Power distribution automation terminal comprehensive detection system and method

Abstract

The invention discloses a power distribution automation terminal comprehensive detection system and method, and relates to the field of power distribution automation terminal comprehensive detection, and the method comprises the steps: building a power distribution automation terminal main feature data set based on the historical data of a power distribution automation terminal; based on Weibull analysis and in combination with a CNN neural network algorithm, the precision and reliability of equipment life detection are improved; a power distribution automation terminal stability comprehensive evaluation model is constructed, and accurate evaluation and scientific judgment of the terminal stability condition are achieved; according to the equipment life cycle, combining the stability evaluation result, and dynamically optimizing the distribution automation terminal stability comprehensive evaluation model; and dynamically optimizing a power distribution automation regulation and control mode according to a communication feedback result of the power distribution automation terminal and the master station in combination with a stability evaluation result so as to adapt to operation requirements. The method has the advantages that power distribution terminal life prediction, stability evaluation and regulation and control mode dynamic optimization are realized, and the overall performance of comprehensive detection of the power distribution automation terminal is improved.

Description

Technical Field

[0001] The present invention relates to the field of comprehensive detection of distribution automation terminals, and in particular to a comprehensive detection system and method for distribution automation terminals. Background Art

[0002] As a key component of the smart grid, the operating status of the distribution automation system's terminal equipment directly affects the reliability and safety of the distribution network. As the scale of the distribution network expands and the level of intelligence increases, the distribution automation terminals face more complex operating environments and longer service cycles, resulting in equipment aging and increased probability of failure. According to statistics, the accidental failure of the distribution terminal will cause more than 30% of unplanned power outages in the distribution network. Therefore, the life prediction, stability assessment and regulation optimization of terminal equipment have become core needs of the industry.

[0003] Current technological development shows a trend of multi-dimensional integration: on the one hand, Weibull analysis, as a classic method in reliability engineering, can model the equipment life cycle through failure data, but traditional Weibull analysis has difficulty capturing the nonlinear aging characteristics of equipment operation. On the other hand, CNN neural networks have shown advantages in feature extraction and pattern recognition, and can mine deep aging characteristics from multi-dimensional monitoring data to provide more accurate parameter input for the Weibull model. In addition, the stability assessment of distribution terminals requires the integration of multi-source data such as equipment status, environmental stress, and communication feedback. The traditional single indicator evaluation method can no longer meet the needs of complex scenarios, and the optimization of dynamic control modes relies more on the deep integration of life cycle and stability assessment. Secondly, in the interaction process between distribution automation terminals and master stations, they tend to obey the global control of the master station and easily ignore the local stability of the distribution automation terminal, which leads to unstable operation of the distribution automation terminal. In this context, building a comprehensive detection system of "data-driven-model fusion-intelligent control" has become the key to solving the problem of distribution terminal reliability management.

[0004] Current technology in distribution automation terminal detection tends to target the data source of a single device and lacks comprehensive multi-data fusion analysis, resulting in one-sided distribution automation terminal detection results and difficulty in reflecting the overall distribution automation terminal detection quality. Secondly, in the interaction process between the distribution automation terminal and the master station, it tends to obey the global control of the master station and easily ignores the local stability of the distribution automation terminal, which leads to unstable operation of the distribution automation terminal. Summary of the Invention

[0005] In order to solve the above technical problems, a comprehensive detection system and method for distribution automation terminals are provided. This technical solution solves the problem proposed in the above background technology that there is a lack of comprehensive multi-data fusion analysis, which leads to one-sided detection results of distribution automation terminals and makes it difficult to reflect the overall detection quality of distribution automation terminals. Secondly, in the interaction process between the distribution automation terminal and the master station, it tends to obey the global control of the master station and easily ignores the local stability of the distribution automation terminal, which leads to unstable operation of the distribution automation terminal.

[0006] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0007] A comprehensive detection method for distribution automation terminals, comprising:

[0008] Based on the historical data of distribution automation terminals, the main feature data that affects the comprehensive detection of distribution automation terminals is obtained, and the main feature data set of distribution automation terminals is established;

[0009] Based on Weibull analysis, the life cycle of distribution automation terminal equipment is predicted, and optimized with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection;

[0010] Build a comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data to achieve accurate assessment and scientific judgment of terminal stability status;

[0011] Dynamically optimize the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability;

[0012] Based on the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the comprehensive evaluation model of the distribution automation terminal stability, the distribution automation control mode is dynamically optimized to adapt to the operation requirements.

