Power transmission and transformation equipment fault risk assessment method and system in extremely severe weather

Abstract

The invention discloses a power transmission and transformation equipment fault risk assessment method and system in extremely severe weather, and the method and system improve the fault detection effect through the cloud-side intelligent cooperation technology, a self-learning federated learning mechanism and the construction of a risk map. Power transmission line fault detection based on cloud edge cooperation is composed of an edge end and a cloud end. The edge end comprises a hidden danger identification module, a data acquisition module and an algorithm iteration module. And the cloud is responsible for integrating and globally updating the model and the map. A self-learning federated learning method is used, the edge ends use own data to train a model, and model parameter data uploaded to the cloud end by each edge end are updated globally at the cloud end and deployed to each edge end. The phenomenon that the recognition capability is low due to unbalanced data collection at the edge end, missing of sample categories and overfitting is avoided, and the problem that the transmission pressure of a network channel is too large is solved.

Description

Technical Field

[0001] The present invention relates to a risk assessment method and system, and in particular to a risk assessment method and system for power transmission and transformation equipment failure under extremely severe weather conditions. Background Art

[0002] With the continuous advancement of science and technology, residents' daily electricity consumption has increased year by year. The security of power transactions plays an important role in smart grids, which has greatly promoted the country's economic development and social stability and prosperity. Climate change has become an important factor affecting my country's energy security and sustainable economic and social development. Under the combined effect of these factors, the balance of power supply and demand faces severe challenges, especially in new power systems. Due to the particularity of their structure and operation mode, the risk of this imbalance between supply and demand is further amplified, and the stable operation of the power system is seriously threatened. Effective measures are urgently needed to deal with it.

[0003] Common factors affecting power transmission and transformation equipment in extreme weather conditions include bypass attacks, counterattacks, wind damage, ice damage, wildfires, and external damage. In order to reduce the number of failures of power transmission and transformation equipment, staff often need to detect power transmission and transformation equipment failures. Traditional power transmission and transformation equipment fault detection is through manual line inspections. Due to the influence of factors such as extreme weather and terrain, manual inspections are inefficient and have high manpower and material costs. With the development of drone technology, the phenomenon of using drones to replace manual channel inspections has become increasingly common. Drones are relatively cost-effective, but drone inspections have limited distances and insufficient inspection accuracy. At present, relevant departments of the power grid use tower visualization monitoring equipment and various sensors to send equipment collected data to cloud servers for fault detection and hidden danger identification in the cloud. However, this leads to excessive pressure on cloud computing and channel networks, and the real-time identification of hidden dangers is low, and abnormal information across provinces and regions cannot be communicated in a timely manner. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide an evaluation method and system for accurate detection, timely maintenance and global early warning of abnormal faults in power transmission and transformation equipment under extremely severe weather conditions.

[0005] Technical solution: The present invention provides a method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions, the method comprising the following steps:

[0006] (1) Collect abnormal data of power transmission and transformation equipment;

[0007] (2) Perform feature engineering on abnormal data to achieve feature extraction, perform data annotation on the abnormal data after feature extraction, and use data standardization to map the feature values ​​to the [0, 1] interval;

[0008] (3) Constructing the MLP model;

[0009] (4) MLP model update: A global model is initialized using MLP random parameters and deployed to the edge. After local training is completed on the edge, the updated model parameters are uploaded to the cloud for aggregation. The aggregated model parameters are updated using the gradient descent algorithm.

[0010] (5) Map building: A local map is built at the edge. Each edge sends its local map to the cloud for aggregation. A global map is created in the cloud and then broadcast to each edge. Each edge obtains the latest abnormal conditions of the transmission lines in a larger area.

[0011] The step (2) is to perform feature engineering on the abnormal data, introduce a tree model, select feature information including number, name, and range, annotate the abnormal data after feature extraction, fill non-existent data with "0", and use data standardization: Map the eigenvalues ​​to the interval [0,1];

[0012] The MLP model in step (3) is a three-layer network structure, including an input layer, a hidden layer and an output layer, wherein the input layer is the data input into the MLP model, the hidden layer maps the feature information of the input data, and the output layer is the output data. The output layer uses the Softmax normalization function to convert the data value into a probability. The Softmax function is defined as: Among them, exp(z i ) represents an exponential function with the natural constant e as its base.

