Distribution line ice disaster prediction method and system based on biological optimization neural network

Abstract

The invention relates to the technical field of power system risk assessment, in particular to a distribution line ice disaster prediction method and system based on a biological optimization neural network. Performing first preprocessing on the first distribution line state data to obtain second distribution line state data; presetting a first disconnection prediction model and inputting the second distribution line state data into the first disconnection prediction model; and performing ice disaster disconnection prediction according to the output of the first disconnection prediction model. The first distribution line state data comprises wind speed data, wind direction data, icing thickness data and wire model data; the first disconnection prediction model is a full-connection neural network model optimized based on a red-mouth blue-magpie optimization algorithm, and comprises an input layer, three hidden layers and an output layer; the red-beak blue-streak optimization algorithm comprises the steps of initializing individuals in a population according to a good point set method, and optimizing neural network parameters through small group exploration or group exploration, prey attack operation and optimal solution updating.

Description

Technical Field

[0001] The present invention relates to the technical field of power system risk assessment, and in particular to a distribution line ice disaster prediction method and system based on a biological optimization neural network. Background Art

[0002] With global climate change, icing is becoming more and more frequent in many areas, posing a serious threat to the safety of distribution lines. Icing increases the weight of the conductors, leading to increased conductor tension, which may eventually lead to line disconnection accidents and cause huge losses to the power system. Therefore, it is of great practical significance to study effective methods for predicting line disconnection under icing conditions.

[0003] Traditional disconnection prediction methods are mostly based on physical models and statistical analysis, which have problems such as complex calculations and poor real-time performance. In recent years, the rapid development of artificial intelligence technology has provided new ideas for disconnection prediction. Neural networks, especially fully connected neural networks, have become a hot topic of research due to their advantages in dealing with nonlinear problems. However, how to effectively train neural networks to improve prediction accuracy remains a challenge. Therefore, in order to ensure the safe operation of distribution lines in icing-prone areas, it is very important to establish a distribution line disconnection prediction model with high prediction accuracy and strong real-time performance.

[0004] The Red-billed Blue Magpie Optimization Algorithm is an emerging intelligent optimization algorithm inspired by the foraging behavior of the Red-billed Blue Magpie. By simulating the group collaboration and information sharing of the Red-billed Blue Magpie, the algorithm can efficiently find the global optimal solution in a large-scale search space. Applying the Red-billed Blue Magpie Optimization Algorithm to the training of fully connected neural networks can effectively optimize the network structure and parameters, and improve the accuracy and efficiency of disconnection prediction.

[0005] The present invention aims to use the red-billed blue magpie optimization algorithm to optimize the training process of the fully connected neural network, so as to achieve accurate prediction of the disconnection of distribution lines under icing conditions. This can not only improve the safety and stability of the power system, but also provide technical support for the intelligent transformation of the power industry. Summary of the invention

[0006] In view of the problems existing in the prior art, the present invention is proposed.

[0007] Therefore, the problem to be solved by the present invention is how to solve the problems that traditional disconnection prediction methods are mostly based on physical models and statistical analysis, and have complex calculations and poor real-time performance. The present invention aims to use the red-billed blue magpie optimization algorithm to optimize the training process of the fully connected neural network, so as to achieve accurate prediction of the disconnection of distribution lines under icing conditions. This can not only improve the safety and stability of the power system, but also provide technical support for the intelligent transformation of the power industry.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, an embodiment of the present invention provides a distribution line ice disaster prediction method based on a biological optimization neural network, which includes acquiring first distribution line state data in a first target area, and performing a first preprocessing on the first distribution line state data to obtain second distribution line state data;

[0010] Presetting a first disconnection prediction model, and inputting the second distribution line state data into the first disconnection prediction model;

[0011] Ice disaster power outage prediction is performed based on the output of the first power outage prediction model.

[0012] As a preferred solution of the distribution line ice disaster prediction method based on biological optimization neural network of the present invention, wherein: the first distribution line state data includes: wind speed data, wind direction data, ice thickness data and conductor model data;

[0013] The first distribution line status data is historical operation data acquired according to a preset collection cycle.

