Power grid full-cycle operation safety supervision system based on artificial intelligence

Abstract

The invention discloses a power grid full-cycle operation safety supervision system based on artificial intelligence, and belongs to the field of power grid safety supervision, and the system comprises the steps: collecting historical charge data and corresponding timestamp records, carrying out the preprocessing, and dividing the obtained time series data into a training sample set and a verification sample set; through constructing a prediction model and carrying out training and optimization evaluation, an optimal prediction model is obtained; real-time charge data are collected and input into the optimal prediction model, predicted charge data of the next time step are obtained, the charge change rate is calculated and compared with a preset change threshold value, and whether the charge data of the next time step are abnormal data or normal data is judged; performing fault type prediction on the abnormal data by using a preset random forest model; and periodically traversing the fault record to obtain all fault types, setting a power grid operation safety level, and performing risk assessment on power grid operation by calculating a risk index to obtain the power grid operation safety level.

Description

Technical Field

[0001] The present invention relates to the field of power grid safety supervision, and in particular to an artificial intelligence-based full-cycle power grid operation safety supervision system. Background Art

[0002] The power grid is one of the indispensable infrastructures in modern society, and its stable operation is crucial to ensuring social production and people's lives. With the continuous growth of electricity demand and the continuous expansion of the power grid, the safety of power grid operation has become increasingly prominent.

[0003] Traditional grid safety supervision methods have limitations: Traditional grid safety supervision methods mainly rely on manual inspections and regular maintenance, which is not only inefficient but also difficult to detect potential safety hazards in a timely manner. In addition, with the complexity of the grid structure and the opening of the power market, traditional supervision methods can no longer meet the needs of modern grid safety management. Summary of the invention

[0004] The purpose of the present invention is to provide an artificial intelligence-based full-cycle operation safety supervision system for power grids to address the deficiencies in the background technology.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A power grid full-cycle operation safety supervision system based on artificial intelligence, comprising:

[0006] The data collection and preprocessing module collects the historical charge data of the power grid operation and the corresponding timestamp records; preprocesses the collected data to obtain time series data, and divides the time series data into a training sample set and a verification sample set according to a certain ratio;

[0007] The prediction model building module builds a prediction model by designing the LSTM network structure, and uses the training sample set and the verification sample set to train and optimize the prediction model to obtain the optimal prediction model;

[0008] The fault diagnosis module collects real-time charge data and inputs it into the optimal prediction model, obtains the predicted charge data of the next time step, and calculates the charge change rate; uses the obtained charge change rate and the preset change threshold to make a judgment, and obtains whether the charge data of the next time step is abnormal data or normal data;

[0009] Use the preset random forest model to predict abnormal data, obtain the corresponding labels of different fault types or normal labels, and generate fault records;

[0010] The operation safety assessment module periodically traverses the fault records to obtain all fault types, sets the power grid operation safety level, conducts risk assessment on the power grid operation by calculating the risk index, and obtains the power grid operation safety level.

[0011] Preferably, preprocessing the collected data includes: removing duplicate data, filling missing values ​​with the mean or median, aligning all data with the recorded timestamps, and normalizing the data so that all data are normalized to the range [0,1] to obtain time series data.

[0012] Preferably, the steps of training the prediction model using the training sample set are as follows:

[0013] S1. Design a multi-layer LSTM network structure and represent the time series data of each training sample as X = [x1, x2, ..., x T ]; where x t The feature vector representing time step t, t = 1, 2, ..., T;

[0014] S2, input the time series data in the training sample set into the multi-layer LSTM network, learn the time series features at different levels, and obtain the final hidden state vector;

[0015] S3. Get the final hidden state vector of the multi-layer LSTM, train it through a fully connected layer, and then normalize the output of the fully connected layer using the softmax function to obtain the weight of each final hidden state vector;

[0016] S4, using the trained weights to calculate the weighted average sum of the final hidden state vector of the multi-layer LSTM to obtain the context vector;

[0017] S5 uses a fully connected layer to perform a linear transformation on the context vector, maps the vector output by the attention layer to the required prediction space, and obtains a trained prediction model.