[0013] The ReLU activation function is used in the MLP model to convert the input of the neuron into a nonlinear output. The ReLU function expression is: f(x)=max(0,x), where x is the input value and f(x) is the output of the activation function.

[0014] The MLP model uses learning rate, weight decay, batch size and number of iterations for optimization and adjustment, divides the data into training set, validation set and test set, uses gradient descent algorithm to optimize parameters, and calculates the loss value and accuracy of the model at the end of each round.

[0015] The step (4) specifically uses the random parameters of the MLP model to initialize a model as a global model, deploys it to each edge terminal, and each edge terminal uses its own data to locally train the model. After the training is completed, the updated model parameters are uploaded to the cloud server for weighted average aggregation, and the aggregated model parameters are updated using the gradient descent optimization algorithm. The above steps are repeated until the preset model performance is achieved.

[0016] The step (5) specifically collects edge data to build a local model, divides the data into three levels: low, medium and high according to the degree of data anomaly, builds a local map at the edge, including the anomaly type, anomaly level and anomaly location, and the edge sends the local local map to the cloud for aggregation, creates a global map in the cloud, and then broadcasts it to each edge, so that each edge obtains the latest abnormal conditions of the transmission lines in a larger area.

[0017] The present invention provides a system for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions, comprising:

[0018] Data acquisition module: used to collect abnormal data of power transmission and transformation equipment and store the data;

[0019] Hidden danger identification module: used to extract features from abnormal data and label the data;

[0020] Algorithm iteration module: used to train the model at the edge and upload the updated model to the cloud for aggregation and updating;

[0021] Map integration module: used to build local maps on the edge and create global maps on the cloud;

[0022] A computer device described in the present invention includes one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather are implemented.

[0023] The computer-readable storage medium described in the present invention stores a computer program thereon, and when the computer program is executed by a processor, the steps of the method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions are implemented.

[0024] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: designing a power transmission and transformation equipment fault detection architecture, realizing resource collaboration, data collaboration and intelligent collaboration between the cloud and the edge, thereby alleviating the network channel pressure between the cloud and the edge, and introducing self-learning federated learning technology, making full use of hidden danger data in the actual environment, and continuously improving the accuracy of hidden danger identification through these technologies; through the federated learning technology, the edge end's own data is iteratively trained on the model, the model parameter information is uploaded to the cloud, and the cloud aggregates and updates the model parameter information and sends it to the edge end, which greatly improves the edge end model recognition ability. The use of federated learning has the advantages of data privacy protection, reducing communication overhead, improving model quality, enhancing data security, and being suitable for real-time sharing of abnormal data between provinces and regions. The edge end and the cloud end collaborate on the local and global construction of the abnormal situation map of the power transmission and transformation equipment. Through the local map, a quick warning can be provided locally, and the cloud map can be broadcast to various regions. Then, each region can obtain the latest abnormal situation of the transmission line across provinces and regions, which helps maintenance personnel to arrive for maintenance in time, select the optimal transmission route, and significantly reduce the damage of dangerous transmission lines and cause greater accidents. It has very important practical significance for ensuring the safe operation of power market transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The figure is an overall framework diagram of the method described in the present invention.

[0026] Figure 2 This is a structural diagram of the MLP model described in the present invention. DETAILED DESCRIPTION

[0027] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0028] A method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions comprises the following steps:

[0029] (1) Collect abnormal data of power transmission and transformation equipment;

[0030] (2) Perform feature engineering on abnormal data to achieve feature extraction, perform data annotation on the abnormal data after feature extraction, and use data standardization to map the feature values ​​to the [0, 1] interval;

[0031] (3) Constructing the MLP model;

[0032] (4) MLP model update: A global model is initialized using MLP random parameters and deployed to the edge. After local training is completed on the edge, the updated MLP model parameters are uploaded to the cloud for aggregation. The aggregated MLP model parameters are updated using the gradient descent algorithm.

[0033] (5) Map building: A local map is built at the edge. Each edge sends its local map to the cloud for aggregation. A global map is created in the cloud and then broadcast to each edge. Each edge obtains the latest abnormal conditions of the transmission lines in a larger area.

