fault prediction model parameters;

[0016] Perform algorithm iteration optimization based on the first fault prediction model parameters to obtain the second fault prediction model parameters;

[0017] Adjust and optimize the pre-stored fault prediction model based on the second fault prediction model parameters and then output to obtain the fault prediction result.

[0018] In an implementable manner of the first aspect, the normalization process of the diesel generator set operation data to obtain the diesel generator set operation data sample set includes:

[0019] Extract features from the diesel generator set operation data to obtain attribute values and feature vectors;

[0020] Calculate the mean and standard deviation of the attribute values to obtain the attribute value mean and the attribute value standard deviation;

[0021] Perform integrated calculation based on the attribute values and the feature vectors to obtain operation data samples;

[0022] Perform normalization processing on the operation data samples, the attribute value mean, and the attribute value standard deviation to obtain the diesel generator set operation data sample set.

[0023] In an implementable manner of the first aspect, the integrated calculation based on the attribute values and the feature vectors to obtain operation data samples includes:

[0024] The operating data sample is calculated through the following formula:

[0025]

[0026] Wherein, represents the operating data sample, represents the th attribute value of the th data sample in the set of operating data samples of the diesel generator set, represents the th eigenvector of the th attribute value, represents the th eigenvector of the th attribute value, is the parameter for controlling the influence range of the neighborhood,

[0027] In one possible implementation of the first aspect, the feature extraction is performed according to the Euclidean distance and the pre-stored fault location occurrence function to obtain local features, including:

[0028] The local features are calculated through the following formula:

[0029]

[0030] Wherein, represents the local feature, is the Euclidean distance between the sample and the sample in the extended operating data sample set, is the Euclidean distance between the sample and the sample corresponding to the occurrence location of the th type of fault in the extended operating data sample set, represents the pre-stored fault occurrence location function, is the th fault category of the th extended operating data sample, is the pre-stored number of fault categories, is the index of the extended operating data sample set,

[0031] In one possible implementation of the first aspect, the conversion according to the first fault category and the first fault occurrence possibility to obtain the second fault category and the second fault occurrence possibility includes:

[0032] The second fault category and the second fault occurrence probability are calculated through the following formula:

[0033] C

[0034]

[0035] where C represents the second fault category, represents the second fault occurrence probability, represents the number of extended operation data samples in the pre-stored extended operation data sample set, represents the th extended operation data sample corresponding to the first fault occurrence probability of the th type of fault, represents the th extended operation data sample corresponding to the first fault category of the th type of fault, represents the number of pre-stored fault categories, represents the th extended operation data sample having the first fault occurrence probability of the th type of fault, represents the th extended operation data sample corresponding to the first fault category of the th type of fault, represents the fault category serial number, represents the extended operation data sample serial number.

[0036] In an implementable manner of the first aspect, the error analysis is performed on the predicted output value and the pre-stored expected output value to obtain the first fault prediction model parameters, including:

[0037] The first fault prediction model parameters are calculated through the following formula:

[0038]

[0039] where, represents the first fault prediction model parameters, represents the number of pre-stored fault categories, represents the number of extended operation data samples in the pre-stored extended operation data sample set, represents the predicted output value, represents the pre-stored expected output value, represents the fault category serial number, represents the extended operation data sample serial number.

[0040] In an implementable manner of the first aspect, the algorithm iteration optimization according to the first fault prediction model parameters to obtain the second fault prediction model parameters includes:

[0041] Using the gradient descent method to iterate according to the first fault prediction model parameters to obtain updated fault prediction model parameters; until the error between the predicted output value and the pre-stored expected output value is less than a preset threshold, terminate the iteration, and use the updated fault prediction model parameters at this time as the second fault prediction model parameters for output.

[0042] In the second aspect, the present invention provides a fault prediction model for a diesel generator set, including:

[0043] A data acquisition module for acquiring the operating data of the diesel generator set;

[0044] A data processing module for performing normalization processing on the operating data of the diesel generator set to obtain a set of operating data samples of the diesel generator set;

[0045] A high-dimensional mapping module for performing high-dimensional mapping on the set of operating data samples of the diesel generator set to obtain an extended set of operating data samples;

[0046] A distance calculation module for calculating the Euclidean distance according to the extended set of operating data samples;

[0047] A feature extraction module for extracting local features according to the Euclidean distance and a pre-stored fault location occurrence function;

