Power consumption information acquisition terminal field fault remote diagnosis system based on BLSTM-GRU model

Abstract

The invention particularly relates to a BLSTM-GRU model-based field fault remote diagnosis system for an electricity consumption information acquisition terminal. The system comprises an electricity utilization information acquisition terminal and a cloud control center, and further comprises a log acquisition module, a processing module, a data set preparation module, a BLSTM-GRU model training module, a fault diagnosis module and a fault report module. According to the method, the BLSTM-GRU model can quickly and accurately detect and classify various fault types in the power utilization information acquisition terminal, and has the advantage of capturing bidirectional information; the system is deployed in a cloud control center, can automatically and remotely obtain and carry out real-time analysis and diagnosis on the fault type of the power utilization information acquisition terminal, eliminates the downloading and analysis time consumed by operation and maintenance personnel on site, and directly improves the fault solving efficiency. And the demand quantity of operation and maintenance personnel is reduced, so that the management cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grid information management, and in particular to an on-site fault remote diagnosis system for an electricity consumption information collection terminal. Background Art

[0002] The electricity consumption information collection terminal is a terminal device installed in an electricity management area. It is connected to multiple user meters in the area and manages all connected user meters through scheduled communication.

[0003] In actual use, the electricity usage information collection terminal will communicate and interact with the user's meter's normal or abnormal electricity usage information and generate an operation log. Abnormal electricity usage information indicates that the corresponding user's meter has failed. The power supply company's operation and maintenance personnel will regularly access the electricity usage information collection terminal offline to physically download and manually analyze the operation log to determine the fault type of the user's meter, thereby facilitating the next step of targeted meter repair and restoration.

[0004] The existing fault diagnosis method for electricity consumption information collection terminals requires maintenance personnel to visit the terminals on-site and download the operation logs via data cables or other communication methods. Since each terminal manages a large number of user meters, the operation logs are often large in size and require a long download time. Furthermore, manual analysis requires additional time, significantly reducing fault resolution efficiency. Furthermore, the significant investment in human resources for offline maintenance results in high management costs. Summary of the Invention

[0005] The purpose of the present invention is to provide a remote fault diagnosis system for electricity consumption information collection terminals based on the BLSTM-GRU model. The system can automatically obtain the fault type of the electricity consumption information collection terminal remotely and perform real-time analysis and diagnosis, eliminating the time spent by operation and maintenance personnel on site for downloading and parsing, directly improving the efficiency of fault resolution; and reducing the number of operation and maintenance personnel required, thereby reducing management costs.

[0006] In order to achieve the above object, the technical solutions adopted by the present invention are as follows:

[0007] A remote fault diagnosis system for an electricity consumption information collection terminal based on a BLSTM-GRU model, comprising an electricity consumption information collection terminal and a cloud control center, and characterized by further comprising:

[0008] The log acquisition module is used to obtain the original operation log of the power consumption information collection terminal and send it to the preprocessing module;

[0009] A preprocessing module, configured to obtain abnormal target message sequences according to the original operation log, classify the target message sequences, and package them to form an original data set;

[0010] The dataset preparation module is used to obtain the original dataset and divide it into training set, test set and validation set according to the preset quantity ratio;

[0011] A BLSTM-GRU model training module is used to iteratively train the constructed BLSTM-GRU model using the training set, update the model parameters, and evaluate using the test set and validation set to obtain the optimal BLSTM-GRU model after training;

[0012] The fault diagnosis module is used to wirelessly obtain the on-site operation log of the power consumption information collection terminal in real time and send it to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis to obtain predictive diagnosis results.

[0013] Therefore, the purpose of this invention is to deploy a system in a cloud control center that can wirelessly obtain the operation logs of the electricity usage information collection terminal and perform fault diagnosis based on the operation logs to obtain predictive diagnostic results. This process is executed periodically and automatically, allowing operation and maintenance personnel to directly obtain the fault type of each user's electricity meter connected to and managed by the electricity usage information collection terminal, eliminating the need for operation and maintenance personnel to visit the electricity usage information collection terminal on-site, download the operation logs, and manually analyze them.

[0014] The present invention firstly forms a remote fault diagnosis network by a cloud control center and a plurality of power consumption information collection terminals connected to the cloud control center by wireless communication.

