Topology identification method and system for low-voltage distribution network based on deep learning

Abstract

The invention provides a topology identification method and system for a low-voltage distribution network based on deep learning, and belongs to the technical field of power grid topology identification, and the method comprises the steps: deploying a multi-source sensing array at a key node position according to the structural characteristics and node distribution of the low-voltage distribution network, collecting first data through the multi-source sensing array, and storing the first data in a database; processing the data into third data; establishing a deep learning model for topology identification of the low-voltage power distribution network, training the deep learning model by using the third data set to obtain a trained deep learning model, and outputting a topological structure prediction result by using the trained deep learning model; a topological structure prediction result is displayed, and connection lines between the nodes are drawn according to the topological relation; and identifying the electrical quantity data and the environmental data of the line. Based on the method, the invention further provides a topology identification system of the low-voltage power distribution network based on deep learning. According to the invention, high-precision and high-adaptability identification of the low-voltage power distribution topological structure is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid topology identification, and in particular relates to a topology identification method and system for a low-voltage distribution network based on deep learning. Background Art

[0002] The low-voltage distribution network is an important part of the power system, responsible for transmitting electrical energy from the substation to the final power user. In modern power systems, the complexity of low-voltage distribution networks is increasing, and the accurate identification of their topology is critical for the optimal operation, fault diagnosis and rapid recovery of the distribution system.

[0003] Traditional topology identification methods mostly rely on manual experience and simple electrical quantity measurements, such as methods based on current amplitude comparison. In complex multi-branch, variable load low-voltage distribution networks, it is easy to cause topology identification errors due to factors such as load fluctuations and line impedance changes. In addition, although some existing automated topology identification technologies have introduced sensors and data analysis methods, there is still much room for improvement in data fusion processing and algorithm accuracy, which makes it difficult to meet the needs of smart grids for refined management of low-voltage distribution. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a topology identification method and system for a low-voltage distribution network based on deep learning, which realizes high-precision and high-adaptability identification of the low-voltage distribution topology structure through the collaborative work of multiple types of intelligent sensors and deep optimization of deep learning algorithms.

[0005] To achieve the above object, the present invention adopts the following technical solutions: A method for topology identification of a low-voltage distribution network based on deep learning comprises the following steps: According to the structural characteristics and node distribution of the low-voltage distribution network, a multi-source sensor array is deployed at the key node position, and the multi-source sensor array is used to collect the first data, and the first data is converted into a digital signal and metadata information is added to obtain the second data; the first data includes electrical quantity data, environmental data and electromagnetic state data; after the integrity check is performed on the second data, data fusion and normalization processing are performed to obtain the third data; Establishing a deep learning model for low-voltage distribution network topology identification, using a third data set to train the deep learning model to obtain a trained deep learning model, and using the trained deep learning model to output a topology structure prediction result of the low-voltage distribution network; the deep learning model uses a CNN-LSTM network model; The prediction results of the topological structure of the low-voltage distribution network are displayed. Each node of the low-voltage distribution network is displayed in a three-dimensional model. The connection lines between the nodes are drawn according to the topological relationship; and the electrical quantity data and environmental data of the lines are marked.

[0006] Furthermore, the key node set is expressed as: ;in Indicates busbar, Indicates branch line, Indicates distribution box, Indicates the access point of the electrical equipment; The node set of low voltage distribution network is expressed as: ;in, is the number of nodes; For each key node , the key node distribution positions are combined as follows: ;in .

[0007] Further, the multi-source sensor array includes a current transformer, a voltage transformer, a temperature sensor, a humidity sensor and an electromagnetic induction sensor; The current transformer converts the sensed secondary current into a digital signal , the conversion formula is: ; in, is the primary current; is the lower limit of current measurement; is the quantization step size; The voltage transformer converts the primary voltage of the low-voltage grid into the secondary voltage: ; in, is the secondary side voltage; is the primary side voltage of the low voltage grid; ;in, is the voltage amplitude, is the angular frequency; is the initial phase; is the change ratio between the primary voltage and the secondary voltage; The temperature sensor senses the ambient temperature as , converted into the corresponding digital signal ; The humidity sensor senses the surrounding humidity as , converted into the corresponding digital signal ; The electromagnetic induction sensor converts the changes in the surrounding electromagnetic field into electrical signals , converted into the corresponding digital signal .

[0008] Furthermore, after integrity checking the second data, a process of performing data fusion and normalization processing to obtain the third data includes: Ensure that the data collected by any sensor is within the preset range, and use linear interpolation to repair missing data; The data collected by different types of sensors are processed using a fusion algorithm based on weighted average and data complementarity; Normalize the fused data.

[0009] Furthermore, the process of using the fusion algorithm based on weighted average and data complementation includes: according to the formula Calculate; where is a weighting factor determined according to the reliability and importance of the data, and , It is the electrical quantity data; for environmental data; To fuse data.