[0013] Preferably, the method of predicting the life cycle of distribution automation terminal equipment based on Weibull analysis and optimizing it in combination with a CNN neural network algorithm to improve the accuracy and reliability of equipment life detection specifically includes:

[0014] Based on the self-test message of the distribution automation terminal equipment, obtain the electrical data and environmental data of the distribution automation terminal equipment;

[0015] Based on the CNN neural network algorithm, feature extraction is performed on the electrical data and environmental data of distribution automation terminal equipment;

[0016] The features output by the CNN neural network algorithm are mapped into Weibull parameters through the fully connected layer to obtain dynamic characteristic life parameters and shape parameters;

[0017] Based on the dynamic characteristic life parameters and shape parameters, a multi-task loss function is constructed to minimize the difference between the Weibull parameter prediction value and the actual data fitting parameters;

[0018] Based on Weibull analysis, the failure probability density value, survival probability value and failure rate value of each device in the distribution automation terminal are obtained, thereby predicting the life of each device in the distribution automation terminal;

[0019] The dynamic characteristic life parameter and characteristic life parameter expression are:

[0020]

[0021] Where, In order to map the features output by the CNN neural network algorithm to the dynamic shape parameter value of Weibull through the fully connected layer, In order to map the features output by the CNN neural network algorithm into the dynamic characteristic life parameter value of Weibull through the fully connected layer, is the Sigmoid activation function, is an exponential function, 、 is the learnable projection matrix, is the feature vector extracted by the CNN neural network algorithm, 、 is the bias term;

[0022] The multi-task loss function expression is:

[0023]

[0024] Where, is the comprehensive loss value of dynamic characteristic life parameter and shape parameter, is the number of samples, To map the output features of the CNN neural network algorithm to the dynamic shape parameters of Weibull through the fully connected layer values, To map the features output by the CNN neural network algorithm into the dynamic characteristic life parameters of Weibull through the fully connected layer values, is the shape parameter target value, Characteristic life parameter target value, The first feature vectors, is the equilibrium constant term, Regularization for extracting feature vectors for CNN neural network algorithms;

[0025] The expression for predicting the life of each device in the distribution automation terminal is:

[0026]

[0027] Where, is the predicted value of the average remaining life of each device in the distribution automation terminal, is the gamma function, which is used to calculate the gamma value related to the shape parameter. When it is an integer, it can be simplified to factorial operation. The time the device has been running.

[0028] Preferably, the construction of a comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data to achieve accurate evaluation and scientific judgment of the terminal stability status specifically includes:

[0029] Based on the main feature data set of distribution automation terminals, the main feature data of lines, transformers and switches are extracted to establish training sets, test sets and validation sets respectively;

[0030] Based on the LSTM neural network model, a four-layer LSTM hierarchical structure is constructed;

[0031] Based on the Pearson correlation formula, the correlation coefficients between the main feature data of lines, transformers, and switches are calculated to construct cross-device correlation features;

[0032] Based on a four-layer LSTM neural network, a comprehensive evaluation model for distribution automation terminal stability is constructed to accurately evaluate and scientifically determine the stability of lines, transformers, switches, and terminals.

[0033] The four-layer LSTM hierarchical structure is specifically as follows:

[0034] The first layer: the temporal stability prediction layer of the line main characteristic data;

[0035] The second layer: the temporal stability prediction layer of the transformer main characteristic data;

[0036] The third layer: the timing stability prediction layer of the switch main characteristic data;

[0037] The fourth layer: the correlation main feature data processing layer.

[0038] Preferably, the dynamically optimizing the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of the distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability specifically includes:

[0039] According to the life expression of each device in the distribution automation terminal, the predicted value of the average remaining life of each device in the distribution automation terminal is obtained;

[0040] According to the comprehensive evaluation model of distribution automation terminal stability, obtain the stability score results of each layer in the comprehensive evaluation model of distribution automation terminal stability;

[0041] Based on the predicted value of the average remaining life of each device and the stability score results of each layer, a comprehensive evaluation and compensation formula for distribution automation terminal stability is constructed;

[0042] Based on the comprehensive evaluation and compensation formula for distribution automation terminal stability, the comprehensive evaluation value of distribution automation terminal stability is dynamically optimized to eliminate the impact of equipment life;

[0043] The comprehensive evaluation and compensation formula for distribution automation terminal stability is:

[0044]

[0045] Where, is the revised comprehensive evaluation value of distribution automation terminal stability, is the correction constant term, is the number of distribution automation terminal devices, For the The impact weight of the equipment life, For distribution automation terminal The predicted value of the average remaining life of each device, For distribution automation terminal The target value of the equipment life.

[0046] Preferably, the dynamically optimizing the distribution automation control mode to adapt to the operation requirements based on the communication feedback results between the distribution automation terminal and the master station and the evaluation results of the comprehensive evaluation model for the stability of the distribution automation terminal specifically includes:

[0047] Obtain communication feedback data between the distribution automation terminal and the master station based on the communication message between the distribution automation terminal and the master station;

[0048] Based on the communication feedback data between the distribution automation terminal and the master station, the input parameter adjustment value of the comprehensive evaluation model of the distribution automation terminal stability is obtained;

[0049] By inputting parameter adjustment amounts, the comprehensive stability evaluation model for distribution automation terminals is used to evaluate the stability of each layer and determine whether the distribution automation terminals are stable after adjustment.

[0050] Set stability boundary thresholds based on historical data or big data analysis;

[0051] Determine whether the absolute difference between the comprehensive evaluation value of the distribution automation terminal stability and the adjusted comprehensive evaluation value of the distribution automation terminal stability is greater than the stability boundary threshold. If so, it means that the current adjustment amount will cause the distribution automation terminal to be unstable and further adjustment is required. If not, it means that the current adjustment amount will not affect the stability of the distribution automation terminal;

[0052] Set the adaptive step size, divide the input parameter adjustment into several small adjustment amounts, and obtain the stability values ​​corresponding to different small adjustment amounts through the comprehensive evaluation model of distribution automation terminal stability;

[0053] Based on the PID algorithm, the distribution automation control mode is dynamically optimized to adapt to the operation requirements through the stability values ​​corresponding to different small adjustment amounts and the original stability values.