[0048] A feature analysis module for outputting through a model according to the local features to obtain a first fault category and a first fault occurrence probability;

[0049] A data conversion module for converting according to the first fault category and the first fault occurrence probability to obtain a second fault category and a second fault occurrence probability;

[0050] A data output module for performing set output according to the second fault category and the second fault occurrence probability to obtain a predicted output value;

[0051] An error analysis module for performing error analysis on the predicted output value and a pre-stored expected output value to obtain the first fault prediction model parameters;

[0052] A parameter iteration module for performing algorithm iteration optimization according to the first fault prediction model parameters to obtain the second fault prediction model parameters;

[0053] The result output module is used to adjust and optimize the pre-stored fault prediction model according to the second fault prediction model parameters and then output to obtain the fault prediction result.

[0054] In an implementable manner of the second aspect, the normalization processing of the diesel generator set operation data to obtain the diesel generator set operation data sample set includes:

[0055] Feature extraction is performed on the diesel generator set operation data to obtain an attribute value and a feature vector;

[0056] The mean and variance of the attribute values are calculated to obtain the attribute value mean and the attribute value standard deviation;

[0057] Integrated calculation is performed on the attribute value and the feature vector to obtain an operation data sample;

[0058] Normalization processing is performed on the operation data sample, the attribute value mean, and the attribute value standard deviation to obtain the diesel generator set operation data sample set.

[0059] In an implementable manner of the second aspect, the integrated calculation of the attribute value and the feature vector to obtain an operation data sample includes:

[0060] The operation data sample is calculated by the following formula:

[0061]

[0062] Among them, represents the operation data sample, represents the th attribute value of the th data sample in the diesel generator set operation data sample set, represents the th feature vector of the attribute value, represents the th feature vector of the attribute value, is the control neighborhood influence range parameter, is the number of neighboring nodes with the highest similarity selected from the data samples in the diesel generator set operation data sample set, represents the Euclidean distance between the feature vectors.

[0063] In an implementable manner of the second aspect, the feature extraction based on the Euclidean distance and the pre-stored fault location occurrence function to obtain local features includes:

[0064] The local features are calculated by the following formula:

[0065]

[0066] Among them, represents a local feature, is the Euclidean distance between the sample in the extended operation data sample set and the sample ; is the Euclidean distance between the sample in the extended operation data sample set and the sample corresponding to the position of the th type of fault occurrence in the extended operation data sample set ; represents a pre-stored fault occurrence position function, is the th fault category of the th extended operation data sample, is the index of the extended operation data sample set, is the weight coefficient.

[0067] In an implementable manner of the second aspect, the obtaining of the second fault category and the second fault occurrence possibility by converting according to the first fault category and the first fault occurrence possibility includes:

[0068] Calculating the second fault category and the second fault occurrence possibility through the following formula:

[0069] C

[0070]

[0071] Among them, C represents the second fault category, represents the second fault occurrence possibility, represents the number of extended operation data samples in the pre-stored extended operation data sample set, is the th first fault occurrence possibility corresponding to the th type of fault of the th extended operation data sample, is the th first fault category corresponding to the th type of fault of the th extended operation data sample, represents the first fault occurrence possibility that the th type of fault occurs in the th extended operation data sample, is the th first fault category corresponding to the Indicates the serial number of the fault category, Indicates the serial number of the extended operation data sample.

[0072] In an implementable manner of the second aspect, the error analysis is performed based on the predicted output value and the pre-stored expected output value to obtain the first fault prediction model parameter, including:

[0073] The first fault prediction model parameter is calculated by the following formula:

[0074]

[0075] Wherein, Indicates the first fault prediction model parameter, Indicates the number of pre-stored fault categories, Indicates the number of extended operation data samples in the pre-stored extended operation data sample set, Indicates the predicted output value, Indicates the pre-stored expected output value, Indicates the serial number of the fault category, Indicates the serial number of the extended operation data sample.

[0076] In an implementable manner of the second aspect, the algorithm iteration optimization is performed based on the first fault prediction model parameter to obtain the second fault prediction model parameter, including:

[0077] Iterate according to the first fault prediction model parameter using the gradient descent method to obtain the updated fault prediction model parameter; until the error between the predicted output value and the pre-stored expected output value is less than the preset threshold, terminate the iteration, and output the updated fault prediction model parameter at this time as the second fault prediction model parameter.

[0078] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the diesel generator set fault prediction method described in any one of the above is implemented.