[0014] As a preferred solution of the distribution line ice disaster prediction method based on biological optimization neural network described in the present invention, wherein: the first disconnection prediction model is a fully connected neural network model optimized based on the red-billed blue magpie optimization algorithm;

[0015] The fully connected neural network model includes an input layer, three hidden layers and an output layer.

[0016] As a preferred solution of the distribution line ice disaster prediction method based on biological optimization neural network of the present invention, the red-billed blue magpie optimization algorithm includes:

[0017] Initialize the individuals in the population according to the best point set method, and add all individuals to the first unvisited set;

[0018] When individuals in the first unvisited set are visited, small group exploration or group exploration is performed;

[0019] When the exploration is complete, the prey attack operation is performed;

[0020] When the attack is completed, the stored optimal solution is updated.

[0021] As a preferred solution of the distribution line ice disaster prediction method based on biological optimization neural network described in the present invention, wherein: 2 to 5 individuals are randomly selected to form an exploration group during the small group exploration process, and more than 10 individuals are randomly selected to form an exploration group during the group exploration process.

[0022] As a preferred solution of the distribution line ice disaster prediction method based on biological optimization neural network of the present invention, wherein: the first disconnection prediction model includes:

[0023] The input layer includes 7 neurons, which receive the sine and cosine components of wind speed and direction, ice thickness, span, sag and suspension point height difference respectively;

[0024] The first hidden layer includes 128 neurons, the second hidden layer includes 64 neurons, and the third hidden layer includes 32 neurons;

[0025] The output layer outputs the disconnection probability.

[0026] As a preferred solution of the distribution line ice disaster prediction method based on biological optimization neural network described in the present invention, wherein: the binary cross entropy loss function is used as the fitness function in the first disconnection prediction model to optimize the neural network parameters.

[0027] In a second aspect, an embodiment of the present invention provides a distribution line ice disaster prediction system based on a biological optimization neural network, which includes a state data acquisition module, which acquires first distribution line state data in a first target area, and performs a first preprocessing on the first distribution line state data to obtain second distribution line state data;

[0028] A model operation module, presetting a first disconnection prediction model, and inputting the second distribution line state data into the first disconnection prediction model;

[0029] The result acquisition module predicts ice-disaster power outages according to the output of the first power outage prediction model.

[0030] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the distribution line ice disaster prediction method based on the biological optimization neural network as described in the first aspect of the present invention are implemented.

[0031] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the distribution line ice disaster prediction method based on a biological optimized neural network as described in the first aspect of the present invention are implemented.

[0032] The beneficial effects of the present invention are as follows: the present invention provides a 10kV distribution line ice disaster disconnection prediction method based on an optimized fully connected neural network, combines historical data, establishes a corresponding database, constructs a fully connected neural network prediction model, and trains and optimizes the model through a red-billed blue magpie algorithm, which can accurately predict the disconnection of the distribution line under icing conditions; it has the following advantages:

[0033] The red-billed blue magpie optimization algorithm is a new type of meta-heuristic algorithm with strong global search capabilities. It can effectively avoid local optimality, find a better combination of weights and thresholds, and improve search efficiency and optimization effects.

[0034] For complex nonlinear problems such as risk assessment of distribution systems under icing conditions, the fully connected neural network algorithm optimized by the red-billed blue magpie can provide stable optimization effects and improve the generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 The flowchart of the ice disaster prediction method for distribution lines based on biological optimization neural network;

[0037] Figure 2 A computer device diagram for the ice disaster prediction method for distribution lines based on biological optimization neural network;

[0038] Figure 3 Schematic diagram of the process of optimizing the fully connected neural network using the red-billed blue magpie optimization algorithm for the improved distribution line ice disaster prediction method based on biological optimization neural network. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0042] Example 1

[0043] Reference Figure 1-2, which is the first embodiment of the present invention, provides a distribution line ice disaster prediction method based on biological optimization neural network, including:

[0044] S100: Acquire first distribution line status data in a first target area, and perform first preprocessing on the first distribution line status data to obtain second distribution line status data;

[0045] S101: The first distribution line status data includes: wind speed data, wind direction data, ice thickness data, and conductor model data;

[0046] The first distribution line status data is historical operation data acquired according to a preset collection cycle.

[0047] In an embodiment of the present application, the first distribution line status data includes: wind speed data, wind direction data, ice thickness data and conductor model data; in addition, it also includes the collection of historical meteorological data, and the historical meteorological data includes environmental parameters such as temperature, humidity, and precipitation.