[0054] Furthermore, this solution proposes a distribution automation terminal comprehensive detection system for implementing the above-mentioned distribution automation terminal comprehensive detection method, including:

[0055] A data processing module, the data processing module is used to obtain main feature data that affects the comprehensive detection of the distribution automation terminal based on the historical data of the distribution automation terminal, and to establish a main feature data set of the distribution automation terminal;

[0056] A stability assessment module is used to predict the life cycle of distribution automation terminal equipment based on Weibull analysis and optimize it in combination with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection. A comprehensive stability assessment model for distribution automation terminals is constructed based on multi-dimensional data to achieve accurate assessment and scientific judgment of terminal stability status. The comprehensive stability assessment model for distribution automation terminals is dynamically optimized based on the life cycle of distribution automation terminal equipment and the assessment results of the comprehensive stability assessment model for distribution automation terminals.

[0057] A dynamic control module is used to dynamically optimize the distribution automation control mode to adapt to operating requirements based on the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the distribution automation terminal stability comprehensive evaluation model.

[0058] Preferably, the stability assessment module includes:

[0059] A life detection unit, which is used to predict the life cycle of distribution automation terminal equipment based on Weibull analysis and optimize it in combination with a CNN neural network algorithm to improve the accuracy and reliability of equipment life detection;

[0060] A stability assessment unit, which is used to construct a comprehensive stability assessment model for distribution automation terminals based on multi-dimensional data to achieve accurate assessment and scientific judgment of the terminal stability status;

[0061] The stability optimization unit is used to dynamically optimize the distribution automation terminal stability comprehensive evaluation model based on the life cycle of the distribution automation terminal equipment and the evaluation results of the distribution automation terminal stability comprehensive evaluation model.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The present invention provides a comprehensive detection method for distribution automation terminals. The method improves the accuracy and reliability of equipment life detection by combining the basic analysis capability of equipment life of Weibull analysis with the capture capability of complex nonlinear relationships of CNN neural network algorithm. Secondly, based on a four-layer LSTM neural network, a comprehensive evaluation model for the stability of distribution automation terminals is constructed, and a comprehensive analysis is performed from local stability to correlation stability. In combination with the life cycle of distribution automation terminal equipment, the comprehensive evaluation model for the stability of distribution automation terminals is dynamically optimized to eliminate the influence of equipment life. Finally, according to the feedback result of communication between the distribution automation terminal and the master station, combined with the evaluation result of the comprehensive evaluation model for the stability of distribution automation terminals, an adaptive step size is set, and the input parameter adjustment amount is divided into several small adjustment amounts. Based on the PID algorithm, the distribution automation control mode is dynamically optimized to adapt to the operation requirements through the stability values ​​corresponding to different small adjustment amounts and the original stability values, thereby realizing the life prediction, stability evaluation and dynamic optimization of the control mode of the distribution automation terminal, and improving the overall performance of the comprehensive detection of distribution automation terminals. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flow chart of a comprehensive detection method for distribution automation terminals of the present invention;

[0065] Figure 2 This is a flowchart of the present invention's method of predicting the life cycle of distribution automation terminal equipment based on Weibull analysis and optimizing it with a CNN neural network algorithm to improve the accuracy and reliability of equipment life detection.

[0066] Figure 3 This is a flowchart for constructing a comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data to achieve accurate evaluation and scientific judgment of terminal stability status;

[0067] Figure 4 This is a flow chart of the present invention for dynamically optimizing the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability;

[0068] Figure 5 The present invention dynamically optimizes the distribution automation control mode to adapt to the operation demand flow chart based on the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the distribution automation terminal stability comprehensive evaluation model. DETAILED DESCRIPTION

[0069] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0070] Reference Figure 1 As shown, a comprehensive detection method for distribution automation terminals includes:

[0071] Based on the historical data of distribution automation terminals, the main feature data that affects the comprehensive detection of distribution automation terminals is obtained, and the main feature data set of distribution automation terminals is established;

[0072] Based on Weibull analysis, the life cycle of distribution automation terminal equipment is predicted, and optimized with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection;

[0073] Build a comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data to achieve accurate assessment and scientific judgment of terminal stability status;

[0074] Dynamically optimize the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability;

[0075] Based on the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the comprehensive evaluation model of the distribution automation terminal stability, the distribution automation control mode is dynamically optimized to adapt to the operation requirements.