[0015] The log acquisition module is used to obtain the original operation log of the electricity consumption information collection terminal and send it to the preprocessing module. The operation log is the content of communication interactions collected in chronological order, which is reflected in the message sequence formed by multiple messages arranged according to timestamps. Each message can be regarded as a data point, and there is sequentiality and potential correlation between data points, that is, the communication information corresponding to one timestamp may affect the communication information corresponding to the next timestamp. For example: the electricity consumption information collection terminal fails to collect data from a user's meter at a certain point in time, so the stored data at that timestamp is empty, and the subsequent active reading of the timestamp data will fail. This fault is a dependency relationship, and the three problems are sequential and correlated. The original operation log is used as a source of pre-order training data and can be obtained from historical data or targeted experimental data.

[0016] The pre-processing module is used to obtain abnormal target message sequences based on the original operation log, classify the target message sequences and package them to form an original data set. Message sequences include two types: those without abnormalities and those with abnormalities. The present invention extracts abnormal message sequences and defines them as target message sequences. There are many types of faults corresponding to target message sequences, such as meter data collection failure, data storage failure, terminal data reading failure, terminal data reporting failure, data call failure and event collection failure. Before generating the original data set, it is necessary to classify the fault types corresponding to the target message sequence. The classification method can be based on engineer experience judgment or equipment manual comparison classification. The classification operation class can be used to mark the fault type labels of each original data set for distinction.

[0017] The dataset preparation module is used to obtain the original dataset and divide it into training set, test set and validation set according to the preset quantity ratio; the preset quantity ratio can be considered as a setting, following the principle of majority training set, minority test set and validation set to ensure the effectiveness and accuracy of model training.

[0018] The BLSTM-GRU model is a pre-built bidirectional long short-term memory (LSTM)-gated recurrent unit (GRU) hybrid network. The BLSTM is a variant of the LSTM (Long Short-Term Memory) network. By processing sequence data in both the forward and backward directions, it can capture bidirectional information within the sequence. This property makes the BLSTM highly effective for tasks that require understanding the context of the entire sequence, such as sentiment analysis in natural language processing and machine translation. The GRU is another variant of the recurrent neural network (RNN) designed to address the vanishing or exploding gradient problems that traditional RNNs are prone to when processing long sequences. Compared to the LSTM, the GRU has a simpler structure, with only two gates (an update gate and a reset gate) to control the flow of information. This makes the GRU more efficient during training while still retaining its ability to capture long-term dependencies.

[0019] In the present invention, a BLSTM network is first used to capture bidirectional information in the target message sequence, and the output of the BLSTM layer is then used as the input to the GRU layer. This structure leverages the BLSTM's advantage in capturing bidirectional information and further extracts and refines this information through the GRU layer. In this case, the BLSTM-GRU model can use its bidirectional learning capabilities to understand the connections between the states of electric meter equipment in different time periods and, through the GRU mechanism, efficiently filter out important features that aid in fault identification. This means that the model can make judgments not only based on current data but also consider past and future information. This is very useful in practice, as equipment failures are often not instantaneous but rather the result of the accumulation of multiple factors over a period of time. Therefore, compared to traditional methods that only consider current input or unidirectional time series models, the BLSTM-GRU model, due to its ability to capture long-term dependencies in time series data, can improve classification accuracy when dealing with tasks such as electric meter equipment fault prediction.

[0020] The BLSTM-GRU model training module is used to iteratively train the constructed BLSTM-GRU model using the training set, update the model parameters, and evaluate it using the test set and validation set to obtain the optimal BLSTM-GRU model after training.

[0021] Deploy the optimal BLSTM-GRU model to the cloud control center.

[0022] The fault diagnosis module is connected to the cloud control center and is used to wirelessly obtain the on-site operation logs of the electricity consumption information collection terminal in real time. Specifically, it periodically downloads the on-site operation logs of the electricity consumption information collection terminal through wireless transmission, eliminating the time-consuming problem of manually downloading large amounts of data at one time; and sends them to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis; the optimal BLSTM-GRU model automatically generates predictive diagnosis results and then sends them to the fault diagnosis module. Operation and maintenance personnel can directly perform targeted repairs on user meters based on the above predictive diagnosis results.