[0048] Specifically, the wind speed data may be real-time wind speed data obtained through meteorological stations installed along the distribution lines, the wind direction data may be the obtained wind direction angle information, the ice thickness data may be ice diameter data measured using an online ice monitoring device, and the conductor model data may include information such as conductor specifications, materials, and mechanical properties.

[0049] In an optional embodiment, the collection methods of various types of data can be flexibly combined. For example, when wind speed data comes from a weather station, ice thickness data can be obtained through an image recognition system. The system will dynamically adjust the weight of the data according to the reliability and real-time performance of different data sources to ensure the overall data quality.

[0050] In an optional embodiment, other types of monitoring data can be added according to actual needs. For example, conductor stress sensor data (i.e., the fifth type of data) can be added to directly monitor the stress state of the conductor, thereby improving the accuracy of the prediction. In addition, conductor surface temperature monitoring can also be introduced to evaluate the impact of the ice-melting process on conductor safety.

[0051] In an optional embodiment, if ice thickness data cannot be obtained temporarily at some monitoring points, interpolation estimation can be performed based on data from surrounding monitoring points. However, in this application, in order to ensure the accuracy of the prediction, it is recommended to retain the direct measurement data.

[0052] It should be noted that the above data collection scheme realizes all-round monitoring of the risk of icing on distribution lines. The system ensures the comprehensiveness and accuracy of the prediction through multi-source data fusion, including meteorological data, physical monitoring data and equipment parameter data. At the same time, the use of a variety of monitoring equipment and technologies, such as meteorological stations, icing monitoring devices, stress sensors, etc., realizes redundant backup of data collection and improves the reliability of the system.

[0053] Specifically, the data cleaning steps include:

[0054] Outlier processing: Use the 3σ criterion to identify and process abnormal data points; Missing value processing: For short-term missing values, use linear interpolation to fill in; for long-term missing values, use historical data from the same period to fill in; Noise filtering: Use wavelet transform method to filter out high-frequency noise.

[0055] The data normalization steps include:

[0056] Numerical normalization: unify data of different dimensions into the interval [0,1]; time alignment: ensure that the sampling time of all data is consistent; format unification: unify the storage format and accuracy requirements of data.

[0057] The data validation steps include:

[0058] Physical constraint verification: ensuring that data satisfies the laws of physics; timing continuity verification: checking the temporal continuity of data; cross-station data consistency verification: checking the data correlation between adjacent monitoring stations.

[0059] In a preferred embodiment, different preprocessing strategies are adopted according to the data type: for wind speed and direction data, the focus is on mutation point detection and smoothing; for ice thickness data, attention is paid to the correction of measurement errors and the timeliness verification of data; for conductor model data, the main focus is on data integrity and consistency checks.

[0060] In the embodiment of the present application, the first preprocessing also includes a data quality assessment mechanism, which comprehensively assesses the preprocessed data by setting multi-dimensional quality indicators (such as completeness, accuracy, timeliness, etc.). Only data that pass the quality assessment will enter the subsequent prediction model.

[0061] It should be noted that a reasonable preprocessing process is the key to ensuring the effectiveness of the prediction model. By establishing a strict data preprocessing system, not only the data quality is improved, but also reliable input is provided for model training. At the same time, various parameters in the preprocessing process can be flexibly adjusted according to the actual application scenario to adapt to the characteristics of different regions.

[0062] S200: Preset a first disconnection prediction model, and input the second distribution line state data into the first disconnection prediction model;

[0063] S201: The first disconnection prediction model is a fully connected neural network model optimized based on the red-billed blue magpie optimization algorithm;

[0064] The fully connected neural network model consists of an input layer, three hidden layers, and an output layer.

[0065] In the embodiment of the present application, the construction process of the first disconnection prediction model includes the following key steps:

[0066] Network structure design:

[0067] Input layer: 7 nodes, corresponding to seven key features; first hidden layer: 128 nodes, using Sigmoid activation function; second hidden layer: 64 nodes, using Sigmoid activation function; third hidden layer: 32 nodes, using LeakyReLU activation function; output layer: 1 node, using Sigmoid activation function, output disconnection probability.