[0076] It can be explained that Weibull analysis, as a classic method of reliability engineering, can model the equipment life cycle through failure data, but traditional Weibull analysis is difficult to capture the nonlinear aging characteristics of equipment operation. Secondly, in the interaction between the distribution automation terminal and the master station, it tends to obey the global control of the master station and easily ignores the local stability of the distribution automation terminal, which leads to the instability of the distribution automation terminal operation. Therefore, this solution combines the basic analysis ability of equipment life of Weibull analysis with the ability of CNN neural network algorithm to capture complex nonlinear relationships to improve the accuracy and reliability of equipment life detection. Secondly, based on the four-layer LSTM neural network, a comprehensive evaluation model for the stability of the distribution automation terminal is constructed, from the perspective of local stability. A comprehensive analysis is conducted from qualitative to associative stability, and combined with the life cycle of distribution automation terminal equipment, the comprehensive evaluation model of distribution automation terminal stability is dynamically optimized to eliminate the impact of equipment life. Finally, according to the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the comprehensive evaluation model of distribution automation terminal stability, the adaptive step size is set, and the input parameter adjustment amount is divided into several small adjustment amounts. Based on the PID algorithm, the distribution automation control mode is dynamically optimized to adapt to the operation requirements through the stability values ​​corresponding to different small adjustment amounts and the original stability values, thereby realizing the distribution terminal life prediction, stability evaluation and dynamic optimization of the control mode, and improving the overall performance of the comprehensive detection of distribution automation terminals.

[0077] The process of obtaining the main feature data that affects the comprehensive detection of the distribution automation terminal based on the historical data of the distribution automation terminal and establishing the main feature data set of the distribution automation terminal specifically includes:

[0078] Based on the historical data of distribution automation terminals, obtain the main characteristic data that affects the comprehensive detection of distribution automation terminals;

[0079] The main characteristic data includes: line current, voltage, power electrical parameters, transformer voltage, current, oil temperature, winding temperature operating parameters, switch status, protection action signals and communication messages between the master station;

[0080] Based on the Kalman filter algorithm, the main characteristic data that affects the comprehensive detection of distribution automation terminals is filtered and denoised;

[0081] According to the normalization formula, the data dimension influence between the main characteristic data that affects the comprehensive detection of distribution automation terminals is eliminated;

[0082] Synchronous acquisition technology and zero-value filling are used to ensure the continuity of the data sequence in the time dimension and the uniformity of the bit width;

[0083] Based on the main feature data that affects the comprehensive detection of distribution automation terminals, a multi-channel multidimensional dataset is constructed, and then integrated to form the main feature dataset of distribution automation terminals;

[0084] Among them, in the multi-channel multidimensional data set, multi-channel refers to the data sources of lines, transformers, switches and master stations, and multi-dimensional data covers the detection parameter dimensions of each channel, as well as the time and space dimensions.

[0085] It can be explained that historical data analysis is an important step in the comprehensive detection and analysis of distribution automation terminals. In order to ensure the integrity and time consistency of the power equipment operation data, the line current, voltage, power and other electrical parameters and the transformer voltage, current, oil temperature, winding temperature and other operating parameters adopt synchronous acquisition technology, strictly align the timestamps, and the switch status, protection action signals and communication messages with the master station are processed through interpolation algorithms, and the time series positions without state changes are filled with zero values ​​to ensure the continuity of the data sequence in the time dimension and the uniformity of the bit width, providing a high-quality time series data set for subsequent analysis.

[0086] Reference Figure 2 As shown, the life cycle of distribution automation terminal equipment is predicted based on Weibull analysis, and optimized in combination with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection. Specifically, the following are included:

[0087] Based on the self-test message of the distribution automation terminal equipment, obtain the electrical data and environmental data of the distribution automation terminal equipment;

[0088] Based on the CNN neural network algorithm, feature extraction is performed on the electrical data and environmental data of distribution automation terminal equipment;

[0089] The features output by the CNN neural network algorithm are mapped into Weibull parameters through the fully connected layer to obtain dynamic characteristic life parameters and shape parameters;

[0090] Based on the dynamic characteristic life parameters and shape parameters, a multi-task loss function is constructed to minimize the difference between the Weibull parameter prediction value and the actual data fitting parameters;

[0091] Based on Weibull analysis, the failure probability density value, survival probability value and failure rate value of each device in the distribution automation terminal are obtained, thereby predicting the life of each device in the distribution automation terminal;

[0092] The dynamic characteristic life parameter and characteristic life parameter expression are:

[0093]

[0094] Where, In order to map the features output by the CNN neural network algorithm to the dynamic shape parameter value of Weibull through the fully connected layer, In order to map the features output by the CNN neural network algorithm into the dynamic characteristic life parameter value of Weibull through the fully connected layer, is the Sigmoid activation function, is an exponential function, 、 is the learnable projection matrix, is the feature vector extracted by the CNN neural network algorithm, 、 is the bias term;

[0095] The multi-task loss function expression is:

[0096]

[0097] Where, is the comprehensive loss value of dynamic characteristic life parameter and shape parameter, is the number of samples, To map the output features of the CNN neural network algorithm to the dynamic shape parameters of Weibull through the fully connected layer values, To map the features output by the CNN neural network algorithm into the dynamic characteristic life parameters of Weibull through the fully connected layer values, is the shape parameter target value, Characteristic life parameter target value, The first feature vectors, is the equilibrium constant term, Regularization for extracting feature vectors for CNN neural network algorithms;

[0098] The expression for predicting the life of each device in the distribution automation terminal is:

[0099]

[0100] Where, is the predicted value of the average remaining life of each device in the distribution automation terminal, is the gamma function, which is used to calculate the gamma value related to the shape parameter. When it is an integer, it can be simplified to factorial operation. The time the device has been running.

