Charging equipment fault diagnosis method and system based on semantic fusion of multi-source heterogeneous data

Abstract

A charging equipment fault diagnosis method and system based on semantic fusion of multi-source heterogeneous data. The method includes obtaining charging equipment fault text data and fault image data, using pre-trained fault keyword recognition model and fault image recognition model respectively, and identifying fault features of multiple dimensions from the charging equipment fault text data and fault image data in turn; based on the extracted fault features of multiple dimensions, constructing a model-level multi-level long short-term memory ML_LSTM feature information fusion model to fuse the fault features; based on the fused fault features, diagnosing the equipment fault through the Softmax fault diagnosis model to obtain the fault diagnosis type; based on the fault diagnosis type, constructing a hierarchical recommendation model to output the fault location and operation and maintenance strategy. The solution of the present invention improves the accuracy of fault diagnosis.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent diagnosis and operation and maintenance of charging equipment faults, and in particular relates to a charging equipment fault diagnosis method and system based on semantic fusion of multi-source heterogeneous data. Background Art

[0002] With the large-scale increase in electric vehicles, there is a large demand for electricity to meet the mileage, and the corresponding charging infrastructure has been widely used. The rapid development of charging infrastructure has also brought a series of safety operation problems, especially some safety accidents have occurred during the charging process of electric vehicles. The calibration work requires a large number of charging piles, the inspection work is stressful, and the operation and maintenance costs have also increased. The traditional regular inspection method of electric vehicle charging equipment is limited by multiple resources such as geographical location, operation and maintenance personnel, and time, resulting in equipment failures that cannot be solved in time. However, the current electric vehicle charging and discharging fault diagnosis and safe operation and maintenance service system is still imperfect, resulting in low efficiency in charging and discharging fault diagnosis, and a lack of effective fault operation and maintenance strategies to efficiently assist the maintenance personnel in their maintenance work.

[0003] The shortcomings of the existing solutions for fault diagnosis and maintenance of charging equipment are that, first, they often rely on a single or limited data source, resulting in insufficient accuracy of fault diagnosis. Second, they lack effective integration of multi-source heterogeneous data and cannot fully reflect the status of the equipment. At the same time, most methods are still passive responses and lack the ability to predict fault trends, and cannot achieve active prevention and efficient operation and maintenance. Therefore, how to establish an efficient fault diagnosis and operation and maintenance model has become a problem that needs to be solved urgently. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a charging equipment fault diagnosis method and system based on semantic fusion of multi-source heterogeneous data to solve the technical problem of improving the accuracy of fault diagnosis.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions.

[0006] The present invention first discloses a charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data, which includes the following steps:

[0007] Step 1: Obtain the charging equipment fault text data and fault image data, and use the pre-trained fault keyword recognition model and fault image recognition model to identify fault features of multiple dimensions from the charging equipment fault text data and fault image data in turn;

[0008] Step 2: Based on the extracted fault features of multiple dimensions, a model-level multi-level long short-term memory ML_LSTM feature information fusion model is constructed to fuse the fault features;

[0009] Step 3: Based on the fused fault features, the equipment fault is diagnosed through the Softmax fault diagnosis model to obtain the fault diagnosis type;

[0010] Step 4: Based on the fault diagnosis type, a hierarchical recommendation model is constructed to output the fault location and operation and maintenance strategy.

[0011] The present invention further includes the following preferred embodiments:

[0012] The identifying fault features of multiple dimensions from the charging device fault text data and the fault picture data in sequence further includes:

[0013] Step 1.1: Obtain charging equipment fault text data, which includes fault description, operation and maintenance report, and equipment log. Extract fault text features F from the charging equipment fault text data based on the pre-trained fault keyword recognition model. text ;

[0014] Step 1.2: Obtain the fault image of the charging device in the system, and extract the fault image feature F based on the pre-trained charging device image fault feature recognition model img ;

[0015] Step 1.3: Obtain structured data transmitted by multiple system interfaces and determine the feature dimension. The horizontal dimension is the charging device F ce 、Fault location F fl 、Fault Item F fm , Fault level F fs , the vertical dimension is time, and the fields related to fault diagnosis are selected as features according to the dimension, and analyzed and calculated to extract the fault feature F feat .