[0023] In the power system environment, parameters such as voltage, current, and frequency vary over time. This data can span very long time periods, and the system itself may exhibit different behavioral patterns over different time periods. For example, power load may vary at different times of the day. This data is stored in operation logs using a time-based message format. When using the BLSTM model to process such power system data logs, its memory unit mechanism enables the model to maintain good performance even when processing long time series data. Specifically, this mechanism can store critical historical information, allowing it to be effectively utilized even when it is separated by long time periods. For example, the BLSTM can use the memory unit mechanism to remember a certain power grid pattern from several months ago and respond or predict accordingly when a similar pattern is discovered in current data.

[0024] In the present invention, the optimal BLSTM-GRU model can quickly and accurately detect and classify various fault types in the electricity consumption information collection terminal, and has the advantage of capturing two-way information; when deployed in the cloud control center, it can automatically obtain remotely and perform real-time analysis and diagnosis of the fault types of the electricity consumption information collection terminal, eliminating the time spent on on-site downloading and parsing by operation and maintenance personnel, directly improving the efficiency of fault resolution; and reducing the number of operation and maintenance personnel required, thereby reducing management costs.

[0025] As a preferred embodiment of the present invention, the specific method for the preprocessing module to classify the target message sequence is: equipment manual comparison classification method; specifically, the message sequence fault schematic code in the production manual of the electricity consumption information collection terminal is manually compared with the data points in the target message sequence to obtain the fault type of the target message sequence.

[0026] As a preferred embodiment of the present invention, the network architecture of the BLSTM-GRU model is specifically: a sequential input layer, a word embedding layer, a BLSTM layer, a first dropout layer, a GRU layer, a second dropout layer, an LSTM layer, a first fully connected layer and a second fully connected layer connected in sequence.

[0027] As a preferred embodiment of the present invention, the second fully connected layer uses a Softmax function to generate the probability of each fault category; the Softmax function is specifically:

[0028]

[0029] Where z represents the input vector, k represents the number of fault types, N represents the total number of data sets, and i represents the data set number.

[0030] Therefore, k represents the number of fault types. For different fault types, the corresponding number of fault types is set when the BLSTM-GRU model is arranged; N represents the total number of data sets, that is, the number of training sets input into the model; i represents the data set number, that is, the number of the specific training set among all the training sets corresponding to the above input.

[0031] As a preferred embodiment of the present invention, the remote fault diagnosis system of the electricity consumption information collection terminal further includes a fault reporting module, which is used to receive the predictive diagnosis results sent by the fault diagnosis module and send them to the personal mobile terminal of the operation and maintenance personnel via wireless communication.

[0032] As a preferred embodiment of the present invention, the fault types corresponding to the target message sequence include: meter data collection failure, data storage failure, terminal data reading failure, terminal data reporting failure, data call failure, event collection failure, event reporting failure, event storage failure, event recording failure, event call failure, overdue control, and unchanged status after control.

[0033] As a preferred embodiment of the present invention, the dataset preparation module divides the training set, test set and validation set according to the preset quantity ratio to obtain the corresponding preset quantity ratio of 14:3:3.

[0034] On the other hand, the present invention also provides a method for remote diagnosis of on-site faults of power consumption information collection terminals based on a BLSTM-GRU model, characterized in that the method comprises:

[0035] S1. Obtain the original operation log of the electricity consumption information collection terminal and send it to the preprocessing module;

[0036] S2. The preprocessing module obtains abnormal target message sequences according to the original operation log, classifies the target message sequences, and packages them to form an original data set;

[0037] S3, the data set preparation module obtains the original data set and divides it into training set, test set and validation set according to the preset quantity ratio;

[0038] S4. Use the training set to iteratively train the constructed BLSTM-GRU model, update the model parameters, and use the test set and validation set for evaluation to obtain the optimal BLSTM-GRU model after training.

[0039] S5. Deploy the optimal BLSTM-GRU model to the cloud control center;

[0040] S6. The fault diagnosis module wirelessly obtains the on-site operation log of the power consumption information collection terminal in real time, and sends it to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis to obtain the predicted diagnosis result.

[0041] S7. The fault reporting module receives the predicted diagnosis result sent by the fault diagnosis module and sends it to the personal mobile terminal of the operation and maintenance personnel via wireless communication.