[0079] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program. Wherein, when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the diesel generator set fault prediction method described in any one of the above.

[0080] Compared with the prior art, the present invention has the following beneficial effects:

[0081] The present invention discloses a fault prediction method for a diesel generator set, including: acquiring the operation data of the diesel generator set; performing normalization processing on the operation data of the diesel generator set to obtain a set of operation data samples of the diesel generator set; performing high-dimensional mapping on the set of operation data samples of the diesel generator set to obtain an extended set of operation data samples; calculating distances based on the extended set of operation data samples to obtain Euclidean distances; extracting local features according to the Euclidean distances and a pre-stored fault location occurrence function; outputting through a model based on the local features to obtain a first fault category and a first probability of fault occurrence; performing conversion according to the first fault category and the first probability of fault occurrence to obtain a second fault category and a second probability of fault occurrence; performing set output according to the second fault category and the second probability of fault occurrence to obtain a predicted output value; performing error analysis on the predicted output value and a pre-stored expected output value to obtain first fault prediction model parameters; and performing algorithm iteration optimization according to the first fault prediction model parameters to obtain second fault prediction model parameters. The present invention acquires various operation data of the diesel generator set (such as load data, fuel flow rate, compressor air flow rate, diesel engine speed, diesel engine operating conditions, ambient temperature) through an information acquisition system; maps these original operation data samples to a higher-dimensional feature space to obtain an extended set of operation data samples, and extracts local features based on this extended set; constructs a fault prediction model according to the local features, and this model outputs the fault category, the probability of fault occurrence, and the corresponding local features of each extended operation data sample; uses the stochastic gradient descent method to minimize the sum of squares of the error between the actual output and the expected output of the fault prediction model, optimizes the model parameters, and further improves the prediction accuracy by iteratively updating the parameters of the fault prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 is a schematic flowchart of the fault prediction method for a diesel generator set provided in the first embodiment of the present invention;

[0083] Figure 2 is a schematic structural diagram of the fault prediction model for a diesel generator set provided in the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0084] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0085] Refer to Figure 1, the first embodiment of the present invention provides a method for predicting faults in a diesel generator set, including the following steps:

[0086] S1, Obtain the operating data of the diesel generator set;

[0087] S2, Perform normalization processing on the operating data of the diesel generator set to obtain a set of operating data samples of the diesel generator set;

[0088] S3, Perform high-dimensional mapping on the set of operating data samples of the diesel generator set to obtain an extended set of operating data samples;

[0089] S4, Calculate the distance based on the extended set of operating data samples to obtain the Euclidean distance;

[0090] S5, Extract features based on the Euclidean distance and the pre-stored fault location occurrence function to obtain local features;

[0091] S6, Output through a model based on the local features to obtain the first fault category and the first fault occurrence probability;

[0092] S7, Perform conversion based on the first fault category and the first fault occurrence probability to obtain the second fault category and the second fault occurrence probability;

[0093] S8, Perform set output based on the second fault category and the second fault occurrence probability to obtain a predicted output value;

[0094] S9, Perform error analysis on the predicted output value and the pre-stored expected output value to obtain the first fault prediction model parameters;

[0095] S10, Perform algorithm iteration optimization based on the first fault prediction model parameters to obtain the second fault prediction model parameters;

[0096] S11, Adjust and optimize the pre-stored fault prediction model based on the second fault prediction model parameters and then output to obtain the fault prediction result.

[0097] In step S1, it is necessary to obtain the operating data of the diesel generator set.

[0098] Exemplarily, the operating data of the diesel generator set includes: the generator voltage of the unit, the generator frequency, and the power output data. The generator parameters include: the rated generator voltage, the rated power generation frequency, the rated power, etc. The acquisition method is specifically obtained by using sensors to track the long-term operation of the diesel generator set and collecting and storing the relevant data during the operation process.

[0099] In step S2, the operating data of the diesel generator set is normalized to obtain a set of operating data samples of the diesel generator set.

[0100] In the above step S2, the operation of normalizing the operating data of the diesel generator set to obtain a set of operating data samples of the diesel generator set specifically further includes the following steps:

[0101] S21. Feature extraction is performed on the operating data of the diesel generator set to obtain attribute values and feature vectors.

