Electric transmission line fire reason identification system and method based on electrical characteristic parameters

Abstract

The invention provides a power transmission line fire reason identification system and method based on electrical characteristic parameters, and relates to the technical field of safety monitoring and fault diagnosis of a power system. Comprising a data acquisition module, a data processing module, a fire risk assessment module, a multi-dimensional analysis module, an early warning module, a multi-source data fusion module, an after-event analysis module and a self-learning mechanism module. Abnormal conditions in line operation can be found in time, fire hazards caused by delayed detection are effectively avoided, the efficiency and timeliness of line monitoring are improved, monitored electrical data are analyzed and processed by adopting a machine learning algorithm, abnormal modes can be automatically recognized, potential causes of fire occurrence can be judged, and the method is suitable for popularization and application. The system not only depends on single threshold alarm, but also provides more accurate fire early warning through deep analysis of historical data and operation trend.

Description

Technical Field

[0001] The present invention relates to the technical field of safety monitoring and fault diagnosis of electric power systems, and in particular to a transmission line fire cause identification system based on electrical characteristic parameters. Background Art

[0002] As an important part of power transmission, the safety and stability of transmission lines directly affect the operation of the entire power grid. Transmission line fires not only cause economic losses, but may also lead to large-scale power outages and affect the normal operation of society. Therefore, how to accurately and timely identify the causes of transmission line fires and reduce fire risks has become a key research direction in the field of power systems.

[0003] At present, the monitoring and identification of the causes of transmission line fires mainly rely on the following technical means:

[0004] Manual inspection: Traditional fire prevention measures include regular manual inspections of transmission lines, using personnel and detection equipment to detect potential risks. Although manual inspections can ensure line safety to a certain extent, this method has problems such as high cost, low efficiency, and poor real-time performance, making it difficult to provide timely warnings and identification of fires.

[0005] Infrared thermal imaging technology: Infrared thermal imaging technology can detect overheated wires, insulators or other equipment by detecting temperature changes in transmission lines. This technology has the advantages of non-contact and strong real-time performance, and can assist in detecting fire precursors. However, it is greatly affected by environmental factors (such as weather, sunshine, etc.), and the cost is high, so it is not suitable for large-scale transmission line monitoring.

[0006] Drone monitoring: With the advancement of drone technology, the use of drones for aerial monitoring of power transmission lines has become an effective means. Drones are equipped with sensors such as infrared imaging and high-definition cameras, which can quickly obtain the operating status of power transmission lines. However, drone monitoring also has problems such as short flight time and difficulty in all-weather monitoring.

[0007] Electrical parameter monitoring: Abnormalities in electrical parameters such as current, voltage, and power can often indicate potential problems in the transmission line. For example, abnormal increase in current may be caused by line short circuit, equipment aging, or other problems, which in turn may cause a fire. By monitoring the electrical characteristic parameters of the transmission line, the health status of the line can be more accurately reflected and provide a basis for early warning of fires.

[0008] Although existing monitoring and identification technologies have played an important role in preventing transmission line fires, they still have many shortcomings:

[0009] Problem 1: Traditional manual inspections are time-consuming, have limited coverage, and cannot achieve real-time monitoring of transmission lines. This makes it difficult to detect potential fire hazards in a timely manner and effectively prevent fires.

[0010] Problem 2: Although the use of advanced equipment such as infrared thermal imaging and drones for monitoring can achieve relatively accurate fire warnings, the equipment is expensive and has high maintenance costs, making it difficult to promote and apply on a large scale, especially on long-distance transmission lines. In addition, infrared imaging is greatly affected by weather and the environment, and its reliability will be reduced under harsh conditions;

[0011] Problem 3: Most of the existing monitoring methods rely on regular inspections or periodic drone patrols, lacking real-time and continuous monitoring capabilities. This non-real-time nature will result in hidden dangers not being discovered in time in the early stages of a fire, thus missing the best time for prevention and control;

[0012] Question 4: Although monitoring electrical characteristic parameters can predict the occurrence of line faults and fires, traditional technologies mainly rely on simple threshold alarms and fail to fully utilize modern data analysis methods (such as machine learning and data mining technology) to conduct in-depth analysis of electrical parameters. Simply relying on set thresholds can easily lead to false alarms and missed alarms, and cannot provide accurate risk assessments. Summary of the invention

[0013] In view of the shortcomings of the prior art, the present invention provides a transmission line fire cause identification system based on electrical characteristic parameters, which solves the problems in traditional technology that inspections are time-consuming, coverage is limited, the causes of transmission line fires cannot be identified in a timely manner, and risk assessment accuracy is low.