[0010] Furthermore, the specific process of using the third data set to train the deep learning model to obtain a trained deep learning model, and using the trained deep learning model to output a topology structure prediction result of the low-voltage power distribution network includes: Dividing the third data set into a training set, a validation set, and a test set according to a preset ratio; using the training set to train the CNN-LSTM network model, and using a stochastic gradient descent or an adaptive learning rate algorithm to adjust the weight and bias parameters of the network; and in each iteration, calculating the loss function, updating the network parameters through a back propagation algorithm according to the gradient information of the loss function, and gradually converging the CNN-LSTM network model by continuously adjusting the network parameters; After multiple iterations of training, when the accuracy of the CNN-LSTM network model on the validation set reaches a preset threshold, the training is stopped to obtain the trained CNN-LSTM network model; The acquired real-time third data is input into the trained CNN-LSTM network model, and the topology structure prediction result of the low-voltage distribution network is output. Then the accuracy of the topology structure prediction result of the low-voltage distribution network is tested using a test set; the topology structure prediction result of the low-voltage distribution network includes node connection relationships, load distribution on the line, and possible fault nodes.

[0011] Furthermore, each node of the low-voltage power distribution network is displayed in a three-dimensional model, and the process of drawing connection lines between nodes according to the topological relationship includes: If the coordinates of the low-voltage distribution network nodes in real space are , the display coordinates in the 3D visualization interface are ; Use linear transformation method to display nodes in three-dimensional model and draw connection lines between nodes according to topological relationship; The linear transformation formula is: ; ; ; in, is a first constant determined according to the layout and scaling of the visualization interface, is a second constant determined according to the layout and scaling of the visualization interface; is a third constant determined according to the layout and scaling of the visualization interface; is a fourth constant determined according to the layout and scaling of the visualization interface; is a fifth constant determined according to the layout and scaling of the visualization interface; is the sixth constant determined according to the visualization interface layout and zoom ratio.

[0012] Furthermore, the process of marking the electrical quantity data of the line is as follows: Use different colors to mark the electrical quantity data of the line; suppose the current is , the voltage level is , define the color mapping function of the current: ; in, is a value of the first color that is less than the first current threshold; is the value of the second color corresponding to the first current threshold and the second current threshold; For greater than The current threshold corresponds to The value of the color; is the first current threshold; is the second current threshold; For the Current threshold; The color mapping function for voltage levels is: ; in, is a value of the first color that is less than the first voltage threshold; is the value of the second color corresponding to the first voltage threshold and the second voltage threshold; For greater than The voltage threshold corresponds to The value of the color; is the first voltage threshold; is the second voltage threshold; For the Voltage threshold.

[0013] Furthermore, the process of identifying the environmental data includes: In the temperature range, the color mapping function that defines the temperature is: ; in, is the value of the first color corresponding to the first temperature threshold value; is the value of the second color corresponding to the first temperature threshold and the second temperature threshold; For greater than The temperature threshold corresponds to The value of the color; is the first temperature threshold; is the second temperature threshold; For the Temperature threshold.

[0014] The present invention also proposes a topology identification system for a low-voltage distribution network based on deep learning, including a data processing module, a prediction module and a topology identification module: The data processing module is used to deploy a multi-source sensor array at a key node position according to the structural characteristics and node distribution of the low-voltage distribution network, collect first data using the multi-source sensor array, and convert the first data into a digital signal and then add metadata information to obtain second data; the first data includes electrical quantity data, environmental data and electromagnetic state data; after performing integrity check on the second data, data fusion and normalization processing are performed to obtain third data; The prediction module is used to establish a deep learning model for low-voltage distribution network topology identification, use a third data set to train the deep learning model to obtain a trained deep learning model, and use the trained deep learning model to output a topology structure prediction result of the low-voltage distribution network; the deep learning model uses a CNN-LSTM network model; The topology identification module is used to display the topology structure prediction results of the low-voltage distribution network, display each node of the low-voltage distribution network in a three-dimensional model, draw the connection lines between the nodes according to the topological relationship; and identify the electrical quantity data and environmental data of the line.