[0101] It can be explained that Weibull analysis, as a classic method of reliability engineering, can model the equipment life cycle through failure data, but traditional Weibull analysis is difficult to capture the nonlinear aging characteristics of equipment operation. Therefore, this scheme uses the CNN neural network algorithm to extract features of the electrical data and environmental data of distribution automation terminal equipment, and maps the features output by the CNN neural network algorithm into Weibull parameters through the fully connected layer to obtain dynamic characteristic life parameters and shape parameters, so that it can adapt to complex or nonlinear data and improve the accuracy and reliability of equipment life detection.

[0102] Reference Figure 3 As shown, the comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data is constructed to achieve accurate evaluation and scientific judgment of the terminal stability status, specifically including:

[0103] Based on the main feature data set of distribution automation terminals, the main feature data of lines, transformers and switches are extracted to establish training sets, test sets and validation sets respectively;

[0104] Based on the LSTM neural network model, a four-layer LSTM hierarchical structure is constructed;

[0105] Based on the Pearson correlation formula, the correlation coefficients between the main feature data of lines, transformers, and switches are calculated to construct cross-device correlation features;

[0106] Based on a four-layer LSTM neural network, a comprehensive evaluation model for distribution automation terminal stability is constructed to accurately evaluate and scientifically determine the stability of lines, transformers, switches, and terminals.

[0107] The four-layer LSTM hierarchical structure is specifically as follows:

[0108] The first layer: the temporal stability prediction layer of the line main characteristic data;

[0109] The second layer: the temporal stability prediction layer of the transformer main characteristic data;

[0110] The third layer: the timing stability prediction layer of the switch main characteristic data;

[0111] The fourth layer: the correlation main feature data processing layer.

[0112] It can be explained that when fusing multi-source data and making accurate assessments and scientific judgments on the terminal stability status based on the LSTM neural network, it is easy to ignore the local stability between devices, resulting in sudden local problems being shared and the overall stability being relatively flat. Therefore, this solution constructs a four-layer LSTM hierarchical structure to obtain overall stability while providing feedback on the stability of each layer, thereby achieving a combination of device-level and system-level evaluations through the four-layer structure, taking into account both local and global stability.

[0113] The construction of a comprehensive evaluation model for distribution automation terminal stability based on a four-layer LSTM neural network specifically includes:

[0114] Based on the stability evaluation results of the LSTM neural network of the first three layers, the stability scores of the outputs of the first three layers are obtained;

[0115] Based on the stability scores output by the first three layers and combined with cross-device correlation features, the fusion feature input of the fourth layer is constructed;

[0116] The loss function uses mean square error, , where is the loss between the true value and the predicted value, is the number of sample groups, For the The true value of the stability score, For the The predicted value of a stability score;

[0117] The fusion feature input expression of the fourth layer is:

[0118]

[0119] Where, is the fusion feature input of the fourth layer, is the splicing function, 、 、 They are the stability scores output by the first three layers of LSTM neural network, is the cross-device correlation feature; express 、 、 The three quantities are along the Dimensional splicing operation.

[0120] What can be explained is that by combining the stability evaluation results of the LSTM neural network of the first three layers with the cross-device correlation features, the fusion feature input of the fourth layer is constructed, and the stability evaluation results of the LSTM neural network of the first three layers and the cross-device correlation features are comprehensively judged and analyzed through the LSTM neural network model of the fourth layer. In this way, while improving the local and global stability, when the data of a single device is missing or abnormal (such as data distortion caused by a switch sensor failure), the fourth layer can cross-validate through the correlation features of other devices (such as the historical correlation between line voltage fluctuations and switch status) to avoid deviations in the evaluation results.

[0121] Reference Figure 4As shown, the dynamic optimization of the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of the distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability specifically includes:

[0122] According to the life expression of each device in the distribution automation terminal, the predicted value of the average remaining life of each device in the distribution automation terminal is obtained;

[0123] According to the comprehensive evaluation model of distribution automation terminal stability, obtain the stability score results of each layer in the comprehensive evaluation model of distribution automation terminal stability;

[0124] Based on the predicted value of the average remaining life of each device and the stability score results of each layer, a comprehensive evaluation and compensation formula for distribution automation terminal stability is constructed;

[0125] Based on the comprehensive evaluation and compensation formula for distribution automation terminal stability, the comprehensive evaluation value of distribution automation terminal stability is dynamically optimized to eliminate the impact of equipment life;

[0126] The comprehensive evaluation and compensation formula for distribution automation terminal stability is:

[0127]

[0128] Where, is the revised comprehensive evaluation value of distribution automation terminal stability, is the correction constant term, is the number of distribution automation terminal devices, For the The impact weight of the equipment life, For distribution automation terminal The predicted value of the average remaining life of each device, For distribution automation terminal The target value of the equipment life.