[0034] like Figure 1 As shown, the following is a detailed description of the various steps of the risk assessment method for power transmission and transformation equipment failure under extreme weather conditions:

[0035] Step 1: Collect abnormal data of power transmission and transformation equipment through sensors on the tower. The sensors include electric field sensors, temperature sensors, and humidity sensors. The abnormal data include shielding failure, backlash, wind damage, ice damage, wildfire, and external damage.

[0036] Step 2: Perform feature engineering on abnormal data, introduce the tree model method, realize feature selection, and select feature information including number, name, and range: 1. Volcanic hazard {0,1}; 2. External rupture hazard {0,1}; 3. Hazard area [0,1]; 4. Temperature [-50,100]; 5. Humidity [0,100]; 6. Wind speed [0,100]; 7. Precipitation [0,10]; 8. Electric field strength [-100,100]; 9. Current amplitude [0,500]. For the abnormal data after feature extraction, data annotation is performed, and non-existent data is filled with "0". During the annotation process, the label and fault type correspond: 0, volcanic fault; 1, external rupture fault; 2, shielding fault; 3, counterattack fault; 4, wind damage fault; 5, ice damage fault. Use 0-1 standardization to preprocess the data and map the feature value to the [0,1] interval. The standardization formula is as follows:

[0037]

[0038] Step 3: If Figure 2 As shown in the figure, a three-layer multi-layer perceptron MLP model is selected, which is a feed-forward neural network with interconnected layers. The MLP contains a three-layer network structure: output layer, hidden layer and output layer. Figure 1 As shown in the figure, each solid circle represents a neuron; the first layer is the input layer, which represents the input neural network data; the second layer is the hidden layer, which maps the feature information of the input data to another dimensional space to obtain its more abstract feature information; the third layer is the output layer, which obtains an output value to be applied to downstream tasks. After being processed by the output layer neurons, the classification task is finally completed through the Softmax layer. The Softmax function is a normalization function used to convert a set of arbitrary real values ​​into a probability distribution. If the input value is large, the corresponding output value (probability) will also be large; conversely, if the input value is small or negative, the corresponding output value (probability) will be small. The formal definition of the Softmax function is as follows:

[0039]

[0040] In the formula, exp(z i ) represents an exponential function with the natural constant e as its base.

[0041] The ReLU activation function is used in the MLP model to convert the neuron input into a nonlinear output. The ReLU function expression is: f(x) = max(0,x), where x is the input value and f(x) is the output of the activation function.

[0042] Four parameters are selected to optimize the model, namely, learning rate (lr), weight decay, batch size and epochs. The range of learning rate is {0.0001, 0.0005, 0.001, 0.005, 0.01, 0.05, 0.1}; the range of weight decay is {0.0001, 0.0003, 0.0005, 0.0007, 0.001}; the range of batch size is {1, 5, 10}; the range of iterations is {50, 100, 200, 300}.

[0043] The data is randomly divided into training set, validation set and test set in a ratio of 8:1:1. The gradient descent algorithm is used to optimize the parameters, and the loss value and accuracy of the model are calculated at the end of each round.

[0044] Step 4: Initialize a model with random parameters of the MLP network as the global model and deploy it to each edge. Each edge will use its own data to train the model without directly sharing data. After each edge completes local training, the updated model parameters are uploaded to the cloud server for weighted average aggregation. Update the aggregated model parameters using the gradient descent optimization algorithm and repeat the above steps until the preset model performance is achieved. Through such an iterative process, federated learning can train models in a distributed environment to achieve the goals of data privacy protection and resource sharing. Each edge can be trained locally without uploading data to the cloud server, thereby protecting the privacy of the data, reducing network transmission pressure, improving model quality, and enhancing data security. At the same time, through the aggregation and update of the model, the global model can be optimized and improved.

[0045] Step 5: Construct a map of abnormal conditions of power transmission and transformation equipment. Local and global maps are generated by local and global models respectively. The local model detects abnormal places based on the collected data from the edge end and classifies them into three severity levels: low, medium, and high. Then, a local map is built on the edge, including the damage type, level, and location. Each edge provides a quick warning by broadcasting its own local map to neighboring users. After that, each edge sends its own local map to the cloud. By aggregating all local maps, a global map is created on the cloud. The cloud broadcasts the global map to each region. Then, each region can obtain the latest abnormal conditions of the transmission lines across provinces and regions, so that maintenance personnel can arrive for maintenance in time, choose the best travel route, and significantly reduce the damage to dangerous transmission lines and cause greater accidents.