[0068] Parameter initialization:

[0069] Weight initialization: using Xavier initialization method; bias initialization: using zero initialization; learning rate: set to 0.001, using adaptive adjustment strategy; batch size: set to 64; number of training rounds: set to 1000 rounds.

[0070] Optimizer configuration:

[0071] Main optimizer: Adam optimizer; momentum parameters: beta1=0.9, beta2=0.999; weight decay: set to 0.0001.

[0072] In an optional embodiment, the network structure can be dynamically adjusted according to the actual data characteristics. For example, when the input characteristics increase, the number of hidden layer nodes can be increased accordingly; or in complex scenarios, the number of hidden layers can be increased to improve the model's expressiveness.

[0073] S202: The red-billed blue magpie optimization algorithm includes:

[0074] Initialize the individuals in the population according to the best point set method, and add all individuals to the first unvisited set;

[0075] When individuals in the first unvisited set are visited, small group exploration or group exploration is performed;

[0076] When the exploration is complete, the prey attack operation is performed;

[0077] When the attack is completed, the stored optimal solution is updated.

[0078] Population initialization:

[0079] Population size: set to 100; search dimension: determined according to the total number of neural network parameters; initialization method: using the improved good point set method; search range: continuous space of [-1,1].

[0080] Small Group Exploration Strategies:

[0081] Group size: randomly select 2-5 individuals; exploration radius: dynamically adjusted according to the progress of iteration; update frequency: updated every 5 iterations

[0082] Local search precision: set to 0.01.

[0083] Group exploration strategy:

[0084] Population size: randomly select 10-20 individuals; information sharing mechanism: based on fitness weighting; search direction: combined with historical optimal solutions; global search step size: decreases with the number of iterations.

[0085] Prey attack strategy:

[0086] Attack conditions: triggered when a better solution is found; Attack intensity: proportional to the gap between the current solution and the optimal solution; Attack range: fine search within a local area; Convergence judgment: stop if there is no improvement for 10 consecutive times.

[0087] In an optional embodiment, various parameters of the algorithm can be adjusted according to the characteristics of the specific problem:

[0088] For simple problems, the population size and number of iterations can be reduced; for complex problems, the search intensity in the exploration phase can be increased; when computing resources are limited, an adaptive parameter adjustment strategy can be adopted.

[0089] It should be noted that the advantages of the red-billed blue magpie optimization algorithm are:

[0090] Through multi-level search strategies, global exploration and local development are balanced; the group collaboration mechanism improves search efficiency; and the improved good point set initialization method improves the quality of the initial solution.

[0091] S203: During the small group exploration process, 2 to 5 individuals are randomly selected to form an exploration group. During the group exploration process, more than 10 individuals are randomly selected to form an exploration group.

[0092] In an embodiment of the present application, the first disconnection prediction model is constructed using a deep learning framework, and a multi-layer neural network is used to accurately predict the risk of disconnection due to icing on distribution lines. During the model training process, binary cross entropy is selected as the main loss function, and the L2 regularization term is introduced to suppress model overfitting, and the regularization coefficient is set to 0.01. Taking into account the difference in the number of disconnection samples and normal samples, the category weights are dynamically adjusted according to the sample ratio during model training to ensure that the model has sufficient learning ability for minority class samples. During the optimization process, a cosine annealing learning rate scheduling strategy is adopted. A large learning rate is maintained at the beginning of training to speed up convergence, and the learning rate is gradually reduced as the training progresses to achieve fine optimization.

[0093] In order to make full use of limited data resources, a scientific data set partitioning scheme is adopted. All data are divided into training set, validation set and test set in a ratio of 7:2:1. The partitioning process adopts a stratified sampling method to ensure that the distribution of samples of different categories in each subset is consistent with the overall data set, avoiding model performance deviation caused by uneven data distribution.

[0094] During the model training process, a strict early stopping strategy is implemented. The system continuously monitors the change in loss value on the validation set. When the performance of the validation set does not improve for 20 consecutive rounds, the early stopping mechanism is triggered. At the same time, the model parameters with the best performance are saved during the training process. When overfitting occurs, it can fall back to the optimal state to ensure the generalization ability of the model.