[0016] The fusing of fault features further includes:

[0017] Step 2.1: Fault text feature F text Input the LSTM model to get the first layer as the input of the next layer. The hidden layer output function is:

[0018]

[0019] Among them, F text is the extracted text features; is the hidden layer output at the previous moment t-1;

[0020] Step 2.2: Extract the image features F img The features and the first layer are input into the LSTM model to obtain the second layer, which is used as the input of the next layer. The hidden layer output function is:

[0021]

[0022] Among them, F img is the extracted image feature;

[0023] Step 2.3: Input the extracted field features and the second layer into the LSTM model to obtain the hidden state of the third layer, and use the LSTM-Layer3 fusion features as the feature fusion result. The hidden layer output function is:

[0024]

[0025] Among them, F feat is the extracted field feature.

[0026] The diagnosing the equipment fault by using the Softmax fault diagnosis model to obtain the fault diagnosis type further includes:

[0027] Step 3.1: Based on the output feature matrix H = [h1, h2, h3…, h i ,…h t ], and through the linear prediction function, we get O = [o1, o2, o3, ..., o i ,…o t-1 ], the prediction result objective function is:

[0028] O=W·H+b

[0029] Among them, b is the bias vector and W is the weight matrix;

[0030] Step 3.2: Process the linear prediction function value through the SoftMax function and output the final fault distribution probability. The objective function is as follows:

[0031]

[0032] Among them, e is a natural constant, o i is the i-th predicted value.

[0033] The constructing of a hierarchical recommendation model based on the fault diagnosis type and outputting the fault location and operation and maintenance strategy further includes:

[0034] Step 4.1: Preliminary screening of the fault location and operation and maintenance recommendation rule base, using existing standard industry data and expert experience to define the relationship between different charging equipment, fault locations, fault items and operation and maintenance strategies;

[0035] Step 4.2: Verify the fault location and operation and maintenance recommended case library, and update the case library in real time based on the system's historical faults and operation and maintenance records. Each case includes charging equipment, fault location, fault item, and actual operation and maintenance strategy. Construct a feature vector for each case, match similar cases through the Euclidean distance similarity algorithm, and confirm the handling method of similar faults.

[0036] The present invention also discloses a charging equipment fault diagnosis system based on multi-source heterogeneous data semantic fusion using the above-mentioned charging equipment fault diagnosis method based on multi-source heterogeneous data semantic fusion, comprising:

[0037] A fault feature extraction module is used to obtain the fault text data and fault image data of the charging equipment, and use the pre-trained fault keyword recognition model and fault image recognition model to identify fault features of multiple dimensions from the fault text data and fault image data of the charging equipment in turn;

[0038] The fault feature fusion module is used to build a model-level multi-level long short-term memory ML_LSTM feature information fusion model based on the extracted fault features of multiple dimensions to fuse the fault features;

[0039] The fault type acquisition module is used to diagnose the equipment fault based on the fused fault features through the Softmax fault diagnosis model to obtain the fault diagnosis type;

[0040] The operation and maintenance strategy recommendation module is used to build a hierarchical recommendation model based on the fault diagnosis type and output the fault location and operation and maintenance strategy.

[0041] Accordingly, the present application also discloses a terminal, including a processor and a storage medium;

[0042] The storage medium is used to store instructions;

[0043] The processor is used to operate according to the instructions to execute the steps of the aforementioned charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data.

[0044] Correspondingly, the present application also discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aforementioned charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data.

[0045] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a charging equipment fault diagnosis method and system based on the semantic fusion of multi-source heterogeneous data, combines multi-source heterogeneous fault semantic recognition, feature fusion, fault classification and hierarchical operation and maintenance recommendation methods, makes full use of the structured and unstructured data in the charging equipment system, realizes fault diagnosis and effective prevention based on existing data, locates the fault site and gives reasonable operation and maintenance suggestions, effectively improves the accuracy of fault diagnosis, reduces operation and maintenance costs, and provides effective guarantee for the safe operation of charging equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Schematic diagram of a flow chart according to an exemplary embodiment of the present invention.

[0047] Figure 2 Schematic diagram of the feature extraction and fusion diagnosis process in an embodiment of the present invention.