[0102] In an implementable manner, the feature extraction from the operating data of the diesel generator set to obtain attribute values and feature vectors includes: for each operating data sample in the operating data of the diesel generator set, starting from its high-dimensional feature projection point, constructing approximate projection points mapped to a low-dimensional space according to the feature dimension form of the sample and the mapping function, and calculating the distances from these approximate projection points to several cycles before the fault; by calculating the Euclidean distance and sorting it, selecting the fault cycles corresponding to the first certain number of minimum distances as local features, and these local features and the original features of the sample together constitute the feature vector, and the specific numerical value on each feature dimension is the attribute value.

[0103] S22. Mean and variance calculations are performed on the attribute values to obtain an attribute value mean and an attribute value standard deviation.

[0104] Exemplarily, the sum of the attribute values of each sample is divided by the number of samples to obtain the attribute value mean; the attribute value standard deviation can be obtained by summing the differences between the attribute values of each sample and the attribute value mean.

[0105] S23. Integration calculations are performed on the attribute values and the feature vectors to obtain operating data samples.

[0106] It should be noted that the operating data samples are calculated through the following formula:

[0107]

[0108] Among them, represents the operating data sample, represents the th attribute value of the th data sample in the set of operating data samples of the diesel generator set, represents the th feature vector of the The number of neighboring nodes with the highest similarity selected from the data samples in the diesel generator set operation data sample set represents the Euclidean distance between feature vectors.

[0109] The formula used in this embodiment is a data expansion method based on local information. By using neighborhood information and distance weighting, it can enhance the data expression ability to a certain extent. For data processing of complex systems (such as diesel generator sets), it helps to improve the accuracy and reliability of subsequent data analysis and fault prediction. By expanding the original data using neighborhood information, the local structure and features of the data can be better captured. When calculating the operation data samples, the formula will find the original samples of the most similar neighboring nodes according to the Euclidean distance. Exemplarily, for the original sample set , when calculating the expanded data of a certain sample , the most similar samples to will be found from the set. These neighboring nodes will participate in the subsequent weighted calculation to generate the operation data samples.

[0110] In this embodiment, the parameters in the control neighborhood influence range parameter need to be adjusted through experiments (such as cross-validation) to find the values most suitable for the specific data set and task, so as to optimize the performance of the algorithm. Exemplarily, in fault prediction, by adjusting this parameter, the model can better utilize the local sample information when predicting the faults of diesel generator sets, improving the prediction accuracy.

[0111] In a realizable manner, the process of obtaining the number of neighboring nodes with the highest similarity selected from the data samples in the diesel generator set operation data sample set includes: calculating a matrix , where represents the similarity (or distance) between sample and ; calculating each element in matrix according to the selected similarity metric; for each sample , sorting the th row (or the th column) of in descending (or ascending) order according to the similarity (or distance), and selecting the sample indices corresponding to the first elements. These samples are the neighboring nodes with the highest similarity.

[0112] S24. Normalize according to the operation data sample, the mean value of the attribute value, and the standard deviation of the attribute value to obtain a set of operation data samples of the diesel generator set.

[0113] It should be noted that the set of operation data samples of the diesel generator set is calculated by the following formula:

[0114]

[0115] Wherein, represents the set of operation data samples of the diesel generator set, is the operation data sample, is the th average value of the attribute value, is the th standard deviation of the attribute value.

[0116] In step S3, perform high-dimensional mapping according to the set of operation data samples of the diesel generator set to obtain an extended set of operation data samples.

[0117] In an implementable manner, the performing high-dimensional mapping according to the set of operation data samples of the diesel generator set to obtain an extended set of operation data samples includes: preprocessing according to the set of operation data samples of the diesel generator set, including normalization processing and constructing a fault model to obtain a sample set; selecting a suitable mapping function, such as calculating projection points using a specific formula and determining the mapping function coefficients, and can also be iteratively optimized by the Lagrange multiplier method; performing a mapping operation on the set of operation data samples of the diesel generator set according to the mapping function, so as to obtain an extended set of operation data samples.