[0095] In an optional embodiment, model training can also adopt a more complex integrated learning strategy. Specifically, a 5-fold cross-validation method is used to train multiple basic models, and each model is trained on a slightly different data subset to capture different characteristics of the data. When predicting, the prediction results of each model are integrated by soft voting, and the voting weight of each model is dynamically determined based on its performance on the validation set. The final disconnection risk prediction is converted into a binary classification result by setting a probability threshold of 0.5.

[0096] In an embodiment of the present application, in order to further improve the performance and practicality of the model, an adversarial training mechanism can also be introduced. By adding subtle perturbations to the training samples, the robustness of the model to data noise can be improved. At the same time, knowledge distillation technology can be used to first train a large-scale model as a teacher network, and then use it to guide the training of a smaller student network to achieve lightweight models. In addition, the SHAP value analysis method is used to quantitatively evaluate the importance of features to guide feature engineering and model optimization. Before actual deployment, redundant parameters are reduced through model pruning technology to improve the operating efficiency of the model while maintaining prediction accuracy.

[0097] S300: Predicting ice disaster power outages according to the output of the first power outage prediction model.

[0098] In an embodiment of the present application, the system implements real-time risk assessment of distribution lines based on a trained disconnection prediction model. For newly collected monitoring data, real-time preprocessing is first performed. The system converts the raw data into a standard format acceptable to the model, and uses standardized parameters consistent with the training process for data transformation to ensure that the prediction input is consistent with the data distribution during model training. For missing values ​​that may appear in real-time data, the system uses an interpolation algorithm based on time series characteristics to fill in, or replaces it with the value of the nearest neighbor data point. At the same time, the system identifies abnormal data points by setting dynamic thresholds to avoid interference of abnormal values ​​on the prediction results.

[0099] After data preprocessing is completed, the system starts the prediction process. Considering the importance of the prediction task, the system adopts a model integration solution. Multiple trained models predict the input data at the same time. The system comprehensively considers the confidence and historical performance of each model and obtains the final disconnection risk probability through weighted average. According to the size of the risk probability, the system divides the prediction results into different risk levels, corresponding to different early warning strategies.

[0100] In an optional embodiment, the prediction system is deeply integrated with the power grid operation management system. When a high-risk state is detected, the system automatically obtains the weather change trend for the next 24 hours from the weather forecast system, and combines similar cases in historical data to conduct a more detailed assessment of the risk. At the same time, the system automatically generates operation and maintenance suggestions based on the prediction results, including recommended inspection routes, maintenance priority sorting, etc., to provide data support for operation and maintenance decisions.

[0101] In order to ensure the timely delivery and effective processing of early warning information, the system has established a complete early warning push mechanism. When the risk of disconnection is detected, the system automatically generates an early warning report in a standard format, which contains key information such as risk level, impact range, and recommended measures. Early warning information is pushed to relevant responsible persons through multiple channels at the same time. The system also records the time of early warning push and processing status, realizing traceable management of the entire early warning process.

[0102] In the embodiment of the present application, the system adopts a distributed architecture design to support real-time processing and analysis of large-scale data. The data acquisition layer is deployed at each monitoring point, and real-time data upload is achieved through an industrial-grade communication network. The data processing layer uses a distributed computing framework that can process data streams from multiple monitoring points in parallel. The model calculation layer uses a GPU cluster to accelerate the inference process to ensure millisecond response time. The application service layer provides Web and mobile interfaces to support operation and maintenance personnel to view system status and warning information at any time.

[0103] The core functional modules of the system include real-time data acquisition, data preprocessing, model prediction, early warning management, visual display and system management. The real-time data acquisition module supports a variety of industrial communication protocols to achieve unified collection of sensor data of different types. The data preprocessing module is responsible for data cleaning, format conversion and feature extraction to ensure the quality of input data. The early warning management module automatically generates early warning information based on the prediction results, and is responsible for the push and tracking of early warnings. The visual display module displays monitoring data and prediction results in an intuitive chart form, supporting multi-dimensional data analysis and query.

[0104] In actual applications, the system also realizes linkage with other business systems. For example, the connection with the GIS system enables operation and maintenance personnel to visually view the location of risk points on the map; the integration with the work order system ensures that the early warning information can be quickly converted into specific operation and maintenance tasks; the linkage with the video surveillance system provides intuitive on-site image support for risk assessment. This multi-system collaborative approach greatly improves the practicality and operability of early warning information.