[0018] Preferably, the calculation formula involved in the LSTM network specifically includes:

[0019] Input gate: i t =σ(W i ·[h t-1 ,x t ]+b i );

[0020] In the formula, i t The data generated by the input gate for the current time step, x t is the feature vector of the current time step t; h t-1 is the hidden state of the previous time step; W i is the weight matrix in the input gate; b i is the bias term of the input gate, σ represents the sigmoid activation function;

[0021] Forget gate: f t =σ(W f·[h t-1 ,x t ]+b f );

[0022] In the formula, f t The data generated by the forget gate at the current time step; W f is the weight matrix in the forget gate; b f is the bias term of the forget gate;

[0023] Cell status update:

[0024] In the formula, C t Update the generated data for the cell state at the current time step; represents point-by-point multiplication; tanh is the hyperbolic tangent activation function; W c is the weight matrix in the cell state update; b c is the bias term for cell state update;

[0025] Output gate: o t =σ(W o ·[h t-1 ,x t ]+b o );

[0026] In the formula, o t The data generated by the output gate for the current time step; W o is the weight matrix in the output gate; b o is the bias term of the output gate;

[0027] Generate the final hidden state vector: h t =o t tanh(C t );

[0028] In the formula, h t is the final hidden state vector at the current time step.

[0029] Preferably, the method for optimizing and evaluating the prediction model using the validation sample set is as follows:

[0030] The validation sample set is input into the trained forecasting model to evaluate the model performance. The validation forecast value output by the load forecasting model and the real observed value in the validation sample set are counted. The mean absolute percentage error (MAPE) is selected as the error evaluation indicator.

[0031] The calculation formula is:

[0032]

[0033] In the formula, n is the number of samples in the validation sample set, y mTo verify the true observations in the sample set, is the output verification prediction value, m is the mth sample in the verification sample set, m = 1, 2, ..., n;

[0034] Compare the calculated MAPE with the preset indicator parameters. If MAPE is less than the preset indicator parameters, it means that the performance of the model meets the standard and the optimal prediction model is obtained; otherwise, the model parameters are continuously adjusted until the model performance meets the standard.

[0035] Preferably, the specific implementation content of the fault diagnosis module includes: collecting real-time charge data and inputting it into the optimal prediction model, obtaining the predicted charge data of the next time step, and calculating the charge change rate;

[0036]

[0037] Where DL represents the charge change rate, D Y Denotes the predicted charge data for the next time step, D X Represents the charge data of the current time step;

[0038] The calculated charge change rate DL is compared with the preset change threshold. If the charge change rate DL is greater than the change threshold, it means that the charge data of the next time step is abnormal, there is a potential fault, and it is recorded as abnormal data, and an early warning signal is generated to remind the operation and maintenance personnel to take preventive measures; if the charge change rate DL is less than or equal to the change threshold, it means that the charge data of the next time step is normal and is recorded as normal data;

[0039] The system automatically identifies warning signals and performs fault diagnosis on abnormal data. It uses the preset random forest model to predict abnormal data, obtains corresponding labels of different fault types or normal labels, and generates fault records.

[0040] Preferably, the specific implementation content of the operation safety assessment module includes: periodically traversing the fault records to obtain all fault types, and using the formula FX = ∑GZ j ×θ j Calculate the risk index to conduct risk assessment on power grid operation; FX is the risk index, GZ j The number of times all fault types are diagnosed, θ j is the weight coefficient corresponding to different fault types, j is a positive integer, representing different fault types;

[0041] Set the grid operation safety level to low, medium, and high;

[0042] The obtained risk index is evaluated for safety level. If 0<FX≤0.3, it means that the safety level of power grid operation is high; if 0.3<FX≤0.7, it means that the safety level of power grid operation is medium; if FX>0.7, it means that the safety level of power grid operation is low.