[0118] Exemplarily, use a normalization function to normalize the operation data of the diesel generator set to avoid the interference of abnormal points on the training process of the entire parameter space. For example, the data can be normalized to a specific interval, such as [0, 1] or [-1, 1], etc. The specific normalization method can be selected according to the characteristics of the data and the requirements of the model. After obtaining the normalized operation data, construct a fault model for all possible faults using expert experience. The fault model includes the fault state and several cycles before the fault, and the corresponding fault cycle characteristics are the original operation data samples. The set of operation data samples of the diesel generator set is:

[0119]

[0120] Wherein, represents the th fault cycle characteristic, represents the fault type corresponding to the fault cycle characteristic given according to expert experience, and expresses it in the form of where represents the number of characteristic features of the diesel generator set operation data sample set, represents the th original feature of the th extended operation data sample, and represents the projection point after mapping to the high-dimensional space, is the mapping function coefficient, is calculated as follows:

[0121]

[0122] where is the th original operation data sample feature in the same dimension as after conversion, is the th original operation data sample feature in the same dimension as after conversion, and is determined by calculating an expression related to the cosine similarity.

[0123] According to the determined mapping function , each sample in the diesel generator set operation data sample set is mapped to convert it from the low-dimensional space to a higher-dimensional feature space, thereby obtaining the extended operation data sample set.

[0124] In step S4, distance calculation is performed based on the extended operation data sample set to obtain the Euclidean distance.

[0125] In this embodiment, the Euclidean distance is used to measure the straight-line distance between two points in the Euclidean space. For two points (or vectors) and , their Euclidean distance is calculated as follows:

[0126]

[0127] Exemplarily, there are two two-dimensional vectors and , then their Euclidean distance is:

[0128]

[0129] In step S5, local features are extracted according to the Euclidean distance and a pre-stored fault location occurrence function.

[0130] It should be noted that the local features are calculated through the following formula:

[0131]

[0132] Where, represents the local feature, is the Euclidean distance between the sample in the extended operation data sample set and the sample , is the Euclidean distance between the sample in the extended operation data sample set and the sample corresponding to the fault occurrence location of the rd type of fault in the extended operation data sample set, , represents the pre-stored fault occurrence location function, is the th fault category of the th extended operation data sample, is the index of the extended operation data sample set, is the weight coefficient.

[0133] In this embodiment, the pre-stored fault occurrence location function refers to a function for determining the association relationship between the sample and the fault occurrence location. Exemplarily, if the data contains the operation data of different components, the fault occurrence location function is a function for judging the component or operation state to which the sample belongs according to the sample data characteristics. If the predicted fault type is "generator stator winding fault", then may judge whether the sample comes from the monitoring points or operation states related to the generator stator winding according to the characteristics such as voltage, current, and temperature in the sample. If so, , this sample is considered when calculating the local features of this fault type; if not, , this sample is excluded from this calculation.

[0134] In step S6, the first fault category and the first fault occurrence possibility are obtained through output by the model according to the local features.

[0135] In an implementable manner, the output by the model based on the local features to obtain the first fault category and the first fault occurrence probability includes: inputting the extracted local features into a pre-constructed fault prediction model, and the internal parameters are continuously adjusted by optimization algorithms such as the stochastic gradient descent method to adapt to the feature patterns and fault rules of the data; after inputting the local features, the model calculates and infers based on the mapping relationship between the features it has learned and the faults; for the obtained output results, the first fault category is determined through a specific conversion method, and according to the probability value sizes corresponding to different fault categories in the output, the category with the largest probability is selected as the first fault category; and the first fault occurrence probability is the probability value corresponding to this category in the model output.

[0136] In step S7, conversion is performed according to the first fault category and the first fault occurrence probability to obtain the second fault category and the second fault occurrence probability.

[0137] It should be noted that the second fault category and the second fault occurrence probability are calculated through the following formula:

[0138] C

[0139]

[0140] where C represents the second fault category, represents the second fault occurrence probability, represents the number of extended operation data samples in the pre-stored extended operation data sample set, represents the th extended operation data sample corresponding to the first fault occurrence probability of the th type of fault, represents the th extended operation data sample corresponding to the first fault category of the th type of fault, represents the number of pre-stored fault categories, represents the th extended operation data sample corresponding to the first fault occurrence probability of the th type of fault, represents the th extended operation data sample corresponding to the first fault category of the th type of fault, represents the fault category serial number, represents the extended operation data sample serial number.

[0141] In this embodiment, the second fault category and the second fault occurrence probability refer to the numerical values obtained after converting the fault category and fault occurrence probability directly output by the model through a certain form.

[0142] In step S8, a set output is performed according to the second fault category and the second fault occurrence probability to obtain a predicted output value.

[0143] Exemplarily, according to the second fault category and the second fault, the fault type identifier with a relatively high probability is obtained through a series of data processing, model operations, and analysis. The second fault occurrence probability represents the degree of occurrence of the fault in the form of probability. When performing the set output, these two pieces of information are integrated to form a predicted output value.