[0048] Figure 3 The figure is a flow chart of a fault classification method in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0050] The embodiments described in this application are only some embodiments of the present invention, not all embodiments. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the protection scope of the present invention.

[0051] In view of the shortcomings of the prior art, the present invention proposes a charging equipment fault diagnosis method and system based on the semantic fusion of multi-source heterogeneous data. The fault features are identified by constructing a multi-source heterogeneous fault semantic feature recognition model; a model-level multi-level long short-term memory ML_LSTM feature information fusion model is constructed based on the fault features to fuse the multi-source fault features; a Softmax fault diagnosis model is constructed based on the fused fault features to diagnose the equipment fault type; a hierarchical recommendation model is constructed based on the fault type to locate the fault location and recommend operation and maintenance strategies. By deeply integrating multi-source heterogeneous data, semantic fusion technology is used to realize intelligent understanding and association of data, and combined with time series prediction algorithms, future fault trends can be accurately predicted to achieve early warning and active prevention of faults.

[0052] See also Figure 1 As shown, the charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data disclosed in the present invention includes the following steps:

[0053] Step 1: Obtain the charging equipment fault text data and fault image data, and use the pre-trained fault keyword recognition model and fault image recognition model to identify fault features of multiple dimensions from the charging equipment fault text data and fault image data respectively.

[0054] The specific steps of step 1 are as follows:

[0055] Step 1.1: Obtain charging equipment fault text data. The fault text data includes fault description, operation and maintenance report and equipment log. Based on the pre-trained fault keyword recognition model, extract the fault text feature F from the charging equipment fault text data. text .

[0056] The step 1.1 comprises:

[0057] Step 111: Preprocess the acquired fault text data. First, clean the data by using the BeautifulSoup library to delete tags, simple regular expressions (regexes) to delete special characters, stop words, extra spaces, numbers, etc. Then, use the jieba tool to segment the text data based on the Chinese stop word list, Baidu stop word list, Harbin Institute of Technology stop word list, etc. Build a synonym list for common faults to ensure that the model can recognize words with similar semantics and ensure the accuracy of the model.

[0058] Step 112: Determine the characteristic dimension, the horizontal dimension is the charging device F ce , Fault Classification F fc 、Fault location F fl 、Fault Item F fm , Fault level F fs , the vertical dimension is time, for example, it can be divided by date.

[0059] Step 113: Based on the Word2Vec training word vector, the Skip-gram model is used to predict the context words according to the central word. The Skip-gram model is usually composed of an input layer, a hidden layer and an output layer. The input layer receives the one-hot encoding of the target word, the projection layer is used to calculate the word embedding, and the output layer outputs the predicted probability of the context word. The word vectors of each word obtained by Word2Vec training are added and then averaged.

[0060] Step 114: Classify the fault features based on bag-of-words and tf-idf models. The bag-of-words model counts the frequency of words appearing in the text set. Each time a word appears, the value represented by the corresponding word in the dictionary is increased by 1. The tf-idf model adds weights to keywords on this basis.

[0061]

[0062] In the above formula, N represents the number of all words in the text, N t is the number of keywords t.

[0063] Step 115: Extract fault keywords through the above trained charging facility fault keyword recognition model, including fault description, operation and maintenance report, equipment log, etc. Through the extracted text feature value of each dimension, the overall text feature value function is obtained as follows:

[0064]

[0065] In the above formula, The extracted charging device dimension F ce The text feature value of is the extracted fault classification dimension F fc The text feature value of F is the extracted fault location dimension fl The text feature value of F is the extracted fault item dimension fm The text feature value of is the extracted fault level dimension F fs The text feature value of .

[0066] Step 1.2: Obtain the fault image of the charging device in the system, and extract the fault image feature F based on the pre-trained charging device image fault feature recognition model img .

[0067] The step 1.2 further comprises:

[0068] Step 121: Determine the characteristic dimension, the horizontal dimension is the charging device F ce , Fault Classification F fc 、Fault location F fl 、Fault Item F fm , Fault level F fs , the vertical dimension is time (divided by date).

[0069] Step 122: Using the convolutional neural network (CNN) model, the obtained fault image of the charging device is input. The image is usually composed of three color channels (red, green, and blue), forming a two-dimensional matrix X, which represents the intensity value of the pixel.