[0043] In order to solve the problem, the present invention also discloses a method for monitoring the safety of power grid full-cycle operation based on artificial intelligence, comprising the following steps:

[0044] Collect historical charge data of power grid operation and corresponding timestamp records, pre-process the collected data to obtain time series data, and divide the time series data into training sample set and verification sample set according to a certain ratio;

[0045] By designing the LSTM network structure and building a prediction model, the prediction model is trained and optimized using the training sample set and the validation sample set to obtain the optimal prediction model.

[0046] Collect real-time charge data and input it into the optimal prediction model, obtain the predicted charge data of the next time step, and calculate the charge change rate; use the obtained charge change rate and the preset change threshold to make a judgment, and obtain whether the charge data of the next time step is abnormal data or normal data;

[0047] Use the preset random forest model to predict abnormal data, obtain the corresponding labels of different fault types or normal labels, and generate fault records;

[0048] Periodically traverse the fault records to obtain all fault types, set the power grid operation safety level, conduct risk assessment on the power grid operation by calculating the risk index, and obtain the power grid operation safety level.

[0049] Compared with the existing solutions, the present invention achieves the following beneficial effects:

[0050] The present invention collects historical charge data and corresponding timestamp records, performs preprocessing, and obtains time series data, which helps to capture the time series characteristics of charge changes and improve the accuracy of model prediction. By designing a long short-term memory network structure and building a prediction model, the limitations of traditional neural networks in processing time series data are effectively solved. Through continuous adjustment and optimization, the optimal prediction model is obtained, ensuring that the model has good generalization ability and prediction performance.

[0051] The present invention realizes real-time monitoring and fault diagnosis of charge data of power grid operation by utilizing the optimal prediction model and the preset random forest model, timely discovers potential safety hazards, formulates more accurate and effective maintenance plans, reduces unnecessary downtime and maintenance costs, and greatly improves the safety performance of the power grid.

[0052] The present invention obtains all fault types by periodically traversing fault records, sets the power grid operation safety level, performs risk assessment on power grid operation by calculating the risk index, and obtains the power grid operation safety level, which is conducive to objectively quantifying the power grid operation safety status and provides a scientific basis for the management of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention will be further described below in conjunction with the accompanying drawings.

[0054] Figure 1 This is a module block diagram of the artificial intelligence-based power grid full-cycle operation safety supervision system of the present invention.

[0055] Figure 2 The present invention is a flowchart of the method for monitoring the safety of power grid operation throughout the entire cycle based on artificial intelligence. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0057] Embodiment 1, as Figure 1 As shown in the figure, the AI-based full-cycle power grid operation safety supervision system includes:

[0058] The data collection and preprocessing module collects the historical charge data of the power grid operation and the corresponding timestamp records; preprocesses the collected data to obtain time series data, and divides the time series data into a training sample set and a verification sample set according to a certain ratio; the charge data includes current, voltage, and power; the certain ratio is usually set to 7:3

[0059] Furthermore, the collected data are preprocessed, including: removing duplicate data, filling missing values ​​with the mean or median, aligning all data with the recorded timestamps, and normalizing the data so that all data are normalized to the range [0,1] to obtain time series data.

[0060] In the embodiment of the present invention, historical charge data of the power grid and the corresponding timestamps are collected and stored to ensure the time sequence of the data; the data is normalized to eliminate the scale differences between different features, which helps to improve the effect of subsequent model training.