[0015] The effects provided in the content of the invention are only the effects of the embodiments, not all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects: The present invention proposes a method and system for topology identification of a low-voltage distribution network based on deep learning, which belongs to the technical field of power grid topology identification. The method comprises the following steps: deploying a multi-source sensor array at a key node position according to the structural characteristics and node distribution of the low-voltage distribution network, collecting first data by using the multi-source sensor array, and adding metadata information after converting the first data into a digital signal to obtain second data; the first data comprises electrical quantity data, environmental data and electromagnetic state data; after integrity verification of the second data, data fusion and normalization are performed to obtain third data; establishing a deep learning model for topology identification of the low-voltage distribution network, training the deep learning model with a third data set to obtain a trained deep learning model, and outputting a topology structure prediction result of the low-voltage distribution network with the trained deep learning model; the deep learning model adopts a CNN-LSTM network model; displaying the topology structure prediction result of the low-voltage distribution network, displaying each node of the low-voltage distribution network in a three-dimensional model, and drawing connection lines between nodes according to topological relationships; and marking the electrical quantity data and environmental data of the line. Based on a topology identification method of a low-voltage distribution network based on deep learning, a topology identification system of a low-voltage distribution network based on deep learning is also proposed. The present invention realizes high-precision and high-adaptability identification of low-voltage distribution topology structure through the collaborative work of multiple types of intelligent sensors and deep optimization of deep learning algorithms, which can effectively improve the accuracy and adaptability of low-voltage distribution topology identification and provide strong support for the intelligent management and operation and maintenance of low-voltage distribution networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of a method for topology identification of a low-voltage distribution network based on deep learning proposed in Embodiment 1 of the present invention; Figure 2 This is an architecture diagram for implementing a method for topology identification of a low-voltage distribution network based on deep learning proposed in Example 1 of the present invention; Figure 3 This is a schematic diagram of the CNN-LSTM network model proposed in Example 1 of the present invention; Figure 4 This is a schematic diagram of a topology identification system for a low-voltage distribution network based on deep learning proposed in Example 2 of the present invention. DETAILED DESCRIPTION

[0017] In order to clearly illustrate the technical features of the present solution, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings. The disclosure below provides many different embodiments or examples for realizing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and settings of specific examples are described below. In addition, the present invention may repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. It should be noted that the components illustrated in the accompanying drawings are not necessarily drawn to scale. The present invention omits the description of known components and processing techniques and processes to avoid unnecessary limitations on the present invention.

[0018] Example 1 Embodiment 1 of the present invention proposes a low-voltage distribution network topology identification method based on deep learning, which is used to solve the technical problems existing in the low-voltage distribution network topology identification in the prior art.

[0019] Figure 1 A flow chart of a method for topology identification of a low-voltage distribution network based on deep learning proposed in Embodiment 1 of the present invention; In order to accurately describe in detail a method for topology identification of a low-voltage distribution network based on deep learning proposed in Example 1 of the present invention, Figure 2 This is an architecture diagram for implementing a method for topology identification of a low-voltage distribution network based on deep learning proposed in Example 1 of the present invention; In step S100, a multi-source sensor array is deployed at key node locations according to the structural characteristics and node distribution of the low-voltage distribution network.

[0020] In this application, key node locations include busbars, branch lines, distribution boxes, and access points for important electrical equipment.

[0021] The key node set is expressed as: ;in Indicates busbar, Indicates branch line, Indicates distribution box, Indicates the access point of the electrical equipment; The node set of low voltage distribution network is expressed as: ;in, is the number of nodes; For each key node , the key node distribution positions are combined as follows: ;in , satisfying certain distribution rules. For example, at the busbar position , which may be selected based on the power flow direction and load balancing principle of the network, such as selecting the front node with the largest inflow power. nodes as busbar position nodes.

[0022] The multi-source sensor array includes: current transformer, voltage transformer, temperature sensor, humidity sensor and electromagnetic induction sensor. Current transformer and voltage transformer are used to obtain electrical quantity data; temperature sensor and humidity sensor are used to obtain environmental data; electromagnetic induction sensor is used to obtain electromagnetic state data.

[0023] With current transformer, for line rated current The measurement range of the line , , (Here it is assumed that under certain conditions the sampling frequency is fixed to , if there are changes according to the line parameters, a functional relationship can be further established).

[0024] For the voltage transformer, the grid voltage level is ,exist and measuring range The relationship can be expressed as: ; in, is the change ratio between the primary voltage and the secondary voltage; is the primary side voltage, is the secondary voltage, and According to voltage level The standard range is determined by

[0025] For environmental sensors, such as temperature sensors, the measurement range , humidity sensor measurement range , which can be simply expressed as (Unit: ); (unit: ).

[0026] In step S200, first data is collected using a multi-source sensor array, and the first data is converted into a digital signal and metadata information is added to obtain second data; the first data includes electrical quantity data, environmental data and electromagnetic state data; after the integrity of the second data is checked, data fusion and normalization are performed to obtain third data.

[0027] Each sensor has independent data collection and preliminary processing capabilities, and can convert the collected raw data into digital signals and add metadata information such as timestamp, location identifier, and sensor type identifier.

[0028] The current transformer follows the set sampling frequency Real-time monitoring of line current. When current flows through the line, according to the principle of electromagnetic induction, the secondary winding of the current transformer will induce a secondary current proportional to the primary current. Assuming the primary current is , and its relationship with time can be expressed as: , here it is assumed to be a sinusoidal alternating current, is the current amplitude, is the angular frequency, The initial phase.