[0070] Step 123: The model is composed of multiple convolutional layers. All convolution kernels extract the features of the input samples through weight sharing as the input of the next convolutional layer. Based on the same convolution mode, the ReLu function is used to activate the convolution result. The output of the lth convolution layer is as follows:

[0071]

[0072] In the above formula, (i, j) is the feature coordinate of the input image of the lth layer, (m, n) is the coordinate of the convolution kernel K, and ReLu(x) = max(0, x) is the activation function.

[0073] Step 124: The pooling layer reduces the computational complexity by reducing the size of the feature map to prevent overfitting. The maximum pooling is used to sparsely process the feature map to further amplify the main features. Assuming the pooling window size is k, the output feature value of the lth layer is:

[0074]

[0075] In the above formula, (i, j) is the coordinate of the output feature map y, and (a, b) represents the coordinate within the pooling window.

[0076] Step 125: The fully connected layer integrates the extracted local features into global features. After feature weighting, the global feature output value of the lth layer is:

[0077] F l =W l ×Z l +b l

[0078] In the above formula, W l is the weight matrix, Z l is the input value of the pooling layer, b l is the bias vector.

[0079] Among them, the weight matrix W l The gradient descent method is used to minimize the loss function, which measures the gap between the model's predicted output and the true label. Through multiple iterations of this process, the weight matrix is ​​gradually adjusted so that the model's predicted output is closer to the true label, thereby improving the performance of the model.

[0080] Step 126: Through the above trained charging equipment image fault feature recognition model, the fault features in the fault image are extracted, and the image feature values ​​of each dimension are calculated to obtain the total image feature function as follows:

[0081]

[0082] In the above formula, The extracted charging device dimension F ce The image feature value, is the extracted fault classification dimension F fc The image feature value, F is the extracted fault location dimension fl The image feature value, F is the extracted fault item dimension fm The image feature value, is the extracted fault level dimension F fs The image feature value.

[0083] Step 1.3: Obtain structured data transmitted by multiple system interfaces and determine the feature dimension. The horizontal dimension is the charging device F ce , Fault Classification F fc 、Fault location F fl 、Fault Item F fm , Fault level F fs The vertical dimension is time (divided by date). According to the dimension, the fields related to fault diagnosis are selected as features, and the analysis and calculation are performed to extract the fault feature F. feat , the objective function is as follows:

[0084]

[0085] In the above formula, Calculate the charging device dimension F for structured data analysis ce The fault characteristic value,

[0086] Calculate the fault classification dimension F for structured data analysis fc The fault characteristic value, Calculate the fault location dimension F for structured data analysis fl The characteristic value of Calculate the fault item dimension F for structured data analysis fm The fault characteristic value, Calculate the fault level dimension F for structured data analysis fs The fault characteristic value.

[0087] Step 2: Based on the extracted fault features of multiple dimensions, a model-level multi-level long short-term memory ML_LSTM feature information fusion model is constructed to fuse the fault features.

[0088] Feature information fusion uses model-level fusion ML-LSTM, which is a deeper fusion method that produces more optimized joint discriminant feature representations for classification and regression tasks. ML-LSTM combines a multi-layer network with a traditional LSTM model. Figure 2 As shown, the specific steps are as follows:

[0089] Step 2.1: Extract the text features F from step 1.1 text Input the LSTM model to get the first layer LSTM-Layer1 as the input of the next layer. The hidden layer output function is as follows:

[0090]

[0091] Among them, F text is the extracted text features; is the hidden layer output at the previous time t-1.

[0092] Step 2.2: Extract the image features F img The features and LSTM-Layer1 input get the second layer LSTM-Layer2, which is used as the input of the next layer. The hidden layer output function is as follows:

[0093]

[0094] Among them, F img is the extracted image feature, To combine features from different sources (F img ) t and

[0095] Step 2.3: The extracted field features are combined with LSTM-Layer2 input to obtain the hidden state of the third layer LSTM-Layer3. The LSTM-Layer3 fusion feature is used as the feature fusion result. The hidden layer output function is as follows:

[0096]

[0097] Among them, F feat is the extracted field feature, To combine features from different sources (F feat ) t and

[0098] Step 3: Based on the fused fault features, the equipment fault is diagnosed through the Softmax fault diagnosis model to obtain the fault diagnosis type.