[0061] The prediction model building module builds the prediction model by designing the LSTM network structure and trains the prediction model using the training sample set;

[0062] The steps to train a prediction model using a training sample set are as follows:

[0063] S1. Design a multi-layer LSTM network structure and represent the time series data of each training sample as X = [x1, x2, ..., x T ]; where x t The feature vector representing time step t, t = 1, 2, ..., T;

[0064] S2, input the time series data in the training sample set into the multi-layer LSTM network to learn the time series features at different levels, that is, the output of the previous LSTM layer is used as the input of the next layer, and is passed layer by layer to obtain the final hidden state vector;

[0065] Specifically, the LSTM network structure includes an input gate, a forget gate, a cell state update, an output gate, and generates a final hidden state vector;

[0066] Furthermore, the input gate determines how much of the input information of the current time step can enter the cell state, and a sigmoid activation function is used to limit the input value to the range of [0,1] to control the flow of information; the cell state is the memory unit in the LSTM, which transmits information along the time step during the entire processing process;

[0067] The calculation formula is:

[0068] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0069] In the formula, i t The data generated by the input gate for the current time step, x t is the feature vector of the current time step t; h t-1 is the hidden state of the previous time step; W i is the weight matrix in the input gate; b i is the bias term of the input gate, σ represents the sigmoid activation function;

[0070] The forget gate determines how much of the cell state at the previous moment needs to be retained to the current moment, and controls the degree of information retention through the sigmoid activation function;

[0071] The calculation formula is:

[0072] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0073] In the formula, f t The data generated by the forget gate at the current time step; Wf is the weight matrix in the forget gate; b f is the bias term of the forget gate;

[0074] The cell state update is determined by the input gate and the forget gate;

[0075] The calculation formula is:

[0076]

[0077] In the formula, C t Update the generated data for the cell state at the current time step; represents point-by-point multiplication; tanh is the hyperbolic tangent activation function; W c is the weight matrix in the cell state update; b c is the bias term for cell state update;

[0078] The output gate determines how much of the cell state at the current time step can be output to the hidden state;

[0079] The calculation formula is:

[0080] o t =σ(W o ·[h t-1 ,x t ]+b o );

[0081] In the formula, o t The data generated by the output gate for the current time step; W o is the weight matrix in the output gate; b o is the bias term of the output gate;

[0082] The final hidden state vector is generated by processing the cell state at the current time step through the tanh activation function and then multiplying it by the output of the output gate, which represents the output of the LSTM layer at the current time step;

[0083] The calculation formula is:

[0084] h t =o t tanh(C t );

[0085] In the formula, h t is the final hidden state vector of the current time step;

[0086] S3. Get the final hidden state vector of the multi-layer LSTM, train it through a fully connected layer, and then normalize the output of the fully connected layer using the softmax function to obtain the weight of each final hidden state vector; the weight indicates the importance of the final hidden state of each time step to the prediction result;

[0087] The calculation formula is:

[0088] S v =tanh(W a h tv +B a );

[0089] α v =softmax(S v );

[0090] In the formula, S v is the attention score corresponding to the final hidden state vector of each layer of LSTM; W a and B a is the weight matrix and bias term of the attention layer; h tv is the final hidden state vector of the vth layer LSTM corresponding to the current time step t, v = 1, 2, ..., k, k is a positive integer; α v is the weight of the final hidden state vector of each LSTM layer;

[0091] S4, using the trained weights to calculate the weighted average sum of the final hidden state vector of the multi-layer LSTM to obtain the context vector;

[0092] The calculation formula is:

[0093]

[0094] In the formula, C t is the context vector, emphasizing the information most relevant to the current time step;

[0095] S5. Use the fully connected layer to perform a linear transformation on the context vector, map the vector output by the attention layer to the required prediction space, and obtain a trained prediction model.

[0096] The linear transformation formula is:

[0097] y=W d C t +B d ;

[0098] In the formula, y is the output prediction result, W d , B d are the weight matrix and bias term of the linear transformation.

[0099] The validation sample set is input into the trained forecasting model to evaluate the model performance. The validation forecast value output by the load forecasting model and the real observed value in the validation sample set are counted. The mean absolute percentage error (MAPE) is selected as the error evaluation indicator.