[0014] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0015] Transmission line fire cause identification system based on electrical characteristic parameters,

[0016] Data acquisition module, which monitors the high-frequency electrical characteristic parameter data of the transmission line in real time and transmits the data to the cloud platform;

[0017] A data processing module, combining random forest and support vector machine algorithms, performs abnormality detection on the electrical characteristic parameters collected by the data collection module;

[0018] A fire risk assessment module, which uses a long short-term memory network to process time series data and constructs a decision tree model to assess the fire risk level based on the electrical characteristic parameters processed by the data processing module;

[0019] The multi-dimensional analysis module uses a method combining K-means clustering and neural network to perform cluster analysis on the electrical characteristic parameters and environmental data processed by the data processing module to identify patterns of different fire causes;

[0020] The early warning module sets thresholds based on historical data, and sets up an alarm mechanism based on the risk levels assessed by the fire risk assessment module and the fire causes identified by the multi-dimensional analysis module. When the electrical characteristic parameter data exceeds the normal range, a real-time early warning is issued, and the threshold is adjusted according to the real-time input electrical characteristic parameter data.

[0021] The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters comprises the following steps:

[0022] Real-time monitoring and collection of high-frequency electrical characteristic parameter data of transmission lines, and transmission of the data to the cloud platform;

[0023] Combining random forest and support vector machine algorithms, abnormality detection is performed on the electrical characteristic parameters collected by the data acquisition module;

[0024] Using a long short-term memory network to process time series data, and constructing a decision tree model based on the processed electrical characteristic parameters to evaluate the fire risk level;

[0025] A method combining K-means clustering and neural network is used to perform cluster analysis on the processed electrical characteristic parameters and environmental data to identify patterns of different fire causes;

[0026] Thresholds are established based on historical data, and an alarm mechanism is established based on the risk levels of the fire risk assessment and the identified causes of fire. When the electrical characteristic parameter data exceeds the normal range, a real-time warning is issued, and the threshold is adjusted based on the real-time input electrical characteristic parameter data.

[0027] The present invention has the following beneficial effects:

[0028] 1. The present invention can timely detect abnormal conditions in line operation through real-time monitoring of electrical characteristic parameters such as current and voltage of the transmission line, effectively avoid fire hazards caused by delayed detection, and improve the efficiency and timeliness of line monitoring. It uses machine learning algorithms to analyze and process the monitored electrical data, and can automatically identify abnormal patterns and determine the potential causes of fire, reducing the probability of false alarms and missed alarms. The system not only relies on a single threshold alarm, but also provides more accurate fire warnings through in-depth analysis of historical data and operating trends.

[0029] 2. The present invention can automatically issue a fire risk alarm based on the analysis results, prompting relevant personnel to take preventive measures to reduce the probability of fire. In addition, the system can provide corresponding early warning strategies for different risk levels to achieve more flexible and reliable fire prevention.

[0030] 3. Compared with traditional manual inspections or high-cost equipment monitoring, the present invention realizes fire warning through the monitoring of electrical characteristic parameters, which is lower in cost and can be widely promoted and applied. It is suitable for all-weather monitoring of long-distance transmission lines, does not rely on infrared or other environmentally sensitive equipment, and can operate normally under various weather conditions. It overcomes the shortcomings of the prior art that are limited by weather and environmental changes, and improves the reliability of the system. The system can monitor the electrical characteristic parameters of the transmission lines in real time to ensure that abnormal changes in current or voltage are discovered at any time to avoid fire risks. It introduces machine learning or data-driven algorithms to conduct in-depth analysis of electrical parameters and can automatically identify abnormal patterns to determine potential causes of fires. When a high fire risk is detected, the system can promptly issue an alarm to remind relevant personnel to take emergency measures to reduce the possibility of fire. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic block diagram of the specific process of the present invention;

[0032] Figure 2 This is a simulation diagram for analyzing the changes in electrical characteristic parameters of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Specific embodiment 1