[0042] In summary, the present invention has the following beneficial effects:

[0043] In the present invention, the optimal BLSTM-GRU model can quickly and accurately detect and classify various fault types in the electricity consumption information collection terminal, and has the advantage of capturing two-way information; when deployed in the cloud control center, it can automatically obtain remotely and perform real-time analysis and diagnosis of the fault types of the electricity consumption information collection terminal, eliminating the time spent on on-site downloading and parsing by operation and maintenance personnel, directly improving the efficiency of fault resolution; and reducing the number of operation and maintenance personnel required, thereby reducing management costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of this specification, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0045] Figure 1 This is the system block diagram of the remote fault diagnosis system for the electricity consumption information collection terminal;

[0046] Figure 2 This is a flow chart of the remote fault diagnosis method for the electricity consumption information collection terminal;

[0047] Figure 3 This is a structural diagram of the BLSTM-GRU model. DETAILED DESCRIPTION

[0048] The following is an explanation and description of the technical solutions of the embodiments of the present invention in conjunction with the drawings of the embodiments of the present invention. However, the following embodiments are only preferred embodiments of the present invention and are not exhaustive. Based on the embodiments in the implementation manner, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention.

[0049] Throughout this specification, the claims, and the accompanying drawings, the terms "first," "second," and so forth are used to distinguish between different items, not to describe a particular order. Furthermore, the term "comprises" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may include other steps or elements inherent to the process, method, product, or apparatus.

[0050] Therefore, the purpose of this invention is to deploy a system in a cloud control center that can wirelessly obtain the operation logs of the electricity usage information collection terminal and perform fault diagnosis based on the operation logs to obtain predictive diagnostic results. This process is executed periodically and automatically, allowing operation and maintenance personnel to directly obtain the fault type of each user's electricity meter connected to and managed by the electricity usage information collection terminal, eliminating the need for operation and maintenance personnel to visit the electricity usage information collection terminal on-site, download the operation logs, and manually analyze them.

[0051] The present invention first comprises a cloud control center and a plurality of electricity consumption information collection terminals wirelessly connected to the cloud control center to form a remote fault diagnosis network. This embodiment uses the cooperation relationship between a cloud control center and a electricity consumption information collection terminal as an example to illustrate the technical solution of the present invention.

[0052] like Figure 1 As shown, the log acquisition module obtains the original operation log of the power consumption information collection terminal and sends it to the preprocessing module. The preprocessing module obtains abnormal target message sequences based on the original operation log, classifies the target message sequences, and packages them into an original data set.

[0053] The target message sequence is an abnormal sequence. In order to ensure the accuracy of subsequent model training and improve the scientificity of the model's prediction, it is necessary to ensure that the number of extracted target message sequences is no less than 1,000.

[0054] Next, before generating the original data set, the fault type corresponding to the target message sequence needs to be classified. The classification method can be engineer experience judgment or equipment manual comparison classification. The classification operation class can be used to label the fault type label of each original data set for distinction.

[0055] Specifically, in this embodiment, the fault types corresponding to the target message sequence include 12 types: meter data collection failure, data storage failure, terminal data reading failure, terminal data reporting failure, data call failure, event collection failure, event reporting failure, event storage failure, event recording failure, event call failure, overdue control, and unchanged status after control.

[0056] The dataset preparation module is used to obtain the original dataset and divide it into training, test, and validation sets according to a preset ratio. In this embodiment, the preset ratio is set to 14:3:3, meaning that 70% of the total dataset is used as the training set, 15% as the test set, and 15% as the validation set. This reasonable division of the dataset ensures the effectiveness of model training and the accuracy of model performance evaluation.

[0057] The BLSTM-GRU model is a pre-built bidirectional long short-term memory network-gated recurrent unit hybrid network, such as Figure 3 As shown in the figure, the network architecture is specifically as follows: sequential input layer, word embedding layer, BLSTM layer, first dropout layer, GRU layer, second dropout layer, LSTM layer, first fully connected layer and second fully connected layer connected in sequence.

[0058] The Ciphertext sequence input layer is used to input the sequentially arranged target message sequence into the network. The word embedding layer then embeds the features into the BLSTM layer. The BLSTM layer processes the sequence data in both the forward and backward directions to capture the bidirectional information in the sequence.