[0100] The calculation formula is:

[0101]

[0102] In the formula, n is the number of samples in the validation sample set, y m To verify the true observations in the sample set, is the output verification prediction value, m is the mth sample in the verification sample set, m = 1, 2, ..., n;

[0103] Compare the calculated MAPE with the preset index parameters. If the MAPE is less than the preset index parameters, it means that the performance of the model meets the standard and the optimal prediction model is obtained. Otherwise, the model parameters are continuously adjusted until the model performance meets the standard. The preset index parameters are determined according to the specific application background of the actual model.

[0104] It should be noted that the range of MAPE is [0,∞). When the verified predicted value is completely consistent with the true value, that is, MAPE is equal to 0, the model is a perfect model; the smaller the MAPE value, the higher the accuracy of the model.

[0105] In the embodiment of the present invention, the design of the LSTM network structure is conducive to processing time series data, can capture long-term dependencies, and improve the accuracy of predictions. Through the design of multi-layer LSTM units, the model's ability to recognize nonlinear and complex patterns is enhanced.

[0106] The fault diagnosis module collects real-time charge data and inputs it into the optimal prediction model, obtains the predicted charge data of the next time step, and calculates the charge change rate;

[0107]

[0108] Where DL represents the charge change rate, D Y Denotes the predicted charge data for the next time step, D X Represents the charge data of the current time step;

[0109] The calculated charge change rate DL is compared with the preset change threshold. If the charge change rate DL is greater than the change threshold, it means that the charge data of the next time step is abnormal, there is a potential fault, and it is recorded as abnormal data, and an early warning signal is generated to remind the operation and maintenance personnel to take preventive measures; if the charge change rate DL is less than or equal to the change threshold, it means that the charge data of the next time step is normal and is recorded as normal data;

[0110] The system automatically identifies warning signals and performs fault diagnosis on abnormal data. It uses the preset random forest model to predict abnormal data, obtains labels of different fault types or normal labels, and generates fault records. Fault types include but are not limited to equipment faults, voltage faults, and current faults.

[0111] In some preferred embodiments, the random forest model is constructed as follows:

[0112] According to the historical fault records, historical abnormal data and corresponding fault types are obtained, and key features related to different fault types, such as charge change rate, are extracted; the random forest model is trained using a labeled data set (containing normal labels and labels of different types of faults), and the random forest model parameters are adjusted through a cross-validation method until the performance of the random forest model meets the standards; the specific algorithm formulas involved in the random forest model are not described here.

[0113] Run the safety assessment module, periodically traverse the fault records to obtain all fault types, and use the formula FX = ∑GZ j ×θ j Calculate risk index to conduct risk assessment on power grid operation;

[0114] In the formula, FX is the risk index, GZ j The number of times all fault types are diagnosed, θ j is the weight coefficient corresponding to different fault types, which is determined by experts in the power grid field; j is a positive integer, representing different fault types; the evaluation period is set to one month;

[0115] Set the grid operation safety level to low, medium, and high;

[0116] The obtained risk index is evaluated for safety level. If 0<FX≤0.3, it means that the safety level of power grid operation is high; if 0.3<FX≤0.7, it means that the safety level of power grid operation is medium; if FX>0.7, it means that the safety level of power grid operation is low.

[0117] It should be noted that the grid operation safety level is used to evaluate the safety of the grid operation in the current cycle; if the grid operation safety level is low, the early warning mechanism should be activated in time and pre-specified risk control measures should be taken to reduce risks and ensure the stable operation of the power system; among them, risk control measures include but are not limited to strengthening equipment maintenance and updating, optimizing grid structure and layout, and strengthening safety management and supervision.

[0118] Embodiment 2, as Figure 2 As shown, the artificial intelligence-based power grid full-cycle operation safety supervision method includes the following steps:

[0119] Collect historical charge data of power grid operation and corresponding timestamp records, pre-process the collected data to obtain time series data, and divide the time series data into training sample set and verification sample set according to a certain ratio;

[0120] By designing the LSTM network structure and building a prediction model, the prediction model is trained and optimized using the training sample set and the validation sample set to obtain the optimal prediction model.