[0144] In step S9, error analysis is performed on the predicted output value and the expected output value stored in advance to obtain the first fault prediction model parameter.

[0145] It should be noted that the first fault prediction model parameter is calculated through the following formula:

[0146]

[0147] Where, represents the first fault prediction model parameter, represents the number of fault categories stored in advance, represents the number of extended operation data samples in the extended operation data sample set stored in advance, represents the predicted output value, represents the expected output value stored in advance, represents the fault category serial number, represents the extended operation data sample serial number.

[0148] In this embodiment, the first fault prediction model parameter refers to the preliminary fault prediction model parameter obtained by analyzing the predicted output value and the expected output value stored in advance by a preset model without iterative update. The expected output value stored in advance refers to the relatively ideal output value output by constructing a fault model using expert experience.

[0149] S10. Perform algorithm iteration optimization according to the first fault prediction model parameter to obtain the second fault prediction model parameter.

[0150] Exemplarily, according to the first fault prediction model parameter, the gradient of the model parameter is calculated using the stochastic gradient descent method. The stochastic gradient descent method randomly selects a small batch of samples in each iteration, calculates the gradients of these samples, and updates the model parameters. The specific update formula is:

[0151]

[0152] wherein are model parameters, is the learning rate, is the error function, and is the gradient of the model parameters. In the present invention, the number of sample faults occurring in the th sample is obtained from historical data, and the optimal solution of the iterative mapping function and the fault type of the th fault of the historical sample are input into the logistic regression model, the output of the logistic regression model is calculated, and the gradient descent parameter is obtained based on the output and the number of historical sample faults occurring, and then the model parameters are updated. The above process is continuously repeated, and the parameters of the fault prediction model are iteratively updated in each training cycle until the model converges. The judgment of convergence can be through monitoring indicators such as the loss function value or accuracy rate on the validation set. When these indicators no longer change significantly, it is considered that the model has converged, and thus the second fault prediction model parameters are obtained.

[0153] In step S11, the pre-stored fault prediction model is adjusted and optimized according to the second fault prediction model parameters and then output to obtain a fault prediction result.

[0154] In a specific embodiment, the weights, biases or other key structures inside the model are adjusted according to the second fault prediction model parameters. Exemplarily, for a model based on a neural network, the weight values of the connections between neurons are updated; the adjusted fault prediction model is used to process and analyze the input data, and these input data are the operation data of the diesel generator set after preprocessing and feature extraction; the model will calculate the possibility of each fault category occurring and related feature information according to its internal algorithm and new parameter settings. Finally, according to specific rules or thresholds, the fault category with the highest possibility is selected from the output of the model as the predicted fault type, and the final fault prediction result is determined from the corresponding confidence level or other relevant information.

[0155] In summary, the present invention discloses a method for predicting faults in a diesel generator set, including: obtaining the operating data of the diesel generator set; performing normalization processing on the operating data of the diesel generator set to obtain a set of operating data samples of the diesel generator set; performing high-dimensional mapping on the set of operating data samples of the diesel generator set to obtain an extended set of operating data samples; calculating the distance based on the extended set of operating data samples to obtain the Euclidean distance; extracting local features based on the Euclidean distance and a pre-stored fault location occurrence function; outputting through a model based on the local features to obtain a first fault category and a first probability of fault occurrence; performing conversion based on the first fault category and the first probability of fault occurrence to obtain a second fault category and a second probability of fault occurrence; performing set output based on the second fault category and the second probability of fault occurrence to obtain a predicted output value; performing error analysis on the predicted output value and a pre-stored expected output value to obtain first fault prediction model parameters; and performing algorithm iteration optimization based on the first fault prediction model parameters to obtain second fault prediction model parameters. The present invention obtains various operating data of the diesel generator set (such as load data, fuel flow rate, compressor air flow rate, diesel engine speed, diesel engine operating conditions, ambient temperature) through an information acquisition system; maps these original operating data samples to a higher-dimensional feature space to obtain an extended set of operating data samples, and extracts local features based on this extended set; constructs a fault prediction model based on the local features, and this model outputs the fault category, the probability of fault occurrence, and the corresponding local features of each extended operating data sample; uses the stochastic gradient descent method to minimize the sum of the squares of the errors between the actual output and the expected output of the fault prediction model, optimizes the model parameters, and further improves the prediction accuracy by iteratively updating the parameters of the fault prediction model.