[0099] The Softmax function is used to handle multi-classification problems. It converts a vector containing arbitrary real numbers into a probability distribution so that the value of each element is between 0 and 1 and the sum of all elements is 1. The specific method is as follows Figure 3 The specific steps of step 3 are as follows:

[0100] Step 3.1: Based on the ML-LSTM model, we finally get the outputs of the hidden layers at multiple times: h1, h2, h3, ..., h i ,...,h t , forming the output feature matrix H, and then passing through the linear prediction function to obtain O = [o1, o2, o3, ..., o i ,…ot-1 ], the prediction result objective function is as follows:

[0101] O=W·H+b

[0102] In the above formula, b is the bias vector and W is the weight matrix.

[0103] Step 3.2: Process the linear prediction function value through the Softmax function and output the final fault distribution probability. The objective function is as follows:

[0104]

[0105] Among them, e is a natural constant, o i is the i-th predicted value.

[0106] The Softmax function converts the predicted value of each category into a probability value, where The larger the value is, the higher the probability of belonging to this category is, and the classification with the highest probability is taken as the fault diagnosis result.

[0107] Step 4: Based on the fault diagnosis type, a hierarchical recommendation model is constructed to output the fault location and operation and maintenance strategy.

[0108] Preferably, the specific process of step 4 is as follows:

[0109] Step 4.1: Conduct a preliminary screening of the fault location and operation and maintenance recommendation rule base, and use existing standard industry data and expert experience to define the relationship between different charging equipment, fault locations, fault items and operation and maintenance strategies.

[0110] Step 4.2: Verify the fault location and operation and maintenance recommendation case library, and update the case library in real time based on the system's historical faults and operation and maintenance records. Each case includes charging equipment, fault location, fault item, and actual operation and maintenance strategy. Construct a feature vector for each case, match similar cases through the Euclidean distance similarity algorithm, confirm the handling method of similar faults, and provide accurate operation and maintenance suggestions.

[0111] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a charging equipment fault diagnosis method and system based on the semantic fusion of multi-source heterogeneous data, combines multi-source heterogeneous fault semantic recognition, feature fusion, fault classification and hierarchical operation and maintenance recommendation methods, makes full use of the structured and unstructured data in the charging equipment system, realizes fault diagnosis and effective prevention based on existing data, locates the fault site and gives reasonable operation and maintenance suggestions, effectively improves the accuracy of fault diagnosis, reduces operation and maintenance costs, and provides effective guarantee for the safe operation of charging equipment.

[0112] The present invention may be a system, a method and / or a computer program product. The present invention also discloses a charging device fault diagnosis system based on multi-source heterogeneous data semantic fusion based on the aforementioned charging device fault diagnosis method based on multi-source heterogeneous data semantic fusion, comprising:

[0113] A fault feature extraction module is used to obtain the fault text data and fault image data of the charging equipment, and use the pre-trained fault keyword recognition model and fault image recognition model to identify fault features of multiple dimensions from the fault text data and fault image data of the charging equipment in turn;

[0114] The fault feature fusion module is used to build a model-level multi-level long short-term memory ML_LSTM feature information fusion model based on the extracted fault features of multiple dimensions to fuse the fault features;

[0115] The fault type acquisition module is used to diagnose the equipment fault based on the fused fault features through the Softmax fault diagnosis model to obtain the fault diagnosis type;

[0116] The operation and maintenance strategy recommendation module is used to build a hierarchical recommendation model based on the fault diagnosis type and output the fault location and operation and maintenance strategy.

[0117] Based on the spirit of the present invention, those skilled in the art can easily think that a computer program product can be obtained based on the aforementioned charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data. The computer program product may include a computer-readable storage medium, which carries computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure. That is, the present application also includes a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the aforementioned charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data.