[0121] Collect real-time charge data and input it into the optimal prediction model, obtain the predicted charge data of the next time step, and calculate the charge change rate; use the obtained charge change rate and the preset change threshold to make a judgment, and obtain whether the charge data of the next time step is abnormal data or normal data;

[0122] Use the preset random forest model to predict abnormal data, obtain the corresponding labels of different fault types or normal labels, and generate fault records;

[0123] Periodically traverse the fault records to obtain all fault types, set the power grid operation safety level, conduct risk assessment on the power grid operation by calculating the risk index, and obtain the power grid operation safety level.

[0124] In the several embodiments provided by the present invention, it should be understood that the disclosed system can be implemented in other ways. For example, the above-described embodiments of the invention are only illustrative, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation.

[0125] The modules described as separate accessories may or may not be physically separated, and the accessories shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0126] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0127] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. The full-cycle operation safety supervision system of power grid based on artificial intelligence is characterized by: include: Data collection and preprocessing module, which collects historical charge data of power grid operation and corresponding timestamp records; Preprocess the collected data to obtain time series data, and divide the time series data into training sample set and validation sample set according to a certain ratio; The prediction model building module builds a prediction model by designing the LSTM network structure, and uses the training sample set and the verification sample set to train and optimize the prediction model to obtain the optimal prediction model; The fault diagnosis module collects real-time charge data and inputs it into the optimal prediction model, obtains the predicted charge data of the next time step, and calculates the charge change rate; uses the obtained charge change rate and the preset change threshold to make a judgment, and obtains whether the charge data of the next time step is abnormal data or normal data; Use the preset random forest model to predict abnormal data, obtain the corresponding labels of different fault types or normal labels, and generate fault records; The operation safety assessment module periodically traverses the fault records to obtain all fault types, sets the power grid operation safety level, conducts risk assessment on the power grid operation by calculating the risk index, and obtains the power grid operation safety level.

2. The artificial intelligence-based full-cycle power grid operation safety supervision system according to claim 1 is characterized in that: The preprocessing of the collected data includes: removing duplicate data, filling missing values ​​with the mean or median, aligning all data with the recorded timestamps, and normalizing the data so that all data are normalized to the range [0,1] to obtain time series data.

3. The artificial intelligence-based full-cycle power grid operation safety supervision system according to claim 2 is characterized in that: The steps to train a prediction model using a training sample set are as follows: S1. Design a multi-layer LSTM network structure and represent the time series data of each training sample as X = [x1, x2, ..., x T ]; where x t The feature vector representing time step t, t = 1, 2, ..., T; S2, input the time series data in the training sample set into the multi-layer LSTM network, learn the time series features at different levels, and obtain the final hidden state vector; S3. Get the final hidden state vector of the multi-layer LSTM, train it through a fully connected layer, and then normalize the output of the fully connected layer using the softmax function to obtain the weight of each final hidden state vector; S4, using the trained weights to calculate the weighted average sum of the final hidden state vector of the multi-layer LSTM to obtain the context vector; S5 uses a fully connected layer to perform a linear transformation on the context vector, maps the vector output by the attention layer to the required prediction space, and obtains a trained prediction model.

4. The artificial intelligence-based full-cycle power grid operation safety supervision system according to claim 3 is characterized in that: The calculation formulas involved in the LSTM network include: Input gate: i t =σ(W i ·[h t-1 ,x t ]+b i ); In the formula, i t The data generated by the input gate for the current time step, x t is the feature vector of the current time step t; h t-1 is the hidden state of the previous time step; W i is the weight matrix in the input gate; b i is the bias term of the input gate, σ represents the sigmoid activation function; Forget gate: f t =σ(W f ·[h t-1 ,x t ]+b f ); In the formula, f t The data generated by the forget gate for the current time step; W f is the weight matrix in the forget gate; b f is the bias term of the forget gate; Cell status update: In the formula, C t Update the generated data for the cell state at the current time step; represents point-by-point multiplication; tanh is the hyperbolic tangent activation function; W c is the weight matrix in the cell state update; b c is the bias term for cell state update; Output gate: o t =σ(W o ·[h t-1 ,x t ]+b o ); In the formula, o t The data generated by the output gate for the current time step; W o is the weight matrix in the output gate; b o is the bias term of the output gate; Generate the final hidden state vector: h t =o t tanh(C t ); In the formula, h t is the final hidden state vector at the current time step.