[0029] The current transformer converts the sensed secondary current into a digital signal , the conversion formula is: ; in, is the primary current; is the lower limit of current measurement; is the quantization step size; after the conversion is completed, the system adds a timestamp to the data The acquisition method can be achieved through the system's built-in clock module, which accurately records the time of data collection and the location mark Mark according to the node location information of the current transformer installation, such as the starting end of branch line 1, etc.; and the sensor type identification , indicating that the data comes from the current transformer.

[0030] The voltage transformer is connected in parallel at both ends of the measured line to sample the line voltage. Assume that the grid voltage is , its relationship with time may also follow the sine law ;in, is the voltage amplitude, is the angular frequency; is the initial phase; The voltage transformer converts the primary voltage of the low-voltage grid into the secondary voltage: ; in, is the secondary side voltage; is the primary side voltage of the low voltage grid; It is the change ratio between the primary side voltage and the secondary side voltage; the secondary side voltage signal is converted into a digital signal by ADC, and the conversion process is similar to that of the current transformer.

[0031] The temperature sensor senses the ambient temperature as , converted into the corresponding digital signal ; The humidity sensor senses the surrounding humidity as , converted into the corresponding digital signal ; The electromagnetic induction sensor converts the changes in the surrounding electromagnetic field into electrical signals , converted into the corresponding digital signal .

[0032] After performing integrity check on the second data, a process of performing data fusion and normalization processing to obtain the third data includes: Ensure that the data collected by any sensor is within the preset range, and use linear interpolation to repair the missing data; use a fusion algorithm based on weighted average and data complementarity to process the data collected by different types of sensors; and normalize the fused data.

[0033] The process of using the fusion algorithm based on weighted average and data complementation includes: according to the formula Calculate; where is a weighting factor determined according to the reliability and importance of the data, and , It is the electrical quantity data; for environmental data; To fuse data, establish a data reliability evaluation system and determine the weighting coefficient based on factors such as sensor accuracy, stability, and historical data accuracy. The specific value of is used to ensure that the fused data can effectively reflect the actual situation.

[0034] Normalize the fused data, such as using the minimum-maximum normalization method, and map the data to interval to ensure that the data meets the input requirements of the deep learning algorithm.

[0035] In step S300, a deep learning model for low-voltage distribution network topology identification is established, the deep learning model is trained using a third data set to obtain a trained deep learning model, and the trained deep learning model is used to output a topology structure prediction result of the low-voltage distribution network; the deep learning model adopts a CNN-LSTM network model; Figure 3 This is a schematic diagram of the CNN-LSTM network model proposed in Example 1 of the present invention; The CNN layer is used to extract the spatial features of the data, specifically to mine potential topological connection clues from the distribution of electrical and environmental quantity data of multiple nodes.

[0036] For the input fusion data matrix X, the convolution operation is performed according to the formula to obtain the feature map C; ; in, are the coordinate indices of the feature maps. is the element index of the convolution kernel. During the training process, let the convolution kernel The weight parameter is , paranoid , then the output after convolution is: ; LSTM layer function: The LSTM layer learns the time series characteristics of data through a unique gating mechanism (including input gate, forget gate and output gate). The input gate determines the input degree of new information, the forget gate controls the retention or discarding of previous information, and the output gate determines the information to be output in the current state. This mechanism enables it to effectively capture the dynamic characteristics of load and environment changes over time in low-voltage distribution networks, such as the fluctuation pattern of load in different time periods, the long-term impact of environmental factors (such as temperature and humidity changes) on electrical parameters, etc.

[0037] According to the characteristics of time series data of low-voltage power distribution network, the hyperparameters of LSTM layer are set. For example, according to the sampling frequency of data and the expected time dependency, the appropriate number of hidden units is determined to balance the learning ability and computational complexity of the model. At the same time, the appropriate activation function (such as tanh or sigmoid function) is selected to enhance the nonlinear expression ability of the model, so as to better adapt to complex dynamic changes.

[0038] In this application, a large number of data samples from low-voltage distribution networks of different types and sizes are collected, and these samples are preprocessed and labeled to ensure the accuracy and completeness of the data, and finally the third data set is obtained.