[0129] It can be explained that the main feature data collected in real time will be affected by equipment aging. Therefore, using this data to conduct a comprehensive evaluation of the stability of the distribution automation terminal may result in a low comprehensive evaluation value of the distribution automation terminal stability, which is easy to trigger false alarms of abnormal alarms. Therefore, it is necessary to compensate for the impact of equipment aging and eliminate the impact of minor equipment aging. Therefore, this solution integrates the equipment life cycle prediction and stability evaluation results to construct a dynamic optimization evaluation model for the compensation formula, thereby realizing accurate "time-varying calibration" of the stability of the distribution automation terminal, eliminating the evaluation deviation caused by different equipment aging stages, and making the evaluation results more in line with the actual operating status of the equipment through the two-way mapping of the life prediction value and the stability score.

[0130] Reference Figure 5 As shown, the dynamic optimization of the distribution automation control mode to adapt to the operation requirements based on the communication feedback results between the distribution automation terminal and the master station and the evaluation results of the comprehensive evaluation model of the distribution automation terminal stability specifically includes:

[0131] Obtain communication feedback data between the distribution automation terminal and the master station based on the communication message between the distribution automation terminal and the master station;

[0132] Based on the communication feedback data between the distribution automation terminal and the master station, the input parameter adjustment value of the comprehensive evaluation model of the distribution automation terminal stability is obtained;

[0133] By inputting parameter adjustment amounts, the comprehensive stability evaluation model for distribution automation terminals is used to evaluate the stability of each layer and determine whether the distribution automation terminals are stable after adjustment.

[0134] Set stability boundary thresholds based on historical data or big data analysis;

[0135] Determine whether the absolute difference between the comprehensive evaluation value of the distribution automation terminal stability and the adjusted comprehensive evaluation value of the distribution automation terminal stability is greater than the stability boundary threshold. If so, it means that the current adjustment amount will cause the distribution automation terminal to be unstable and further adjustment is required. If not, it means that the current adjustment amount will not affect the stability of the distribution automation terminal;

[0136] Set the adaptive step size, divide the input parameter adjustment into several small adjustment amounts, and obtain the stability values ​​corresponding to different small adjustment amounts through the comprehensive evaluation model of distribution automation terminal stability;

[0137] Based on the PID algorithm, the distribution automation control mode is dynamically optimized to adapt to the operation requirements through the stability values ​​corresponding to different small adjustment amounts and the original stability values.

[0138] It can be explained that in the interaction process between the distribution automation terminal and the master station, it tends to obey the global control of the master station and easily ignores the local stability of the distribution automation terminal, which leads to the instability of the distribution automation terminal operation. Therefore, when obeying the global control of the master station, it is necessary to reduce the global control process of the master station and obtain the optimal parameter adjustment through multi-step adaptive decomposition and adjustment.

[0139] The adaptive step size expression is:

[0140]

[0141] Where, For the Adjust the step size, is the initial adjustment step size, is the attenuation coefficient, For the The absolute difference between the secondary comprehensive evaluation value of distribution automation terminal stability and the adjusted comprehensive evaluation value of distribution automation terminal stability.

[0142] Furthermore, based on the same inventive concept as the above-mentioned distribution automation terminal comprehensive detection method, this solution proposes a distribution automation terminal comprehensive detection system, including:

[0143] A data processing module, the data processing module is used to obtain main feature data that affects the comprehensive detection of the distribution automation terminal based on the historical data of the distribution automation terminal, and to establish a main feature data set of the distribution automation terminal;

[0144] A stability assessment module is used to predict the life cycle of distribution automation terminal equipment based on Weibull analysis and optimize it in combination with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection. A comprehensive stability assessment model for distribution automation terminals is constructed based on multi-dimensional data to achieve accurate assessment and scientific judgment of terminal stability status. The comprehensive stability assessment model for distribution automation terminals is dynamically optimized based on the life cycle of distribution automation terminal equipment and the assessment results of the comprehensive stability assessment model for distribution automation terminals.

[0145] A dynamic control module is used to dynamically optimize the distribution automation control mode to adapt to the operation requirements based on the feedback results of the communication between the distribution automation terminal and the master station, combined with the evaluation results of the comprehensive evaluation model of the distribution automation terminal stability;

[0146] The stability assessment module includes:

[0147] A life detection unit, which is used to predict the life cycle of distribution automation terminal equipment based on Weibull analysis and optimize it in combination with a CNN neural network algorithm to improve the accuracy and reliability of equipment life detection;

[0148] A stability assessment unit, which is used to construct a comprehensive stability assessment model for distribution automation terminals based on multi-dimensional data to achieve accurate assessment and scientific judgment of the terminal stability status;

[0149] The stability optimization unit is used to dynamically optimize the distribution automation terminal stability comprehensive evaluation model based on the life cycle of the distribution automation terminal equipment and the evaluation results of the distribution automation terminal stability comprehensive evaluation model.

[0150] In summary, the advantages of the present invention are: realizing life prediction, stability evaluation and dynamic optimization of control mode of distribution terminals, and improving the overall performance of comprehensive detection of distribution automation terminals.

[0151] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A comprehensive detection method for distribution automation terminals, characterized in that: include: Based on the historical data of distribution automation terminals, the main feature data that affects the comprehensive detection of distribution automation terminals is obtained, and the main feature data set of distribution automation terminals is established; Based on Weibull analysis, the life cycle of distribution automation terminal equipment is predicted, and optimized with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection; Build a comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data to achieve accurate assessment and scientific judgment of terminal stability status; Dynamically optimize the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability; Based on the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the comprehensive evaluation model of the distribution automation terminal stability, the distribution automation control mode is dynamically optimized to adapt to the operation requirements.