5. The artificial intelligence-based full-cycle power grid operation safety supervision system according to claim 3 is characterized in that: The method of using the validation sample set to optimize the prediction model is as follows: The validation sample set is input into the trained forecasting model to evaluate the model performance. The validation forecast value output by the load forecasting model and the real observed value in the validation sample set are counted. The mean absolute percentage error (MAPE) is selected as the error evaluation indicator. The calculation formula is: In the formula, n is the number of samples in the validation sample set, y m To verify the true observations in the sample set, is the output verification prediction value, m is the mth sample in the verification sample set, m = 1, 2, ..., n; Compare the calculated MAPE with the preset indicator parameters. If MAPE is less than the preset indicator parameters, it means that the performance of the model meets the standard and the optimal prediction model is obtained; otherwise, the model parameters are continuously adjusted until the model performance meets the standard.

6. The artificial intelligence-based full-cycle power grid operation safety supervision system according to claim 5 is characterized in that: The specific implementation of the fault diagnosis module includes: collecting real-time charge data and inputting it into the optimal prediction model, obtaining the predicted charge data for the next time step, and calculating the charge change rate; Where DL represents the charge change rate, D Y Denotes the predicted charge data for the next time step, D X Represents the charge data of the current time step; The calculated charge change rate DL is compared with the preset change threshold. If the charge change rate DL is greater than the change threshold, it means that the charge data of the next time step is abnormal, there is a potential fault, and it is recorded as abnormal data, and an early warning signal is generated to remind the operation and maintenance personnel to take preventive measures; if the charge change rate DL is less than or equal to the change threshold, it means that the charge data of the next time step is normal and is recorded as normal data; The system automatically identifies warning signals and performs fault diagnosis on abnormal data. It uses the preset random forest model to predict abnormal data, obtains corresponding labels of different fault types or normal labels, and generates fault records.

7. The artificial intelligence-based full-cycle power grid operation safety supervision system according to claim 6 is characterized in that: The specific implementation of the operation safety assessment module includes: periodically traversing the fault records to obtain all fault types, and using the formula FX = ∑GZ j ×θ j Calculate the risk index to conduct risk assessment on power grid operation; FX is the risk index, GZ j The number of times all fault types are diagnosed, θ j is the weight coefficient corresponding to different fault types, j is a positive integer, representing different fault types; Set the grid operation safety level to low, medium, and high; The obtained risk index is evaluated for safety level. If 0<FX≤0.3, it means that the safety level of power grid operation is high; if 0.3<FX≤0.7, it means that the safety level of power grid operation is medium; if FX>0.7, it means that the safety level of power grid operation is low.

8. A method for monitoring the safety of power grid operation throughout its entire cycle based on artificial intelligence, comprising the following steps: Collect historical charge data of power grid operation and corresponding timestamp records, pre-process the collected data to obtain time series data, and divide the time series data into training sample set and verification sample set according to a certain ratio; By designing the LSTM network structure and building a prediction model, the prediction model is trained and optimized using the training sample set and the validation sample set to obtain the optimal prediction model. Collect real-time charge data and input it into the optimal prediction model, obtain the predicted charge data of the next time step, and calculate the charge change rate; use the obtained charge change rate and the preset change threshold to make a judgment, and obtain whether the charge data of the next time step is abnormal data or normal data; Use the preset random forest model to predict abnormal data, obtain the corresponding labels of different fault types or normal labels, and generate fault records; Periodically traverse the fault records to obtain all fault types, set the power grid operation safety level, conduct risk assessment on the power grid operation by calculating the risk index, and obtain the power grid operation safety level.

Images