[0118] Computer readable storage medium can be a tangible device that can keep and store the instructions used by the instruction execution device. Computer readable storage medium can be, for example, - but not limited to - electrical storage device, magnetic storage device, optical storage device, electromagnetic storage device, semiconductor storage device or any suitable combination of the above. More specific examples (non-exhaustive list) of computer readable storage medium include: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, for example, punch card or groove protrusion structure with instructions stored thereon and any suitable combination of the above. Computer readable storage medium used here is not interpreted as instantaneous signal itself, such as radio wave or other free propagating electromagnetic wave, electromagnetic wave propagated by waveguide or other transmission medium (for example, light pulse by optical fiber cable) or electrical signal transmitted by wire.

[0119] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0120] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages-such as Smalltalk, C++, etc., and conventional procedural programming languages-such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network-including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA) may be personalized by utilizing the state information of a computer-readable program instruction, and the electronic circuit may execute a computer-readable program instruction, thereby realizing various aspects of the present disclosure.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data, characterized in that: The following steps are involved: Step 1: Obtain the charging equipment fault text data and fault image data, and use the pre-trained fault keyword recognition model and fault image recognition model to identify fault features of multiple dimensions from the charging equipment fault text data and fault image data in turn; Step 2: Based on the extracted fault features of multiple dimensions, a model-level multi-level long short-term memory network ML-LSTM feature information fusion model is constructed to fuse the fault features; Step 3: Based on the fused fault features, the equipment fault is diagnosed through the Softmax fault diagnosis model to obtain the fault diagnosis type; Step 4: Based on the fault diagnosis type, a hierarchical recommendation model is constructed to output the fault location and operation and maintenance strategy.

2. The charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data according to claim 1 is characterized in that: The identifying fault features of multiple dimensions from the charging device fault text data and the fault picture data in sequence further includes: Step 1.1: Obtain charging equipment fault text data, which includes fault description, operation and maintenance report, and equipment log. Extract fault text features F from the charging equipment fault text data based on the pre-trained fault keyword recognition model. text ; Step 1.2: Obtain the fault image of the charging device in the system, and extract the fault image feature F based on the pre-trained charging device image fault feature recognition model img ; Step 1.3: Obtain structured data transmitted by multiple system interfaces and determine the feature dimension. The horizontal dimension is the charging device F ce , Fault Classification F fc 、Fault location F fl 、Fault Item F fm , Fault level F fs , the vertical dimension is time, and the fields related to fault diagnosis are selected as features according to the dimension, and analyzed and calculated to extract the fault feature F feat .

3. The charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data according to claim 2 is characterized in that: The fusing of fault features further includes: Step 2.1: Transform the fault text feature F text Input the LSTM model to get the first layer as the input of the next layer. The hidden layer output function is: Among them, F text is the extracted text features; is the hidden layer output at the previous moment t-1; Step 2.2: Extract the image features F img The features and the first layer are input into the LSTM model to obtain the second layer, which is used as the input of the next layer. The hidden layer output function is: Among them, F img is the extracted image feature; Step 2.3: Input the extracted field features and the second layer into the LSTM model to obtain the hidden state of the third layer, and use the LSTM-Layer3 fusion features as the feature fusion result. The hidden layer output function is: Among them, F feat is the extracted field feature.

4. The charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data according to claim 3 is characterized in that: The diagnosing the equipment fault by using the Softmax fault diagnosis model to obtain the fault diagnosis type further includes: Step 3.1: Based on the output feature matrix H = [h1, h2, h3…, h i ,…h t ], and through the linear prediction function, we get O = [o1, o2, o3, ..., o i ,…o t-1 ], the prediction result objective function is: O=W·H+b Among them, b is the bias vector and W is the weight matrix; Step 3.2: Process the linear prediction function value through the SoftMax function and output the final distribution probability of the fault. The objective function is as follows: Among them, e is a natural constant, o i is the i-th predicted value.

5. The charging equipment fault diagnosis system based on semantic fusion of multi-source heterogeneous data according to claim 4 is characterized in that: The constructing of a hierarchical recommendation model based on the fault diagnosis type and outputting the fault location and operation and maintenance strategy further includes: Step 4.1: Preliminary screening of the fault location and operation and maintenance recommendation rule base, using existing standard industry data and expert experience to define the relationship between different charging equipment, fault locations, fault items and operation and maintenance strategies; Step 4.2: Verify the fault location and operation and maintenance recommended case library, and update the case library in real time based on the system's historical faults and operation and maintenance records. Each case includes charging equipment, fault location, fault item, and actual operation and maintenance strategy. Construct a feature vector for each case, match similar cases through the Euclidean distance similarity algorithm, and confirm the handling method of similar faults.