[0039] The third data set is divided into a training set, a validation set, and a test set according to a preset ratio, for example, 70% for the training set, 20% for the validation set, and 10% for the test set; the CNN-LSTM network model is trained using the training set, and the weight and bias parameters of the network are adjusted using a random gradient descent or adaptive learning rate algorithm; the adaptive learning rate algorithm can use an adaptive learning rate algorithm such as Adagrad or Adadelta. And in each iteration, the loss function such as (cross entropy loss function) is calculated, and the network parameters are updated through the back propagation algorithm according to the gradient information of the loss function. By continuously adjusting the network parameters, the CNN-LSTM network model gradually converges; let the predicted topological structure be The actual topology is , then the loss function is: ; in, It is the node or connection index of the topological structure. During the training process, the stochastic gradient descent algorithm is used to adjust the weight and parameter deviation of the network. , the formula is: ( is the learning rate); For each iteration, the loss function is calculated with respect to the weights Gradient , for example, for the weights of the fully connected layer , its gradient calculation can be expanded according to the chain rule: ; in, is the output layer node value; is the value before the activation function.

[0040] After multiple iterations of training, when the accuracy of the CNN-LSTM network model on the validation set reaches a preset threshold (such as 95%), the training is stopped to obtain the trained CNN-LSTM network model; The acquired real-time third data is input into the trained CNN-LSTM network model, and the topological structure prediction result of the low-voltage distribution network is output. Then, the accuracy of the topological structure prediction result of the low-voltage distribution network is tested using the test set; the topological structure prediction result of the low-voltage distribution network includes the node connection relationship, the load distribution on the line, and the possible fault nodes. At the same time, in order to further improve the generalization ability and adaptability of the model, the model is incrementally learned and updated regularly using newly collected data.

[0041] In step S400, the topological structure prediction result of the low-voltage distribution network is displayed, each node of the low-voltage distribution network is displayed in a three-dimensional model, and the connection lines between the nodes are drawn according to the topological relationship; and the electrical quantity data and environmental data of the lines are marked.

[0042] In the three-dimensional visualization interface, each node of the low-voltage distribution network is displayed as a three-dimensional model, and the connection lines between the nodes are drawn based on the topological relationship.

[0043] If the coordinates of the low-voltage distribution network nodes in real space are , the display coordinates in the 3D visualization interface are ; Use linear transformation method to display nodes in three-dimensional model and draw connection lines between nodes according to topological relationship; The linear transformation formula is: ; ; ; in, is a first constant determined according to the layout and scaling of the visualization interface, is a second constant determined according to the layout and scaling of the visualization interface; is a third constant determined according to the layout and scaling of the visualization interface; is a fourth constant determined according to the layout and scaling of the visualization interface; is a fifth constant determined according to the layout and scaling of the visualization interface; is the sixth constant determined according to the visualization interface layout and zoom ratio.

[0044] The process of marking the electrical quantity data of the line is as follows: using different colors to mark the electrical quantity data of the line; assuming the current size is , the voltage level is , define the color mapping function of the current: ; in, is a value of the first color that is less than the first current threshold; is the value of the second color corresponding to the first current threshold and the second current threshold; For greater than The current threshold corresponds to The value of the color; is the first current threshold; is the second current threshold; For the Current threshold; The color mapping function for voltage levels is: ; in, is a value of the first color that is less than the first voltage threshold; is the value of the second color corresponding to the first voltage threshold and the second voltage threshold; For greater than The voltage threshold corresponds to The value of the color; is the first voltage threshold; is the second voltage threshold; For the Voltage threshold.

[0045] The process of marking environmental data includes: within the temperature range, the color mapping function of the temperature is defined as: ; in, is the value of the first color corresponding to the first temperature threshold value; is the value of the second color corresponding to the first temperature threshold and the second temperature threshold; For greater than The temperature threshold corresponds to The value of the color; is the first temperature threshold; is the second temperature threshold; For the Temperature threshold.

[0046] The process of identifying the humidity environment is similar to that of temperature identification.

[0047] This application also provides users with interactive operation functions, including zooming and rotating views, so that users can view the topology from different angles. Users can click on nodes to view detailed information, such as electrical parameters, environmental parameters, etc. Users are allowed to manually correct and annotate the topology to deal with possible errors or special situations.

[0048] This application can record topological structure information at different time points and provide historical data query function, which is convenient for users to analyze and study the topological evolution of low-voltage distribution networks.

[0049] Embodiment 1 of the present invention proposes a topology identification method for a low-voltage distribution network based on deep learning. Through the collaborative work of multiple types of intelligent sensors and the deep optimization of deep learning algorithms, high-precision and high-adaptability identification of low-voltage distribution topology structures is achieved, which can effectively improve the accuracy and adaptability of low-voltage distribution topology identification and provide strong support for the intelligent management and operation and maintenance of low-voltage distribution networks.