[0105] In order to continuously improve system performance, the prediction model is regularly updated online. The system collects warning feedback information, compares and analyzes actual disconnection events with historical warning records, and continuously optimizes warning thresholds and prediction strategies. Through this closed-loop feedback mechanism, the system's prediction accuracy is continuously improved, and the practicality of warnings is continuously enhanced.

[0106] S301: The first disconnection prediction model includes:

[0107] The input layer includes 7 neurons, which receive the sine and cosine components of wind speed and direction, ice thickness, span, sag and suspension point height difference respectively;

[0108] The first hidden layer includes 128 neurons, the second hidden layer includes 64 neurons, and the third hidden layer includes 32 neurons;

[0109] The output layer outputs the disconnection probability.

[0110] S302: In the first disconnection prediction model, a binary cross entropy loss function is used as a fitness function to optimize neural network parameters.

[0111] Furthermore, this embodiment also provides a distribution line ice disaster prediction system based on biological optimization neural network, comprising:

[0112] A status data acquisition module, which acquires the status data of the first distribution line in the first target area, and performs a first preprocessing on the status data of the first distribution line to obtain the status data of the second distribution line;

[0113] A model operation module presets a first disconnection prediction model and inputs the second distribution line state data into the first disconnection prediction model;

[0114] The result acquisition module predicts ice disaster disconnection according to the output of the first disconnection prediction model.

[0115] In summary, by building a multi-source data collection system (including wind speed, wind direction, ice thickness and conductor model data) and combining the preset collection cycle to continuously monitor the historical operation data, we can achieve all-round perception of the status of distribution lines, improve the comprehensiveness and timeliness of the data, and provide high-quality basic data support for subsequent predictions.

[0116] The fully connected neural network is optimized by introducing the red-billed blue magpie optimization algorithm, which simulates the group foraging behavior of the red-billed blue magpie and adopts a combination of small group exploration and group exploration, which significantly enhances the global search capability of the algorithm, effectively avoids falling into the problem of local optimality, and greatly improves the efficiency of neural network parameter optimization.

[0117] By adopting the good point set method to initialize the population and combining it with the dynamic maintenance mechanism of the unvisited set, uniform coverage and efficient exploration of the search space are achieved, the starting quality and convergence speed of the optimization algorithm are improved, and the network training process is made more stable and efficient.

[0118] By designing a multi-level neural network structure (7 neurons in the input layer and 128 / 64 / 32 neurons in the three hidden layers respectively) and using different activation functions, the model's ability to extract and express nonlinear features is enhanced, and the modeling ability for complex ice-covered and broken line scenarios is improved.

[0119] By using the binary cross entropy loss function as the fitness function and combining it with the dynamic update mechanism of the red-billed blue magpie optimization algorithm, precise optimization of network parameters is achieved, which significantly improves the model in terms of prediction accuracy and generalization ability.

[0120] The overall solution breaks through the limitations of traditional physical models' computational complexity and poor real-time performance through the innovative combination of biological optimization algorithms and deep learning, realizes intelligent prediction of the risk of power line disconnection caused by icing, and provides more reliable technical guarantees for the safe operation of the power system.

[0121] The solution also has good scalability and adaptability. It can not only be used to predict ice-covered and disconnected distribution lines, but can also be extended to status monitoring and fault warning of other power equipment, providing a new technical path for the intelligent operation and maintenance of power systems.

[0122] Example 2

[0123] Reference Figure 1 - Figure 3, which is the second embodiment of the present invention.

[0124] A distribution line ice disaster prediction method based on biological optimization neural network includes the following steps:

[0125] S101: The first distribution line status data includes: wind speed data, wind direction data, ice thickness data and conductor model data;

[0126] The first distribution line status data also includes historical meteorological data, including wind speed and wind direction; the conductor model data includes conductor model, line corridor information, and conductor breakage information.

[0127] The collected information is preprocessed, including data cleaning, normalization, and feature engineering, to obtain a data set.

[0128] Specifically, data cleaning, delete missing values. If some records lack key fields such as wind speed, ice thickness, etc., you can consider deleting these data; fill missing values. If some data are missing and the amount is large, deletion may lead to information loss, which can be filled by using the average value.