[0035] like Figure 1-2 As shown, a method and system for identifying the cause of a power transmission line fire based on electrical characteristic parameters includes a data acquisition module, a data processing module, a fire risk assessment module, a multi-dimensional analysis module, an early warning module, a multi-source data fusion module, a post-analysis module, and a self-learning mechanism module;

[0036] The data acquisition module monitors the high-frequency electrical characteristic parameters of the transmission line, such as current, voltage, frequency, temperature, humidity and wind speed, in real time, and transmits the data to the cloud platform through wireless transmission technology;

[0037] The data processing module combines random forest and support vector machine (SVM) algorithms to detect anomalies in electrical characteristic parameters and improve recognition accuracy;

[0038] The fire risk assessment module uses the long short-term memory network (LSTM) to process time series data and combines the decision tree algorithm to assess the fire risk level;

[0039] The multi-dimensional analysis module uses a combination of K-means clustering and neural networks to perform cluster analysis on electrical characteristics and environmental data, thereby identifying patterns of different fire causes;

[0040] The early warning module sets up thresholds and alarm mechanisms. When the electrical characteristic parameters exceed the normal range, real-time warnings are issued through mobile phone APP and text messages.

[0041] The multi-source data fusion module realizes data layer fusion, feature layer fusion, decision layer fusion, time series fusion and space fusion;

[0042] The post-event analysis module automatically generates a retrospective report of the fire event, analyzes the relationship between the change of electrical characteristic parameters and the occurrence of the fire, and provides a basis for subsequent prevention;

[0043] The self-learning mechanism module has a self-learning function. As new data accumulates, the model is continuously optimized to improve the accuracy of fire cause identification.

[0044] The data acquisition module collects monitoring data from current sensors, voltage sensors, temperature sensors, humidity sensors, wind speed sensors and light sensors in real time, and simultaneously monitors the corresponding electrical characteristic parameters. At the same time, it uses the dual data mining algorithms of association rule mining and cluster analysis to deeply explore the potential relationship between electrical characteristic parameters and fire causes, and sends the data to the cloud in real time through the wireless transmission technology combining LoRa and NB-I oT.

[0045] The data processing module combines random forest, support vector machine (SVM) and deep learning algorithms to perform multi-level anomaly detection on electrical characteristic parameters to improve the accuracy and reliability of fire identification.

[0046] The fire risk assessment module uses the long short-term memory network (LSTM) to process multi-dimensional time series data, combines it with an integrated learning algorithm to conduct a comprehensive assessment of the fire risk level, and formulates an early warning strategy based on the assessment results to send early warning information to relevant personnel in a timely manner.

[0047] The multi-dimensional analysis module adopts a method that combines K-means clustering with adaptive neural networks to dynamically adjust the analysis strategy and identify the clustering patterns of different fire causes. At the same time, it introduces principal component analysis technology to reduce the dimensionality of high-dimensional electrical characteristic parameters and efficiently identify the clustering patterns of different fire causes. At the same time, a dynamic update mechanism is established to continuously optimize the analysis strategy as new data is continuously input.

[0048] The early warning module has an adaptive threshold setting function, can dynamically adjust the alarm mechanism according to historical data, and issue real-time warnings through mobile phone apps, text messages, emails, etc. It can also be linked with the sound and light alarm system of the monitoring center, and dynamically adjust the warning frequency and intensity according to real-time changes in fire risks.

[0049] The multi-source data fusion module adopts a variety of fusion technologies, including data layer fusion, feature layer fusion, decision layer fusion, time series fusion and space fusion, to comprehensively improve the accuracy of fire risk identification. At the same time, a data quality assessment model is established to conduct quality assessment and screening of data before fusion to improve the accuracy of the fusion results.

[0050] The post-analysis module generates a visual retrospective report, which analyzes in detail the relationship between the changes in electrical characteristic parameters and the time, location and mode of the fire, providing empirical basis for subsequent preventive measures. It also covers the detailed analysis process of the cause of the fire and improvement suggestions, and establishes a case database to classify and store historical fire events for quick reference and comparative analysis of subsequent similar events.

[0051] The self-learning mechanism module has the ability to learn online. It optimizes the algorithm model through continuously updated real-time data to improve the accuracy and adaptability of identifying new fire causes. At the same time, it establishes a model performance monitoring indicator system to evaluate the accuracy and adaptability of the model in real time to ensure that the system always maintains good performance.