[0050] Example 2 Based on the method for topology identification of a low-voltage distribution network based on deep learning in embodiment 1 of the present invention, embodiment 2 of the present invention further proposes a topology identification system for a low-voltage distribution network based on deep learning. Figure 4 This is a schematic diagram of a topology identification system for a low-voltage distribution network based on deep learning proposed in Example 2 of the present invention. The system includes: a data processing module, a prediction module and a topology identification module: The data processing module is used to deploy a multi-source sensor array at a key node position according to the structural characteristics and node distribution of the low-voltage distribution network, collect first data using the multi-source sensor array, and convert the first data into a digital signal and then add metadata information to obtain second data; the first data includes electrical quantity data, environmental data and electromagnetic state data; after performing integrity check on the second data, data fusion and normalization processing are performed to obtain third data; The prediction module is used to establish a deep learning model for low-voltage distribution network topology identification, use the third data set to train the deep learning model to obtain a trained deep learning model, and use the trained deep learning model to output a topology structure prediction result of the low-voltage distribution network; the deep learning model uses a CNN-LSTM network model; The topology identification module is used to display the topological structure prediction results of the low-voltage distribution network, display each node of the low-voltage distribution network in a three-dimensional model, draw the connection lines between the nodes based on the topological relationship; and identify the electrical quantity data and environmental data of the line.

[0051] In the data processing module, the key node set is represented as: ;in Indicates busbar, Indicates branch line, Indicates distribution box, Indicates the access point of the electrical equipment; The node set of low voltage distribution network is expressed as: ;in, is the number of nodes; For each key node , the key node distribution positions are combined as follows: ;in .

[0052] The multi-source sensor array includes current transformers, voltage transformers, temperature sensors, humidity sensors, and electromagnetic induction sensors; The current transformer converts the sensed secondary current into a digital signal , the conversion formula is: ; in, is the primary current; is the lower limit of current measurement; is the quantization step size; The voltage transformer converts the primary voltage of the low-voltage grid into the secondary voltage: ; in, is the secondary side voltage; is the primary side voltage of the low voltage grid; ;in, is the voltage amplitude, is the angular frequency; is the initial phase; is the change ratio between the primary voltage and the secondary voltage; The temperature sensor senses the ambient temperature as , converted into the corresponding digital signal ; The humidity sensor senses the surrounding humidity as , converted into the corresponding digital signal ; The electromagnetic induction sensor converts the changes in the surrounding electromagnetic field into electrical signals , converted into the corresponding digital signal .

[0053] After the integrity check of the second data, the process of performing data fusion and normalization to obtain the third data includes: ensuring that the data collected by any sensor is within a preset range, and repairing the missing data by linear interpolation; processing the data collected by different types of sensors by a fusion algorithm based on weighted average and data complementarity; and normalizing the fused data.

[0054] The process of using the fusion algorithm based on weighted average and data complementation includes: according to the formula Calculate; where is a weighting factor determined according to the reliability and importance of the data, and , It is the electrical quantity data; for environmental data; To fuse data.

[0055] In the prediction module, the deep learning model is trained using the third data set to obtain a trained deep learning model, and the specific process of using the trained deep learning model to output the topology structure prediction result of the low-voltage power distribution network includes: Dividing the third data set into a training set, a validation set, and a test set according to a preset ratio; using the training set to train the CNN-LSTM network model, and using a stochastic gradient descent or an adaptive learning rate algorithm to adjust the weight and bias parameters of the network; and in each iteration, calculating the loss function, updating the network parameters through a back propagation algorithm according to the gradient information of the loss function, and gradually converging the CNN-LSTM network model by continuously adjusting the network parameters; After multiple iterations of training, when the accuracy of the CNN-LSTM network model on the validation set reaches a preset threshold, the training is stopped to obtain the trained CNN-LSTM network model; The acquired real-time third data is input into the trained CNN-LSTM network model, and the topology structure prediction result of the low-voltage distribution network is output. Then the accuracy of the topology structure prediction result of the low-voltage distribution network is tested using a test set; the topology structure prediction result of the low-voltage distribution network includes node connection relationships, load distribution on the line, and possible fault nodes.

[0056] In the topology identification module, each node of the low-voltage distribution network is displayed in a three-dimensional model. The process of drawing the connection lines between nodes based on the topological relationship includes: If the coordinates of the low-voltage distribution network nodes in real space are , the display coordinates in the 3D visualization interface are ; Use linear transformation method to display nodes in three-dimensional model and draw connection lines between nodes according to topological relationship; The linear transformation formula is: ; ; ; in, is a first constant determined according to the layout and scaling of the visualization interface, is a second constant determined according to the layout and scaling of the visualization interface; is a third constant determined according to the layout and scaling of the visualization interface; is a fourth constant determined according to the layout and scaling of the visualization interface; is a fifth constant determined according to the layout and scaling of the visualization interface; is the sixth constant determined according to the visualization interface layout and zoom ratio.