6. A charging equipment fault diagnosis system based on semantic fusion of multi-source heterogeneous data, characterized in that: include: A fault feature extraction module is used to obtain the fault text data and fault image data of the charging equipment, and use the pre-trained fault keyword recognition model and fault image recognition model to identify fault features of multiple dimensions from the fault text data and fault image data of the charging equipment in turn; The fault feature fusion module is used to build a model-level multi-level long short-term memory ML_LSTM feature information fusion model based on the extracted fault features of multiple dimensions to fuse the fault features; The fault type acquisition module is used to diagnose the equipment fault based on the fused fault features through the Softmax fault diagnosis model to obtain the fault diagnosis type; The operation and maintenance strategy recommendation module is used to build a hierarchical recommendation model based on the fault diagnosis type and output the fault location and operation and maintenance strategy.

7. The charging equipment fault diagnosis system based on semantic fusion of multi-source heterogeneous data according to claim 6 is characterized in that: The fault feature extraction module is further used to: Obtain charging equipment fault text data, which includes fault description, operation and maintenance report, and equipment log. Extract fault text features F from the charging equipment fault text data based on the pre-trained fault keyword recognition model. text ; Obtain the fault image of the charging device in the system, and extract the fault image feature F based on the pre-trained charging device image fault feature recognition model img ; Obtain structured data transmitted by multiple system interfaces and determine the characteristic dimension. The horizontal dimension is the charging device F ce 、Fault location F fl 、Fault Item F fm , Fault level F fs , the vertical dimension is time, and the fields related to fault diagnosis are selected as features according to the dimension, and analyzed and calculated to extract the fault feature F feat .

8. The charging equipment fault diagnosis system based on semantic fusion of multi-source heterogeneous data according to claim 7 is characterized in that: The fault feature fusion module is further used to: Fault text feature F text Input the LSTM model to get the first layer as the input of the next layer. The hidden layer output function is: Among them, F text is the extracted text features; is the hidden layer output at the previous moment t-1; The extracted image features F img The features and the first layer are input into the LSTM model to obtain the second layer, which is used as the input of the next layer. The hidden layer output function is: Among them, F img is the extracted image feature; The extracted field features and the second layer are input into the LSTM model to obtain the hidden state of the third layer. The LSTM-Layer3 fusion feature is used as the feature fusion result. The hidden layer output function is: Among them, F feat is the extracted field feature.

9. The charging equipment fault diagnosis model based on semantic fusion of multi-source heterogeneous data according to claim 8 is characterized in that: The fault type acquisition module is further used to: Based on the output feature matrix H=[h1, h2, h3…, h i ,…h t ], and through the linear prediction function, we get O = [o1, o2, o3, ..., o i ,…o t-1 ], the prediction result objective function is: O=W·H+b Among them, b is the bias vector and W is the weight matrix; The linear prediction function value is processed by the SoftMax function to output the final distribution probability of the fault. The objective function is as follows: Among them, e is a natural constant, o i is the i-th predicted value.

10. The charging equipment fault diagnosis system based on semantic fusion of multi-source heterogeneous data according to claim 9 is characterized in that: The operation and maintenance strategy recommendation module is further used to: Conduct a preliminary screening of the fault location and operation and maintenance recommendation rule base, and use existing standard industry data and expert experience to define the relationship between different charging equipment, fault locations, fault items, and operation and maintenance strategies; Verify the fault location and operation and maintenance recommended case library, and update the case library in real time based on the system's historical faults and operation and maintenance records. Each case includes charging equipment, fault location, fault item and actual operation and maintenance strategy. Construct a feature vector for each case, match similar cases through the Euclidean distance similarity algorithm, and confirm the handling method of similar faults.

11. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instruction to execute the steps of the charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data according to any one of claims 1-5.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the charging equipment fault diagnosis method based on semantic fusion of multi-source heterogeneous data described in any one of claims 1 to 5 are implemented.

Images