2. A comprehensive detection method for distribution automation terminals according to claim 1, characterized in that: The process of obtaining the main feature data that affects the comprehensive detection of the distribution automation terminal based on the historical data of the distribution automation terminal and establishing the main feature data set of the distribution automation terminal specifically includes: Based on the historical data of distribution automation terminals, obtain the main characteristic data that affects the comprehensive detection of distribution automation terminals; The main characteristic data includes: line current, voltage, power electrical parameters, transformer voltage, current, oil temperature, winding temperature operating parameters, switch status, protection action signals and communication messages between the master station; Based on the Kalman filter algorithm, the main characteristic data that affects the comprehensive detection of distribution automation terminals is filtered and denoised; According to the normalization formula, the data dimension influence between the main characteristic data that affects the comprehensive detection of distribution automation terminals is eliminated; Synchronous acquisition technology and zero-value filling are used to ensure the continuity of the data sequence in the time dimension and the uniformity of the bit width; Based on the main feature data that affects the comprehensive detection of distribution automation terminals, a multi-channel multidimensional dataset is constructed, and then integrated to form the main feature dataset of distribution automation terminals; Among them, in the multi-channel multidimensional data set, multi-channel refers to the data sources of lines, transformers, switches and master stations, and multi-dimensional data covers the detection parameter dimensions of each channel, as well as the time and space dimensions.

3. A comprehensive detection method for distribution automation terminals according to claim 2, characterized in that: The method of predicting the life cycle of distribution automation terminal equipment based on Weibull analysis and optimizing it with a CNN neural network algorithm to improve the accuracy and reliability of equipment life detection specifically includes: Based on the self-test message of the distribution automation terminal equipment, obtain the electrical data and environmental data of the distribution automation terminal equipment; Based on the CNN neural network algorithm, feature extraction is performed on the electrical data and environmental data of distribution automation terminal equipment; The features output by the CNN neural network algorithm are mapped into Weibull parameters through the fully connected layer to obtain dynamic characteristic life parameters and shape parameters; Based on the dynamic characteristic life parameters and shape parameters, a multi-task loss function is constructed to minimize the difference between the Weibull parameter prediction value and the actual data fitting parameters; Based on Weibull analysis, the failure probability density value, survival probability value and failure rate value of each device in the distribution automation terminal are obtained, thereby predicting the life of each device in the distribution automation terminal; The dynamic characteristic life parameter and characteristic life parameter expression are: Where, In order to map the features output by the CNN neural network algorithm to the dynamic shape parameter value of Weibull through the fully connected layer, In order to map the features output by the CNN neural network algorithm into the dynamic characteristic life parameter value of Weibull through the fully connected layer, is the Sigmoid activation function, is an exponential function, 、 is the learnable projection matrix, It is the feature vector extracted by CNN neural network algorithm. 、 is the bias term; The multi-task loss function expression is: Where, is the comprehensive loss value of dynamic characteristic life parameter and shape parameter, is the number of samples, To map the output features of the CNN neural network algorithm to the dynamic shape parameters of Weibull through the fully connected layer values, To map the features output by the CNN neural network algorithm into the dynamic characteristic life parameters of Weibull through the fully connected layer values, is the shape parameter target values, Characteristic life parameter target value, The first feature vectors, is the equilibrium constant term, Regularization of feature vector extraction for CNN neural network algorithm; The expression for predicting the life of each device in the distribution automation terminal is: Where, is the predicted value of the average remaining life of each device in the distribution automation terminal, is the gamma function, which is used to calculate the gamma value related to the shape parameter. When it is an integer, it can be simplified to factorial operation. The time the device has been running.

4. A comprehensive detection method for distribution automation terminals according to claim 3, characterized in that: The comprehensive evaluation model for distribution automation terminal stability based on multi-dimensional data is constructed to achieve accurate evaluation and scientific judgment of the terminal stability status, specifically including: Based on the main feature data set of distribution automation terminals, the main feature data of lines, transformers and switches are extracted to establish training sets, test sets and validation sets respectively; Based on the LSTM neural network model, a four-layer LSTM hierarchical structure is constructed; Based on the Pearson correlation formula, the correlation coefficients between the main feature data of lines, transformers, and switches are calculated to construct cross-device correlation features; Based on a four-layer LSTM neural network, a comprehensive evaluation model for distribution automation terminal stability is constructed to accurately evaluate and scientifically determine the stability of lines, transformers, switches, and terminals. The four-layer LSTM hierarchical structure is specifically as follows: The first layer: the temporal stability prediction layer of the line main characteristic data; The second layer: the temporal stability prediction layer of the transformer main characteristic data; The third layer: the timing stability prediction layer of the switch main characteristic data; The fourth layer: the correlation main feature data processing layer.