[0052] The system has a distributed architecture that supports edge computing, delegating data processing and analysis capabilities to edge devices close to the data source, thereby improving real-time response capabilities and reducing data transmission delays. In addition, edge computing nodes have intelligent decision-making capabilities to perform preliminary fire cause identification and early warning locally, reducing data transmission volume and delays. Specific embodiment 2

[0054] like Figure 1-2 As shown, the key algorithm mentioned in Example 1 is analyzed in detail below, including its core mathematical formula and explanation:

[0055] 1. Data acquisition module,

[0056] 1. Electrical characteristic parameter monitoring:

[0057]

[0058] P(t)=V(t).I(t) (power calculation)

[0059] Where I(t) is the current value at time t, V(t) is the voltage value at time t, R is the resistance value, and P(t) is the power value at time t.

[0060] 2. Association rule mining algorithm:

[0061] In the Support formula:

[0062] Assume item set A and item set B, support

[0063] The number of transactions in the dataset that contain both itemset A and itemset B.

[0064] Total Transactions: The total number of transactions in the dataset.

[0065] In the confidence formula:

[0066] Confidence

[0067] The number of transactions containing item set A: The number of transactions containing item set A in the data set. In the implementation, first, the collected electrical characteristic parameter data is preprocessed and converted into a transaction data set suitable for association rule mining. Then, the association rule mining algorithm is used to mine frequent item sets and calculate the support and confidence. Strong association rules are screened out according to the set threshold for subsequent analysis and fire cause identification.

[0068] 3. Cluster analysis algorithm:

[0069]

[0070] Where k: the number of tight classes, that is, the number of groups into which the data is to be divided, C i : The i-th cluster is a subset of the data set.

[0071] x: sample point, that is, a data point in the data set; μ i : The i-th cluster center is a point representing the cluster;

[0072] x-μ i 2: Sample point x to cluster center μ i The square of the Euclidean distance.

[0073] 2. Data processing module

[0074] 1. Random Forest:

[0075] Through the integration of decision trees:

[0076]

[0077] Among them, h i (x) is the prediction result of the i-th decision tree. When implementing, first, determine the parameters of the random forest, such as the number of decision trees, the maximum depth of each decision tree, the number of randomly selected features, etc. Then, randomly extract samples and features from the original data to build multiple decision trees. For new input data, each decision tree makes a prediction, and finally the final prediction result of the random forest is obtained by voting or averaging.

[0078] 2. Support Vector Machine (SVM):

[0079] The goal is to find the optimal hyperplane:

[0080]

[0081] Where w is the normal vector of the hyperplane and b is the bias.

[0082] Random forests classify electrical features by building multiple decision trees, and SVM is used to distinguish between normal and abnormal states and output abnormal alarms.

[0083] 3. Deep Learning Algorithms:

[0084] y=f(W x +b)

[0085] Where x is the input vector, that is, a sample in the data set;

[0086] W is the weight matrix, which is used to perform linear transformation on the input;

[0087] b is the bias vector, which is used to adjust the result of the linear transformation;

[0088] f is the activation function, which is used to introduce nonlinearity;

[0089] y is the output vector, which is the prediction result of the model for the input sample.

[0090] 3. Fire risk assessment module,

[0091] 1. Long Short-Term Memory Network Algorithm:

[0092] f t =σ(W f ·[h t-1 ,x t ]+b f )middle:

[0093] f t: The output of the forget gate at time step t determines which information is discarded from the cell state;

[0094] σ: sigmoid activation function, which compresses the value between 0 and 1;

[0095] W f : The weight matrix of the forget gate;

[0096] h t-1 : The hidden state of the previous time step;

[0097] x t : Input of the current time step;

[0098] b f : Bias term of forget gate.

[0099] i t =σ(W i ·[h t-1 ,x t ]+b i )middle:

[0100] i t : The output of the input gate at time step t determines what new information will be stored in the cell state;

[0101] W i : weight matrix of input gate;

[0102] b i : Bias term for input gate.

[0103] middle:

[0104] Output of candidate cell state at time step t;

[0105] W C : Weight matrix used to calculate candidate cell states;

[0106] b C : Bias term.

[0107] middle:

[0108] C t : The cell state at the current time step;

[0109] C t-1 : The cell state at the previous time step.