[0059] A first dropout layer is also placed between the BLSTM and GRU layers to prevent overfitting. (BLSTM layers learn the dependencies and dynamics of sequential data in a bidirectional manner, which is important for learning the discriminative features of the data at each time step. Dropout layers are often used in deep learning methods to prevent overfitting.) The first dropout layer randomly drops a certain number of neurons to improve the network's generalization ability. This prevents overfitting.

[0060] The output of the BLSTM layer is then used as the input to the GRU layer. In this example, the GRU layer of the model has 200 hidden GRU units. The GRU is another type of recurrent neural network unit that is more efficient than the traditional LSTM unit. By combining the input gate and forget gate in the LSTM, the GRU simplifies the structure and reduces unnecessary computation. It controls the flow of information through reset and update gates, enabling the model to learn more efficiently and extract contextual features at a lower computational cost than the LSTM.

[0061] Next comes the second dropout layer, followed by a regular LSTM layer with 200 hidden units. This is followed by a first fully connected layer (FC) with 60 neurons. The final layer is a second fully connected layer (FC) with the same number of neurons as the number of classes in the dataset.

[0062] Finally, the Softmax function is used to generate the probability of each fault category. The Softmax function is as follows:

[0063]

[0064] Where z represents the input vector, k represents the number of fault types, N represents the total number of data sets, and i represents the data set number.

[0065] Therefore, k represents the number of fault types. For different fault types, the corresponding number of fault types is set when the BLSTM-GRU model is arranged; N represents the total number of data sets, that is, the number of training sets, test sets, or validation sets input to the model; i represents the data set number, that is, the number of the specific training set among all the training sets corresponding to the above input.

[0066] Use MATLAB or Python's deep learning library to build a BLSTM-GRU network;

[0067] The BLSTM-GRU network model input, output and parameter settings are as follows:

[0068] Input: target message sequence;

[0069] Output: probability distribution of each fault;

[0070] Learning rate, batch size, embedding dimension.

[0071] The following table shows the BLSTM-GRU network model parameter settings.

[0072] Table 1 BLSTM-GRU network model parameter setting table

[0073]

[0074]

[0075] The BLSTM-GRU model training module is used to iteratively train the constructed BLSTM-GRU model using the training set, update the model parameters, and evaluate it using the test set and validation set. A cross-validation strategy is used during training to ensure the generalization ability of the model and obtain the optimal BLSTM-GRU model after training.

[0076] Deploy the optimal BLSTM-GRU model to the cloud control center.

[0077] The fault diagnosis module is connected to the cloud control center and is used to wirelessly obtain the on-site operation log of the electricity consumption information collection terminal in real time. Specifically, it periodically downloads the on-site operation log of the electricity consumption information collection terminal through wireless transmission, eliminating the time-consuming problem of manually downloading large amounts of data at one time; the collection period can be set to 1h, 2h, 6h, 8h, 12h, 24h, etc. according to the actual user type or maintenance period; and sends it to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis; the optimal BLSTM-GRU model automatically generates a predictive diagnosis result and then sends it to the fault diagnosis module. The predictive diagnosis result is a fault type with the highest predicted probability.

[0078] The fault reporting module is connected to the fault diagnosis module, and is used to receive the predicted diagnosis result sent by the fault diagnosis module, that is, the predicted fault type, and send it to the personal mobile terminal of the operation and maintenance personnel through wireless communication.

[0079] Operation and maintenance personnel can directly carry out targeted repairs on users' electricity meters based on the fault type displayed on their personal mobile terminals.

[0080] In another possible embodiment, Figure 2 As shown, the present invention also discloses a remote fault diagnosis method for an electricity consumption information collection terminal based on BLSTM-GRU, the method comprising:

[0081] S1. Obtain the original operation log of the electricity consumption information collection terminal and send it to the preprocessing module;

[0082] S2. The preprocessing module obtains abnormal target message sequences according to the original operation log, classifies the target message sequences, and packages them to form an original data set;

[0083] S3, the data set preparation module obtains the original data set and divides it into training set, test set and validation set according to the preset quantity ratio;

[0084] S4. Use the training set to iteratively train the constructed BLSTM-GRU model, update the model parameters, and use the test set and validation set for evaluation to obtain the optimal BLSTM-GRU model after training.