[0057] The process of marking the electrical quantity data of the line is as follows: using different colors to mark the electrical quantity data of the line; assuming the current size is , the voltage level is , define the color mapping function of the current: ; in, is a value of the first color that is less than the first current threshold; is the value of the second color corresponding to the first current threshold and the second current threshold; For greater than The current threshold corresponds to The value of the color; is the first current threshold; is the second current threshold; For the Current threshold; The color mapping function for voltage levels is: ; in, is a value of the first color that is less than the first voltage threshold; is the value of the second color corresponding to the first voltage threshold and the second voltage threshold; For greater than The voltage threshold corresponds to The value of the color; is the first voltage threshold; is the second voltage threshold; For the Voltage threshold.

[0058] The process of marking environmental data includes: within the temperature range, the color mapping function of the temperature is defined as: ; in, is the value of the first color corresponding to the first temperature threshold value; is the value of the second color corresponding to the first temperature threshold and the second temperature threshold; For greater than The temperature threshold corresponds to The value of the color; is the first temperature threshold; is the second temperature threshold; For the Temperature threshold.

[0059] Embodiment 2 of the present invention proposes a topology identification system for a low-voltage distribution network based on deep learning. Through the collaborative work of multiple types of intelligent sensors and the deep optimization of deep learning algorithms, high-precision and high-adaptability identification of low-voltage distribution topology structures is achieved, which can effectively improve the accuracy and adaptability of low-voltage distribution topology identification, and provide strong support for the intelligent management and operation and maintenance of low-voltage distribution networks.

[0060] The description of the relevant parts of a low-voltage distribution network topology identification system based on deep learning provided in Example 2 of the present application can be found in the detailed description of the corresponding parts of a low-voltage distribution network topology identification method based on deep learning provided in Example 1 of the present application, and will not be repeated here.

[0061] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment that includes a series of elements are inherent to the elements. In the absence of more restrictions, the elements limited by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or equipment that includes the elements. In addition, the above-mentioned technical solution provided in the embodiment of the present application is consistent with the corresponding technical solution in the prior art in principle, and the part is not described in detail, so as not to repeat too much.

[0062] Although the above describes the specific implementation of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. For those skilled in the art, other different forms of modifications or deformations can be made on the basis of the above description. It is not necessary and impossible to list all the implementation methods here. On the basis of the technical solution of the present invention, various modifications or deformations that can be made by those skilled in the art without creative work are still within the scope of protection of the present invention.

Claims

1. A topology identification method for low-voltage distribution network based on deep learning, characterized in that: The following steps are involved: According to the structural characteristics and node distribution of the low-voltage distribution network, a multi-source sensor array is deployed at the key node position, and the multi-source sensor array is used to collect the first data, and the first data is converted into a digital signal and metadata information is added to obtain the second data; the first data includes electrical quantity data, environmental data and electromagnetic state data; after the integrity check is performed on the second data, data fusion and normalization processing are performed to obtain the third data; Establishing a deep learning model for low-voltage distribution network topology identification, using a third data set to train the deep learning model to obtain a trained deep learning model, and using the trained deep learning model to output a topology structure prediction result of the low-voltage distribution network; The deep learning model adopts a CNN-LSTM network model; The prediction results of the topological structure of the low-voltage distribution network are displayed. Each node of the low-voltage distribution network is displayed in a three-dimensional model. The connection lines between the nodes are drawn according to the topological relationship; and the electrical quantity data and environmental data of the lines are marked.

2. According to a method for topology identification of a low-voltage distribution network based on deep learning according to claim 1, it is characterized in that: The key node set is expressed as: ;in Indicates busbar, Indicates branch line, Indicates distribution box, Indicates the access point of the power-consuming equipment; The node set of low voltage distribution network is expressed as: ;in, is the number of nodes; For each key node , the key node distribution positions are combined as follows: ;in .

3. According to a method for topology identification of a low-voltage distribution network based on deep learning according to claim 2, it is characterized in that: The multi-source sensor array includes a current transformer, a voltage transformer, a temperature sensor, a humidity sensor and an electromagnetic induction sensor; The current transformer converts the sensed secondary current into a digital signal , the conversion formula is: ； in, is the primary current; is the lower limit of current measurement; is the quantization step size; The voltage transformer converts the primary voltage of the low-voltage grid into the secondary voltage: ； in, is the secondary side voltage; is the primary side voltage of the low voltage grid; ;in, is the voltage amplitude, is the angular frequency; is the initial phase; is the change ratio between the primary voltage and the secondary voltage; The temperature sensor senses the ambient temperature as , converted into the corresponding digital signal ; The humidity sensor senses the surrounding humidity as , converted into the corresponding digital signal ; The electromagnetic induction sensor converts the changes in the surrounding electromagnetic field into electrical signals , converted into the corresponding digital signal .

4. According to a method for topology identification of a low-voltage distribution network based on deep learning according to claim 1, it is characterized in that: After performing integrity check on the second data, a process of performing data fusion and normalization processing to obtain the third data includes: Ensure that the data collected by any sensor is within the preset range, and use linear interpolation to repair missing data; The data collected by different types of sensors are processed using a fusion algorithm based on weighted average and data complementarity; Normalize the fused data.