[0110] o t =σ(W o ·[h t-1 ,x t ]+bo )middle:

[0111] o t : The output of the output gate at time step t determines which information will be output;

[0112] W o : The weight matrix of the output gate;

[0113] b o : Bias term for the output gate.

[0114] h t =o t *tanh(C t )

[0115] where h t : The hidden state at the current time step.

[0116] LSTM processes time series data, outputs fire risk levels, and generates early warning strategies in combination with decision trees. When implementing, first, build the LSTM network structure and determine parameters such as the number of layers and the number of hidden units. Take multi-dimensional time series data as input for training. During the training process, the network parameters are updated through the back propagation algorithm. For new input data, use the trained LSTM model to perform fire risk assessment.

[0117] 2. Decision Tree:

[0118] By dividing the features, a tree structure is constructed. Each internal node represents a test of the feature, each branch represents a test result, and the leaf node corresponds to a classification result or regression value.

[0119] Decision tree construction process:

[0120] (1) Entropy: It is used to measure the degree of disorder of a data set. For a data set D containing k categories, the formula for calculating entropy is:

[0121]

[0122] where p i It is the proportion of samples of the i-th category in the data set D. The smaller the value of information entropy, the higher the purity of the data set.

[0123] (2) Conditional Entropy: Given a feature A, conditional entropy represents the degree of confusion of the data set when the value of feature A is known. The calculation formula is:

[0124]

[0125] where Values(A) is all possible values ​​of feature A, and D v is the data subset when feature A takes value v, |D v | and |D| are D v and the number of samples of D.

[0126] (3) Information Gain: Information gain indicates the degree to which the chaos of the data set is reduced after being divided using feature A. The calculation formula is:

[0127] Gain(D,A)=Entropy(D)-Entropy(D|A).

[0128] During the construction of the decision tree, the feature with the largest information gain will be selected as the partition feature of the current node.

[0129] Another indicator to measure the purity of a data set. For a data set D containing k categories, the calculation formula of the Gini index is:

[0130]

[0131] Similarly, the conditional Gini index Gini(D|A) given feature A can be calculated, and the information gain can be calculated as Gini(D)-Gini(D|A).

[0132] Steps to build a decision tree:

[0133] (1) Starting from the root node, calculate the information gain or Gini index of all features.

[0134] (2) Select the feature with the largest information gain (or the smallest Gini index) as the partition feature of the current node.

[0135] (3) Divide the data set into multiple subsets based on the different values ​​of the selected features.

[0136] (4) Repeat the above steps for each subset until the stopping condition is met, such as all samples belong to the same category, there are no remaining features for classification, or the preset maximum depth is reached.

[0137] Application of decision tree to transmission line fire risk assessment

[0138] In the fire risk assessment of power transmission lines based on electrical characteristic parameters, a decision tree model can be constructed to determine the fire risk level based on different electrical characteristic parameters and environmental factors. For example, electrical characteristic parameters such as current, voltage, temperature, and humidity can be used as input features of the decision tree, and the fire risk level (high, medium, and low) can be used as the output category. By training on a large amount of historical data, the decision tree can learn the relationship between different feature combinations and fire risks, thereby accurately assessing the fire risk of new input data.

[0139] 4. Multi-dimensional analysis module,

[0140] 1. K-means clustering:

[0141] Objective function:

[0142]

[0143] Among them, μ i is the center of the ith class.

[0144] Principal Component Analysis (PCA):

[0145] This is achieved through eigenvalue decomposition:

[0146] Z=XW

[0147] Where W is the eigenvector matrix.

[0148] K-means clustering is combined with neural networks to dynamically adjust the analysis strategy and reduce the dimension through PCA to improve clustering efficiency and accuracy.

[0149] 5. Early warning module

[0150] Core mathematical formula:

[0151] Adaptive threshold setting:

[0152] Based on statistical analysis:

[0153] T=μ+k·σ

[0154] Among them, μ is the mean of historical data, σ is the standard deviation, and k is the safety factor. The threshold is dynamically adjusted according to historical data, and multi-channel early warning is carried out in combination with the sound and light alarm system.