[0046] A system for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions, comprising:

[0047] Data acquisition module: used to collect abnormal data of power transmission and transformation equipment and store the data;

[0048] Hidden danger identification module: used to extract features from abnormal data and label the data;

[0049] Algorithm iteration module: used to train the model at the edge and upload the updated model to the cloud for aggregation and updating;

[0050] Map integration module: used to build local maps on the edge and create global maps on the cloud.

Claims

1. A method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions, characterized in that: The method comprises the following steps: (1) Collect abnormal data of power transmission and transformation equipment; (2) Perform feature engineering on abnormal data to achieve feature extraction, perform data annotation on the abnormal data after feature extraction, and use data standardization to map the feature values ​​to the [0, 1] interval; (3) Constructing the MLP model; (4) MLP model update: A global model is initialized using MLP random parameters and deployed to the edge. After local training is completed on the edge, the updated model parameters are uploaded to the cloud for aggregation. The aggregated model parameters are updated using the gradient descent algorithm. (5) Map building: A local map is built at the edge. Each edge sends its local map to the cloud for aggregation. A global map is created in the cloud and then broadcast to each edge. Each edge obtains the latest abnormal conditions of the transmission lines in a larger area.

2. According to claim 1, a method for assessing the risk of power transmission and transformation equipment failure under extreme weather conditions is characterized in that The step (2) is to perform feature engineering on the abnormal data, introduce a tree model, select feature information including number, name, and range, annotate the abnormal data after feature extraction, fill non-existent data with "0", and use data standardization: Map the eigenvalues ​​to the interval [0,1].

3. The method for assessing the risk of power transmission and transformation equipment failure under extreme weather conditions according to claim 1 is characterized in that The MLP model in step (3) is a three-layer network structure, including an input layer, a hidden layer and an output layer, wherein the input layer is the data input into the MLP model, the hidden layer maps the feature information of the input data, and the output layer is the output data. The output layer uses the Softmax normalization function to convert the data value into a probability. The Softmax function is defined as: Among them, exp(z i ) represents an exponential function with the natural constant e as its base.

4. The method for assessing the risk of power transmission and transformation equipment failure under extreme weather conditions according to claim 1 is characterized in that The ReLU activation function is used in the MLP model to convert the input of the neuron into a nonlinear output. The ReLU function expression is: f(x)=max(0,x), where x is the input value and f(x) is the output of the activation function.

5. The method for assessing the risk of power transmission and transformation equipment failure under extreme weather conditions according to claim 1 is characterized in that The MLP model uses learning rate, weight decay, batch size and number of iterations for optimization and adjustment, divides the data into training set, validation set and test set, uses gradient descent algorithm to optimize parameters, and calculates the loss value and accuracy of the model at the end of each round.

6. The method for assessing the risk of power transmission and transformation equipment failure under extreme weather conditions according to claim 1 is characterized in that The step (4) specifically uses the random parameters of the MLP model to initialize a model as a global model, deploys it to each edge terminal, and each edge terminal uses its own data to locally train the model. After the training is completed, the updated model parameters are uploaded to the cloud server for weighted average aggregation, and the aggregated model parameters are updated using the gradient descent optimization algorithm. The above steps are repeated until the preset model performance is achieved.

7. A method for assessing the risk of power transmission and transformation equipment failure under extreme weather conditions according to claim 1, characterized in that The step (5) specifically collects edge data to build a local model, divides the data into three levels: low, medium and high according to the degree of data anomaly, builds a local map at the edge, including the anomaly type, anomaly level and anomaly location, and the edge sends the local local map to the cloud for aggregation, creates a global map in the cloud, and then broadcasts it to each edge.

8. A system for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions, characterized in that: include: Data acquisition module: used to collect abnormal data of power transmission and transformation equipment and store the data; Hidden danger identification module: used to extract features from abnormal data and label the data; Algorithm iteration module: used to train the model at the edge and upload the updated model to the cloud for aggregation and updating; Map integration module: used to build local maps on the edge and create global maps on the cloud.

9. A computer device, characterized in that: The method comprises one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of a method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for assessing the risk of power transmission and transformation equipment failure under extremely severe weather conditions are implemented as described in any one of claims 1 to 7.

Images