5. According to a method for topology identification of a low-voltage distribution network based on deep learning according to claim 4, it is characterized in that: The process of using the fusion algorithm based on weighted average and data complementation includes: according to the formula Calculate; where is a weighting factor determined according to the reliability and importance of the data, and , It is the electrical quantity data; for environmental data; To fuse data.

6. According to a method for topology identification of a low-voltage distribution network based on deep learning according to claim 1, it is characterized in that: The specific process of using the third data set to train the deep learning model to obtain a trained deep learning model, and using the trained deep learning model to output a topology structure prediction result of the low-voltage power distribution network includes: Dividing the third data set into a training set, a validation set, and a test set according to a preset ratio; using the training set to train the CNN-LSTM network model, and using a stochastic gradient descent or an adaptive learning rate algorithm to adjust the weight and bias parameters of the network; and in each iteration, calculating the loss function, updating the network parameters through a back propagation algorithm according to the gradient information of the loss function, and gradually converging the CNN-LSTM network model by continuously adjusting the network parameters; After multiple iterations of training, when the accuracy of the CNN-LSTM network model on the validation set reaches a preset threshold, the training is stopped to obtain the trained CNN-LSTM network model; The acquired real-time third data is input into the trained CNN-LSTM network model, and the topology structure prediction result of the low-voltage distribution network is output. Then the accuracy of the topology structure prediction result of the low-voltage distribution network is tested using a test set; the topology structure prediction result of the low-voltage distribution network includes node connection relationships, load distribution on the line, and possible fault nodes.

7. The method for topology identification of a low-voltage distribution network based on deep learning according to claim 1, characterized in that: The process of displaying each node of the low-voltage power distribution network in a three-dimensional model and drawing the connection lines between the nodes based on the topological relationship includes: If the coordinates of the low-voltage distribution network nodes in real space are , the display coordinates in the 3D visualization interface are ; Use linear transformation method to display nodes in three-dimensional model and draw connection lines between nodes according to topological relationship; The linear transformation formula is: ; ; ; in, is a first constant determined according to the layout and scaling of the visualization interface, is a second constant determined according to the layout and scaling of the visualization interface; is a third constant determined according to the layout and scaling of the visualization interface; is a fourth constant determined according to the layout and scaling of the visualization interface; is a fifth constant determined according to the layout and scaling of the visualization interface; is the sixth constant determined according to the visualization interface layout and zoom ratio.

8. The method for topology identification of a low-voltage distribution network based on deep learning according to claim 1, characterized in that: The process of marking the electrical quantity data of the line is as follows: Use different colors to mark the electrical quantity data of the line; suppose the current is , the voltage level is , define the color mapping function of the current: ； in, is a value of the first color that is less than the first current threshold; is the value of the second color corresponding to the first current threshold and the second current threshold; For greater than The current threshold corresponds to The value of the color; is the first current threshold; is the second current threshold; For the Current threshold; The color mapping function for voltage levels is: ； in, is a value of the first color that is less than the first voltage threshold; is the value of the second color corresponding to the first voltage threshold and the second voltage threshold; For greater than The voltage threshold corresponds to The value of the color; is the first voltage threshold; is the second voltage threshold; For the Voltage threshold.

9. The method for topology identification of a low-voltage distribution network based on deep learning according to claim 1, characterized in that: The process of labeling environmental data includes: In the temperature range, the color mapping function that defines the temperature is: ； in, is the value of the first color corresponding to the first temperature threshold value; is the value of the second color corresponding to the first temperature threshold and the second temperature threshold; For greater than The temperature threshold corresponds to The value of the color; is the first temperature threshold; is the second temperature threshold; For the Temperature threshold.

10. A topology identification system for low-voltage distribution network based on deep learning, characterized in that: Including data processing module, prediction module and topology recognition module: The data processing module is used to deploy a multi-source sensor array at a key node position according to the structural characteristics and node distribution of the low-voltage distribution network, collect first data using the multi-source sensor array, and convert the first data into a digital signal and then add metadata information to obtain second data; the first data includes electrical quantity data, environmental data and electromagnetic state data; after performing integrity check on the second data, data fusion and normalization processing are performed to obtain third data; The prediction module is used to establish a deep learning model for low-voltage distribution network topology identification, use a third data set to train the deep learning model to obtain a trained deep learning model, and use the trained deep learning model to output a topology structure prediction result of the low-voltage distribution network; The deep learning model adopts a CNN-LSTM network model; The topology identification module is used to display the topology structure prediction results of the low-voltage distribution network, display each node of the low-voltage distribution network in a three-dimensional model, draw the connection lines between the nodes according to the topological relationship; and identify the electrical quantity data and environmental data of the line.

Images