[0155] The core algorithm here first collects a certain amount of historical data when it is implemented. Then, the initial threshold is calculated based on the statistical characteristics of the historical data, such as the mean, standard deviation, etc. As new data is continuously input, the threshold is updated using methods such as the moving average method or the exponential weighted average method. When the electrical characteristic parameters exceed the normal range, real-time warnings are issued through mobile phone APP, SMS, emails, etc. At the same time, it can be linked with the sound and light alarm system of the monitoring center to dynamically adjust the warning frequency and intensity according to the real-time changes in fire risks.

[0156] 6. Multi-source data fusion module,

[0157] Core mathematical formula:

[0158] Data layer fusion:

[0159] Weighted average fusion:

[0160]

[0161] Among them, w i is the weight, D i Is the data source.

[0162] The data layer, feature layer, decision layer and other multi-level fusion are realized to improve the accuracy of fire risk identification and establish a data quality assessment model. The data layer fusion can directly splice or weighted average the data from different data sources. The feature layer fusion can use principal component analysis, independent component analysis and other methods to extract common features. The decision layer fusion can use voting method, weighted average method and other methods to fuse the decision results of different models. Time series fusion and spatial fusion can be determined according to the specific fusion method. For example, for time series fusion, sliding window method, recursive neural network and other methods can be used. The use of multiple fusion technologies can comprehensively improve the accuracy of fire risk identification. The data layer fusion retains the information of the original data, the feature layer fusion extracts more representative feature information, the decision layer fusion integrates the decision results of different models, the time series fusion considers the time series characteristics of the data, and the spatial fusion combines the spatial location information of the transmission line. First, determine the level and method of fusion. For data layer fusion, the data from different data sources can be directly spliced ​​or weighted averaged. For feature layer fusion, principal component analysis, independent component analysis and other methods can be used to extract common features. For decision-level fusion, voting method, weighted average method and other methods can be used to fuse the decision results of different models. For time series fusion and spatial fusion, they can be implemented according to specific fusion methods. At the same time, a data quality assessment model is established to evaluate and screen the quality of the data before fusion, remove low-quality and unreliable data, and improve the accuracy of the fusion results.

[0163] 7. Post-analysis module

[0164] Correlation analysis:

[0165]

[0166] Where Cov(X,Y) is the covariance, σ x and σ Y is the standard deviation. Through the correlation analysis of the post-analysis module, a visual backtracking report is automatically generated. The relationship between the change of electrical characteristics and the occurrence of fire is analyzed in detail. A case database is established for storage and analysis. When implementing, first collect the relevant data of the fire event, including the change of electrical characteristic parameters, time, location and other information. Then, use data analysis and visualization tools, such as Python's Matplotlib, Seaborn and other libraries, to generate a visual backtracking report. In the report, the changing trend of electrical characteristic parameters, the time and cause of abnormal data, and the impact of different factors on fire are analyzed in detail. At the same time, improvement suggestions are put forward, and historical fire events are stored in the case database for subsequent reference and comparative analysis.

[0167] 8. Self-learning mechanism module,

[0168] Online Learning Algorithms:

[0169] Update rules:

[0170]

[0171] Among them, η is the learning rate, y is the actual value, It is a predicted value. When applied, the model is optimized through real-time data, and a performance monitoring indicator system is established to ensure that the system performance continues to be good.

[0172] 9. Distributed architecture

[0173] Core mathematical formula: edge computing decision,

[0174] Calculate the delay:

[0175] T total =T transmission +T processing

[0176] When applied, preliminary fire cause identification and early warning are carried out at the edge node to reduce data transmission delays and improve response capabilities. These formulas and implementation instructions provide the mathematical basis and technical details for each module of the system, which helps to have a deeper understanding of its working principles and implementation methods. Specific embodiment 3

[0178] like Figure 1-2 As shown, the following is a method for identifying the cause of a transmission line fire based on electrical characteristic parameters:

[0179] The steps include:

[0180] Real-time monitoring and collection of high-frequency electrical characteristic parameter data of transmission lines, and transmission of the data to the cloud platform;

[0181] Combining random forest and support vector machine algorithms, abnormality detection is performed on the electrical characteristic parameters collected by the data acquisition module;

[0182] Using a long short-term memory network to process time series data, and constructing a decision tree model based on the processed electrical characteristic parameters to evaluate the fire risk level;

[0183] A method combining K-means clustering and neural network is used to perform cluster analysis on the processed electrical characteristic parameters and environmental data to identify patterns of different fire causes;

[0184] Thresholds are established based on historical data, and an alarm mechanism is established based on the risk levels of the fire risk assessment and the identified causes of fire. When the electrical characteristic parameter data exceeds the normal range, a real-time warning is issued, and the threshold is adjusted based on the real-time input electrical characteristic parameter data.