[0129] Specifically, data normalization is used to normalize the numerical data, wind speed and ice thickness. The formula is as follows:

[0130]

[0131] Specifically, feature engineering converts wind direction into two features: sine and cosine components to eliminate the angle periodicity problem. The formula is as follows:

[0132] wind_dir_sin=sin(wind_direction)

[0133] wind_dir_cos=cos(wind_direction)

[0134] The conductor model is converted into a numerical feature using a unique coding method. For example, if there are three conductor models, A, B, and C, they are converted into 100, 010, and 001 respectively after coding.

[0135] The wire breakage information is encoded as 0 and 1, where 0 represents that the wire is not broken and 1 represents that the wire is broken.

[0136] Based on the optimization of the fully connected neural network based on the red-billed blue magpie algorithm, a distribution line disconnection prediction model under icing conditions was constructed.

[0137] Based on the optimization of the fully connected neural network by the Red-billed Blue Magpie algorithm, the specific steps of constructing the distribution line disconnection prediction model under icing conditions are as follows:

[0138] A fully connected neural network is constructed. The input layer includes 7 neurons, namely the sine and cosine components of wind speed and wind direction, ice thickness, span, sag and suspension point height difference. The first hidden layer has 128 neurons, the second hidden layer has 64 neurons, and the activation function is ReLU.

[0139] ReLU(x)=max(0,x)

[0140] The third hidden layer has 32 neurons, and the activation function uses LeakyReLU to prevent the problem of neuron dead zone. The formula is as follows:

[0141]

[0142] Here α is a very small positive number, which is 0.01.

[0143] The output layer is probability, and the Sigmoid activation function is used to compress the output value into the range of [0,1]. The formula is as follows:

[0144]

[0145] For binary classification problems, the binary cross entropy loss function is used to measure the gap between the actual label and the predicted probability. The formula is as follows and used as the fitness function.

[0146]

[0147] where y i is the true label, is the predicted probability.

[0148] The algorithm for the red-billed blue magpie includes population initialization, finding food, attacking prey, and storing food;

[0149] Population initialization: Use the good point set initialization method to initialize the red-billed blue magpie population;

[0150] Assume that Gs is an s-dimensional Euclidean space, then

[0151] r∈Gs,

[0152] So

[0153] P n (i)=(r1i i ,r2i2,r3i3,...r n i n ),i=1,2,3,...n.

[0154] Where n represents the number of samples, P n (i) represents the good point set, r represents the good point, and is generally taken as

[0155]

[0156] or

[0157] r={e j i}

[0158] Where k is the smallest prime number satisfying (k-3) / 2≥S;

[0159] Calculate the r value,

[0160] r=(r1,r2,...r n )

[0161] in

[0162]

[0163] m i represents the i-th individual;

[0164] Construct a set of m good points:

[0165] P n (i) = {(r1i1, r2i2, ... r n i n )},i=1,2,3,...n

[0166] P n Mapped to the feasible domain where the population is located:

[0167]

[0168] where a j Indicates the lower limit of the current dimension, b j Indicates the upper limit of the current dimension;

[0169] In the process of searching for food, red-billed blue magpies usually act in small groups (2 to 5) or groups (more than 10) to improve search efficiency. When a small group explores for food, the following formula is used for iteration:

[0170]

[0171] Where t represents the current iteration number, X i (t+1) represents the i-th new search agent position, p represents the number of red-billed blue magpies in a small group of 2 to 5 randomly selected from all search individuals, and X m (t) represents the mth individual selected randomly, X i (t) represents the i-th individual, X rs (t) represents the randomly selected search agent in the current iteration, and Rand1 represents a random number between 0 and 1;

[0172] The following formula is used to iterate as the group explores for food:

[0173]

[0174] Where q represents the number of search agents in the cluster when exploring for food, ranging from 10 to n. Again, it is randomly selected from the entire population;

[0175] Attacking prey, when a small group attacks prey, the following formula is used to iterate:

[0176]

[0177] When attacking prey in groups, the following formula is used for iteration:

[0178]

[0179] Where X food (t) indicates the location of the food, Randn represents the random number used to generate the standard normal distribution (mean 0, standard deviation 1);