5. A comprehensive detection method for distribution automation terminals according to claim 4, characterized in that: The construction of a comprehensive evaluation model for distribution automation terminal stability based on a four-layer LSTM neural network specifically includes: Based on the stability evaluation results of the LSTM neural network of the first three layers, the stability scores of the outputs of the first three layers are obtained; Based on the stability scores output by the first three layers and combined with cross-device correlation features, the fusion feature input of the fourth layer is constructed; The loss function uses mean square error, , where is the loss between the true value and the predicted value, is the number of sample groups, For the The true value of the stability score, For the The predicted value of a stability score; The fusion feature input expression of the fourth layer is: Where, is the fusion feature input of the fourth layer, is the splicing function, 、 、 They are the stability scores output by the first three layers of LSTM neural network, is the cross-device correlation feature; express 、 、 The three quantities are along the Dimensional splicing operation.

6. A comprehensive detection method for distribution automation terminals according to claim 5, characterized in that: The dynamic optimization of the comprehensive evaluation model for distribution automation terminal stability based on the life cycle of the distribution automation terminal equipment and the evaluation results of the comprehensive evaluation model for distribution automation terminal stability specifically includes: According to the life expression of each device in the distribution automation terminal, the predicted value of the average remaining life of each device in the distribution automation terminal is obtained; According to the comprehensive evaluation model of distribution automation terminal stability, obtain the stability score results of each layer in the comprehensive evaluation model of distribution automation terminal stability; Based on the predicted value of the average remaining life of each device and the stability score results of each layer, a comprehensive evaluation and compensation formula for distribution automation terminal stability is constructed; Based on the comprehensive evaluation and compensation formula for distribution automation terminal stability, the comprehensive evaluation value of distribution automation terminal stability is dynamically optimized to eliminate the impact of equipment life; The comprehensive evaluation and compensation formula for distribution automation terminal stability is: Where, is the revised comprehensive evaluation value of distribution automation terminal stability, is the correction constant term, is the number of distribution automation terminal devices, For the The impact weight of the equipment life, For distribution automation terminal The predicted value of the average remaining life of each device, For distribution automation terminal The target value of the equipment life.

7. A comprehensive detection method for distribution automation terminals according to claim 6, characterized in that: The dynamic optimization of the distribution automation control mode to adapt to the operation requirements based on the communication feedback results between the distribution automation terminal and the master station and the evaluation results of the comprehensive evaluation model of the distribution automation terminal stability specifically includes: Obtain communication feedback data between the distribution automation terminal and the master station based on the communication message between the distribution automation terminal and the master station; Based on the communication feedback data between the distribution automation terminal and the master station, the input parameter adjustment value of the comprehensive evaluation model of the distribution automation terminal stability is obtained; By inputting parameter adjustment amounts, the comprehensive stability evaluation model for distribution automation terminals is used to evaluate the stability of each layer and determine whether the distribution automation terminals are stable after adjustment. Set stability boundary thresholds based on historical data or big data analysis; Determine whether the absolute difference between the comprehensive evaluation value of the distribution automation terminal stability and the adjusted comprehensive evaluation value of the distribution automation terminal stability is greater than the stability boundary threshold. If so, it means that the current adjustment amount will cause the distribution automation terminal to be unstable and further adjustment is required. If not, it means that the current adjustment amount will not affect the stability of the distribution automation terminal; Set the adaptive step size, divide the input parameter adjustment into several small adjustment amounts, and obtain the stability values ​​corresponding to different small adjustment amounts through the comprehensive evaluation model of distribution automation terminal stability; Based on the PID algorithm, the distribution automation control mode is dynamically optimized to adapt to the operation requirements through the stability values ​​corresponding to different small adjustment amounts and the original stability values.

8. A comprehensive detection system for distribution automation terminals, characterized in that: A method for implementing a comprehensive detection method for a distribution automation terminal according to any one of claims 1 to 7, comprising: A data processing module, the data processing module is used to obtain main feature data that affects the comprehensive detection of the distribution automation terminal based on the historical data of the distribution automation terminal, and to establish a main feature data set of the distribution automation terminal; A stability assessment module is used to predict the life cycle of distribution automation terminal equipment based on Weibull analysis and optimize it in combination with the CNN neural network algorithm to improve the accuracy and reliability of equipment life detection. A comprehensive stability assessment model for distribution automation terminals is constructed based on multi-dimensional data to achieve accurate assessment and scientific judgment of terminal stability status. The comprehensive stability assessment model for distribution automation terminals is dynamically optimized based on the life cycle of distribution automation terminal equipment and the assessment results of the comprehensive stability assessment model for distribution automation terminals. A dynamic control module is used to dynamically optimize the distribution automation control mode to adapt to operating requirements based on the communication feedback results between the distribution automation terminal and the master station, combined with the evaluation results of the distribution automation terminal stability comprehensive evaluation model.

9. A distribution automation terminal comprehensive detection system according to claim 8, characterized in that: The stability assessment module includes: A life detection unit, which is used to predict the life cycle of distribution automation terminal equipment based on Weibull analysis and optimize it in combination with a CNN neural network algorithm to improve the accuracy and reliability of equipment life detection; A stability assessment unit, which is used to construct a comprehensive stability assessment model for distribution automation terminals based on multi-dimensional data to achieve accurate assessment and scientific judgment of the terminal stability status; The stability optimization unit is used to dynamically optimize the distribution automation terminal stability comprehensive evaluation model based on the life cycle of the distribution automation terminal equipment and the evaluation results of the distribution automation terminal stability comprehensive evaluation model.

Images