[0085] S5. Deploy the optimal BLSTM-GRU model to the cloud control center;

[0086] S6. The fault diagnosis module wirelessly obtains the on-site operation log of the power consumption information collection terminal in real time, and sends it to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis to obtain the predicted diagnosis result.

[0087] The above description is merely an illustration of the preferred embodiments disclosed in this application and the technical principles employed. Those skilled in the art should understand that the scope of protection provided by this disclosure is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned disclosed concepts. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this disclosure.

[0088] In addition, although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of this disclosure. Certain features described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination.

Claims

1. A remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model, comprising a power consumption information collection terminal and a cloud control center, characterized in that: Also includes: The log acquisition module is used to obtain the original operation log of the power consumption information collection terminal and send it to the preprocessing module; A preprocessing module, used for obtaining an abnormal target message sequence according to the original operation log, classifying the target message sequence, and packaging it to form an original data set; The data set preparation module is used to obtain the original data set and divide it into training set, test set and validation set according to the preset quantity ratio; A BLSTM-GRU model training module is used to iteratively train the constructed BLSTM-GRU model using the training set, update model parameters, and evaluate using the test set and the validation set to obtain the optimal BLSTM-GRU model after training; The fault diagnosis module is used to wirelessly obtain the on-site operation log of the power consumption information collection terminal in real time, and send it to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis to obtain the prediction diagnosis result.

2. According to claim 1, a remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model, characterized in that: The specific method for the preprocessing module to classify the target message sequence is: equipment manual comparison classification method; specifically, manually comparing the message sequence fault schematic code in the production manual of the power consumption information collection terminal with the data points in the target message sequence to obtain the fault type of the target message sequence.

3. According to claim 1, a remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model, characterized in that: The network architecture of the BLSTM-GRU model is specifically: a sequential input layer, a word embedding layer, a BLSTM layer, a first dropout layer, a GRU layer, a second dropout layer, an LSTM layer, a first fully connected layer, and a second fully connected layer connected in sequence.

4. According to claim 3, a remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model is characterized in that: The second fully connected layer uses a Softmax function to generate the probability of each fault category; the Softmax function is specifically: Where z represents the input vector, k represents the number of fault types, N represents the total number of data sets, and i represents the data set number.

5. According to claim 1, a remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model, characterized in that: This type of remote fault diagnosis system for electricity consumption information collection terminals also includes a fault reporting module, which is used to receive the predictive diagnosis results sent by the fault diagnosis module and send them to the personal mobile terminal of the operation and maintenance personnel through wireless communication.

6. According to claim 1, a remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model, characterized in that: The fault types corresponding to the target message sequence include: meter data collection failure, data storage failure, terminal data reading failure, terminal data reporting failure, data call failure, event collection failure, event reporting failure, event storage failure, event recording failure, event call failure, overdue control, and unchanged status after control.

7. According to claim 4, a remote fault diagnosis system for power consumption information collection terminal based on BLSTM-GRU model is characterized in that: The preset quantity ratio corresponding to the data set preparation module when the training set, the test set and the validation set are obtained by dividing the data set preparation module according to the preset quantity ratio is specifically 14:3:

3.

8. A remote fault diagnosis method for power consumption information collection terminal based on BLSTM-GRU, characterized in that: The method includes: S1. Obtain the original operation log of the electricity consumption information collection terminal and send it to the preprocessing module; S2. The preprocessing module obtains an abnormal target message sequence according to the original operation log, classifies the target message sequence, and packages it to form an original data set; S3, the data set preparation module obtains the original data set and divides it into a training set, a test set, and a validation set according to a preset quantity ratio; S4. Use the training set to iteratively train the constructed BLSTM-GRU model, update the model parameters and use the test set and validation set for evaluation to obtain the optimal BLSTM-GRU model after training. S5. Deploy the optimal BLSTM-GRU model to the cloud control center; S6. The fault diagnosis module wirelessly obtains the on-site operation log of the power consumption information collection terminal in real time, and sends it to the optimal BLSTM-GRU model deployed in the cloud control center for fault diagnosis to obtain the predictive diagnosis result.

Images