[0185] Description of the specific application logic steps of each module and algorithm:

[0186] 1. Data acquisition module application steps:

[0187] Sp1 (Sensor Deployment): Install current, voltage, temperature, humidity, wind speed and light sensors on the transmission lines to monitor electrical characteristic parameters in real time.

[0188] Sp2 (data transmission): Using LoRa and NB-IoT technologies, the monitored data is transmitted to the cloud platform in real time to ensure the real-time and accuracy of the data.

[0189] Sp3 (Data Preprocessing): Clean and standardize the collected data and prepare them into a format suitable for subsequent analysis.

[0190] 2. Data processing module application steps:

[0191] Sp1 (Feature Selection): Use correlation analysis to identify electrical characteristic parameters that have a greater impact on fire identification.

[0192] Sp2 (Random Forest Model): Combined with the random forest algorithm, multiple decision trees are constructed to detect anomalies in electrical characteristics and output the prediction results of each tree.

[0193] Sp3 (SVM classification): Use support vector machine (SVM) to classify abnormal data, determine whether there is a fire risk, and output abnormal alarm.

[0194] 3. Fire risk assessment module application steps

[0195] Sp1 (LSTM network construction): Build a long short-term memory network (LSTM), input multi-dimensional time series data, and determine the number of layers and hidden units of the network.

[0196] Sp2 (model training): Use historical data to train the LSTM model and update the network parameters through the back-propagation algorithm.

[0197] Sp3 (risk level assessment): The new input data is fed into the trained LSTM model, the fire risk level is output, and the corresponding early warning strategy is generated in combination with the decision tree.

[0198] 4. Multi-dimensional analysis module application steps:

[0199] Sp1 (K-means clustering): Use the K-means clustering algorithm to cluster the electrical characteristics and environmental data and determine the number of clusters K.

[0200] Sp2 (Clustering Pattern Recognition): Dynamically adjust the analysis strategy through an adaptive neural network to identify clustering patterns of different fire causes.

[0201] Sp3 (Principal Component Analysis): Principal component analysis (PCA) is introduced to reduce the dimensionality of high-dimensional electrical features and improve clustering efficiency.

[0202] 5. Early warning module application steps:

[0203] Sp1 (Threshold Setting): Calculate the adaptive threshold based on historical data and dynamically adjust the threshold using the mean and standard deviation.

[0204] Sp2 (real-time monitoring): Real-time monitoring of electrical characteristic parameters. Once the threshold is exceeded, the early warning mechanism is triggered.

[0205] Sp3 (multi-channel warning): real-time warning is issued through mobile phone APP, SMS, email, etc., and the sound and light alarm system is linked.

[0206] 6. Multi-source data fusion module application steps:

[0207] Sp1 (data layer fusion): stitching or weighted averaging of data from different sensors to achieve data layer fusion.

[0208] Sp2 (feature layer fusion): Use principal component analysis or independent component analysis to extract common features and achieve feature layer fusion.

[0209] Sp3 (decision-making layer fusion): Use voting method or weighted average method to integrate the decision results of different models to improve the accuracy of risk identification.

[0210] 7. Post-analysis module application steps:

[0211] Sp1 (Data Collection): Collect data related to fire events, including changes in electrical characteristics, time and location, etc.

[0212] Sp2 (Correlation Analysis): Analyze the relationship between changes in electrical characteristics and fire occurrence by calculating covariance and standard deviation.

[0213] Sp3 (Visual Report Generation): Generates visual backtracking reports, analyzes abnormal data and its impact in detail, and stores historical fire events in the case database.

[0214] 8. Self-learning mechanism module application steps:

[0215] Sp1 (online learning): Continuously optimize the algorithm model through real-time data to improve the accuracy of identifying new fire causes.

[0216] Sp2 (Performance Monitoring): Establish a model performance monitoring indicator system to evaluate the accuracy and adaptability of the model in real time.