[0180] Storing food. After completing the foraging attack, the red-billed blue magpie will store excess food, which makes it easier to find the global optimal solution:

[0181]

[0182] in, and Respectively represent the fitness values ​​of the i-th red-billed blue magpie before and after the position update;

[0183] Construct a fully connected neural network, use the red-billed blue magpie optimization algorithm in the step "Red-billed blue magpie algorithm, including population initialization, food search, attacking prey and food storage" to train and optimize the weights and thresholds of each layer of the fully connected neural network, and recalculate the fitness of each parameter according to the updated position parameters. Repeat the above process until the maximum number of iterations is reached.

[0184] The trained prediction model is used to predict the disconnection of distribution lines under icing conditions.

[0185] Example 3

[0186] This embodiment also provides a computer device, which is applicable to a distribution line ice disaster prediction method based on a biological optimization neural network, and includes a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement a forced oscillation detection and positioning method for a distribution network as proposed in the above embodiment.

[0187] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, a forced oscillation detection and positioning method for a distribution network is implemented as proposed in the above embodiment.

[0188] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0189] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0190] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0191] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0192] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0193] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distribution line ice disaster prediction method based on biological optimization neural network, characterized by: The method comprises acquiring first distribution line status data in a first target area, and performing a first preprocessing on the first distribution line status data to obtain second distribution line status data; Presetting a first disconnection prediction model, and inputting the second distribution line state data into the first disconnection prediction model; Ice disaster power outage prediction is performed based on the output of the first power outage prediction model.

2. The method for predicting ice disasters for power distribution lines based on biologically optimized neural networks according to claim 1, characterized in that: The first distribution line status data includes: wind speed data, wind direction data, ice thickness data and conductor model data; The first distribution line status data is historical operation data acquired according to a preset collection cycle.

3. The method for predicting ice disasters for power distribution lines based on biological optimization neural network according to claim 2, characterized in that: The first disconnection prediction model is a fully connected neural network model optimized based on the red-billed blue magpie optimization algorithm; The fully connected neural network model includes an input layer, three hidden layers and an output layer.

4. The method for predicting ice disasters for power distribution lines based on biologically optimized neural networks according to claim 3, characterized in that: The red-billed blue magpie optimization algorithm comprises: Initialize the individuals in the population according to the best point set method, and add all individuals to the first unvisited set; When individuals in the first unvisited set are visited, small group exploration or group exploration is performed; When the exploration is complete, the prey attack operation is performed; When the attack is completed, the stored optimal solution is updated.

5. The method for predicting ice disasters for power distribution lines based on biologically optimized neural networks according to claim 4, characterized in that: During the small group exploration process, 2 to 5 individuals are randomly selected to form an exploration group, and during the group exploration process, more than 10 individuals are randomly selected to form an exploration group.

6. The method for predicting ice disasters for power distribution lines based on biologically optimized neural networks according to claim 5, characterized in that: The first disconnection prediction model comprises: The input layer includes 7 neurons, which receive the sine and cosine components of wind speed and direction, ice thickness, span, sag and suspension point height difference respectively; The first hidden layer includes 128 neurons, the second hidden layer includes 64 neurons, and the third hidden layer includes 32 neurons; The output layer outputs the disconnection probability.

7. The method for predicting ice disasters for power distribution lines based on biologically optimized neural networks according to claim 6, characterized in that: The first disconnection prediction model uses a binary cross entropy loss function as a fitness function to optimize neural network parameters.

8. A distribution line ice disaster prediction system based on biological optimization neural network, based on the distribution line ice disaster prediction method based on biological optimization neural network according to any one of claims 1 to 7, characterized in that: It also includes a status data acquisition module, which acquires the first distribution line status data in the first target area, and performs a first preprocessing on the first distribution line status data to obtain the second distribution line status data; A model operation module, presetting a first disconnection prediction model, and inputting the second distribution line state data into the first disconnection prediction model; The result acquisition module predicts ice-disaster power outages according to the output of the first power outage prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distribution line ice disaster prediction method based on biological optimization neural network according to any one of claims 1 to 7 are implemented.

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 the distribution line ice disaster prediction method based on biological optimization neural network according to any one of claims 1 to 7 are implemented.

Images