[0156] Referring to Figure 2 , the second embodiment of the present invention provides a fault prediction model for a diesel generator set, including:

[0157] A data acquisition module 101, configured to obtain the operating data of the diesel generator set;

[0158] A data processing module 102, configured to perform normalization processing on the operating data of the diesel generator set to obtain a set of operating data samples of the diesel generator set;

[0159] A high-dimensional mapping module 103, configured to perform high-dimensional mapping on the set of operating data samples of the diesel generator set to obtain an extended set of operating data samples;

[0160] A distance calculation module 104, configured to calculate the distance based on the extended set of operating data samples to obtain the Euclidean distance;

[0161] The feature extraction module 105 is configured to perform feature extraction according to the Euclidean distance and a pre-stored fault location occurrence function to obtain local features;

[0162] The feature analysis module 106 is configured to output through a model according to the local features to obtain a first fault category and a first fault occurrence probability;

[0163] The data conversion module 107 is configured to perform conversion according to the first fault category and the first fault occurrence probability to obtain a second fault category and a second fault occurrence probability;

[0164] The data output module 108 is configured to perform set output according to the second fault category and the second fault occurrence probability to obtain a predicted output value;

[0165] The error analysis module 109 is configured to perform error analysis according to the predicted output value and a pre-stored expected output value to obtain first fault prediction model parameters;

[0166] The parameter iteration module 110 is configured to perform algorithm iteration optimization according to the first fault prediction model parameters to obtain second fault prediction model parameters.

[0167] The result output module 111 is configured to adjust and optimize a pre-stored fault prediction model according to the second fault prediction model parameters and then output to obtain a fault prediction result.

[0168] It should be noted that a diesel generator set fault prediction model provided by an embodiment of the present invention is used to execute all process steps of a diesel generator set fault prediction method in the above embodiment, and the working principles and beneficial effects of the two correspond one by one, so they will not be elaborated here.

[0169] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a diesel generator set fault prediction program. When the processor executes the computer program, it implements the steps in each of the above embodiments of the diesel generator set fault prediction method, such as Figure 1 Step S1 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in each of the above device embodiments, such as the data acquisition module.

[0170] Exemplarily, the computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the electronic device.

[0171] The electronic device may be a computing device such as a desktop computer, a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine certain components, or different components. For example, the electronic device may further include input / output devices, network access devices, a bus, etc.

[0172] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects various parts of the entire electronic device through various interfaces and circuits.

[0173] The memory can be used to store the computer programs and / or modules. By running or executing the computer programs and / or modules stored in the memory, and invoking the data stored in the memory, the processor can implement various functions of the electronic device. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices.

[0174] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-described various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0175] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0176] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. In particular, for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A diesel generator set fault prediction method, characterized in that: Executed by a computer, including: Obtain diesel generator set operation data; Performing normalization processing on the diesel generator set operation data to obtain a diesel generator set operation data sample set; Performing high-dimensional mapping on the diesel generator set operation data sample set to obtain an extended operation data sample set; Performing distance calculation based on the extended running data sample set to obtain a Euclidean distance; Extracting features according to the Euclidean distance and a pre-stored fault location occurrence function to obtain local features; Outputting the local features through a model to obtain a first fault category and a first fault occurrence probability; Converting the first fault category and the first fault occurrence probability to obtain a second fault category and a second fault occurrence probability; Performing a collective output according to the second fault category and the second fault occurrence probability to obtain a predicted output value; Performing error analysis based on the predicted output value and the pre-stored expected output value to obtain a first fault prediction model parameter; Performing algorithm iterative optimization according to the first fault prediction model parameters to obtain second fault prediction model parameters; According to the second fault prediction model parameters, the pre-stored fault prediction model is adjusted and optimized and then output to obtain a fault prediction result; The extracting of features according to the Euclidean distance and the pre-stored fault location occurrence function to obtain local features includes: The local features are calculated by the following formula: in, Represents local features, Samples in the extended run data sample set With sample The Euclidean distance, Samples in the extended run data sample set The extended run data sample set Samples corresponding to the location where the class fault occurs The Euclidean distance, Indicates the pre-stored fault location function, For the Fault categories of extended operation data samples, is the number of pre-stored fault categories, The index of the sample set of data for the extended run, is the weight coefficient; The converting according to the first fault category and the first fault occurrence possibility to obtain the second fault category and the second fault occurrence possibility includes: The second fault category and the second fault occurrence probability are calculated using the following formula: in, Indicates the second fault category, Indicates the possibility of the second fault occurring. Indicates the number of extended operation data samples in the pre-stored extended operation data sample set, Indicates The extended operation data sample corresponds to The probability of the first fault of the type of fault, Indicates The extended operation data sample corresponds to The first fault category of the class fault, Indicates the number of pre-stored fault categories, Indicates The extended run data sample occurs The probability of the first fault of the type of fault, Indicates The extended operation data sample corresponds to The first fault category of the class fault, Indicates the fault category number, Indicates the extended running data sample number.

2. The diesel generator set fault prediction method according to claim 1, characterized in that: The normalization process is performed on the diesel generator set operation data to obtain a diesel generator set operation data sample set, including: Perform feature extraction based on the diesel generator set operation data to obtain attribute values ​​and feature vectors; Calculate the mean and variance of the attribute value to obtain the attribute value mean and the attribute value standard deviation; Performing integrated calculation based on the attribute value and the feature vector to obtain an operation data sample; Normalization processing is performed according to the operation data samples, the mean value of the attribute values ​​and the standard deviation of the attribute values ​​to obtain a set of operation data samples of the diesel generator set.

3. The diesel generator set fault prediction method according to claim 2, characterized in that: The step of performing integrated calculation based on the attribute value and the feature vector to obtain the operation data sample includes: The running data sample is calculated by the following formula: in, Indicates the running data sample, Indicates the number of samples in the diesel generator set operation data set. The data sample attribute values, Indicates The feature vector of attribute values, Indicates The feature vector of attribute values, To control the neighborhood influence range parameters, The number of neighboring nodes with the highest similarity selected for the data samples in the diesel generator set operation data sample set, Represents the Euclidean distance between feature vectors.

4. The diesel generator set fault prediction method according to claim 1, characterized in that: The performing error analysis based on the predicted output value and the pre-stored expected output value to obtain the first fault prediction model parameter includes: The first fault prediction model parameters are calculated by the following formula: in, represents the first fault prediction model parameter, Indicates the number of pre-stored fault categories, Indicates the number of extended operation data samples in the pre-stored extended operation data sample set, Represents the predicted output value, Indicates the expected output value stored in advance, Indicates the fault category number, Indicates the extended running data sample number.

5. The diesel generator set fault prediction method according to claim 1, characterized in that: The iterative optimization of the algorithm according to the first fault prediction model parameters to obtain the second fault prediction model parameters includes: Iterate using the gradient descent method according to the first fault prediction model parameters to obtain updated fault prediction model parameters; until the error between the predicted output value and the pre-stored expected output value is less than a preset threshold, terminate the iteration, and output the updated fault prediction model parameters at this time as the second fault prediction model parameters.

6. A diesel generator set fault prediction model, characterized in that: A method for predicting a fault of a diesel generator set according to any one of claims 1 to 5, comprising: A data acquisition module is used to acquire the operating data of the diesel generator set; A data processing module, used for performing normalization processing on the diesel generator set operation data to obtain a diesel generator set operation data sample set; A high-dimensional mapping module, used for performing high-dimensional mapping according to the diesel generator set operation data sample set to obtain an extended operation data sample set; A distance calculation module, used for performing distance calculation according to the extended running data sample set to obtain a Euclidean distance; A feature extraction module, used for extracting features according to the Euclidean distance and a pre-stored fault location occurrence function to obtain local features; A feature analysis module, configured to obtain a first fault category and a possibility of occurrence of a first fault by outputting the local features through a model; A data conversion module, configured to convert the first fault category and the first fault occurrence possibility to obtain a second fault category and a second fault occurrence possibility; A data output module, configured to perform a collective output according to the second fault category and the probability of occurrence of the second fault to obtain a predicted output value; an error analysis module, used to perform error analysis based on the predicted output value and the pre-stored expected output value to obtain a first fault prediction model parameter; A parameter iteration module, used to perform algorithm iterative optimization according to the first fault prediction model parameters to obtain second fault prediction model parameters; The result output module is used to adjust and optimize the pre-stored fault prediction model according to the second fault prediction model parameters and then output the result to obtain the fault prediction result.

7. An electronic device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the diesel generator set fault prediction method as claimed in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the diesel generator set fault prediction method according to any one of claims 1 to 5.

Images