[0217] 9. Distributed architecture application steps:

[0218] Sp1 (Edge Computing Node Deployment): Deploy computing nodes on edge devices close to data sources to perform preliminary fire cause identification.

[0219] Sp2 (improved real-time response capability): Make intelligent decisions at edge nodes, reduce data transmission delays, and respond quickly to fire risks. Specific embodiment 4

[0221] like Figure 1-2 As shown, the following is a detailed hardware composition and hardware description of each module in Example 1:

[0222] 1. Data acquisition module,

[0223] Hardware composition:

[0224] Current sensor: used to monitor the current value of the transmission line in real time. Hall effect sensors or shunts are usually selected, which have high accuracy and response speed.

[0225] Voltage Sensor: Monitors the voltage of transmission lines, usually using high voltage voltage sensors to ensure safe operation in high voltage environments.

[0226] Temperature Sensors: Use thermocouples or thermistors to monitor temperature changes in the transmission line and the surrounding environment to identify overheating.

[0227] Humidity sensor: monitors ambient humidity, using capacitive or resistive humidity sensors to help analyze the impact of humidity on fire risk.

[0228] Wind Speed ​​Sensor: Use a thermal anemometer or mechanical anemometer to obtain real-time wind speed data to assess the impact of the environment on fire spread.

[0229] Light sensor: Use photoresistors or photoelectric sensors to monitor light intensity and identify the potential impact of external light on fire.

[0230] Data acquisition module: An integrated microcontroller (such as Ardu i no or Raspberry Pi) is responsible for reading sensor data and transmitting the data to the cloud via a LoRa or NB-IoT module.

[0231] 2. Data processing module,

[0232] Hardware composition:

[0233] Server: A high-performance server or cloud computing platform, responsible for receiving, storing and processing data from the data acquisition module. The server configuration must have a multi-core CPU, large-capacity memory and SSD storage.

[0234] GPU acceleration unit: For training and inference of deep learning algorithms (such as LSTM and random forest), configure a GPU (such as NVIDIA series) to accelerate the computing process.

[0235] Software framework: Install corresponding data processing software (such as TensorFlow, PyTorch, Scikit-learn, etc.) for model training and prediction.

[0236] 3. Fire risk assessment module,

[0237] Hardware composition:

[0238] Server / cloud platform: Same as the data processing module, supports large-scale data analysis and the operation of complex models.

[0239] High-performance computing unit: Required for deep learning models (such as LSTM) to ensure the ability to process large-scale time series data.

[0240] 4. Multi-dimensional analysis module,

[0241] Hardware composition:

[0242] Analysis server: A high-performance server that can handle clustering and multi-dimensional data analysis and has high memory and computing power.

[0243] Data visualization tools: Configure data visualization tools (such as Tableau, PowerBI, etc.) to facilitate the display of analysis results.

[0244] 5. Early warning module,

[0245] Hardware composition:

[0246] Alarm device: including sound and light alarm, which can give out sound and flash alarm when fire risk occurs.

[0247] Communication module: Hardware that integrates mobile phone apps and SMS services, usually a cloud-based service interface.

[0248] Server: A cloud server that receives data and manages the warning logic.

[0249] 6. Multi-source data fusion module,

[0250] Hardware composition:

[0251] Fusion server: A server with powerful computing capabilities, responsible for the fusion and analysis of data from different sources.

[0252] Data quality assessment module: can be configured as independent software or integrated into the analysis server to ensure the accuracy and reliability of data fusion.

[0253] 7. Post-analysis module,

[0254] Hardware composition:

[0255] Database server: A database server that stores historical data and fire event analysis results to ensure data security and fast access.

[0256] Data Analysis Workstation: A workstation equipped with a powerful processor and memory for running complex statistical analyses and generating visual reports.

[0257] 8. Self-learning mechanism module,

[0258] Hardware composition:

[0259] Learning server: A server dedicated to online learning and model updating, which must have powerful computing resources and support the processing of real-time data streams.

[0260] 9. Distributed architecture,

[0261] Hardware composition:

[0262] Edge computing devices: Small computing devices (such as edge gateways) deployed near transmission lines for preliminary data processing and intelligent decision-making to reduce latency.

[0263] Connection equipment: used to ensure a stable connection between edge computing nodes and cloud servers, usually including routers and switches.

[0264] 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 any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprising a reference structure" do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0265] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A power transmission line fire cause identification system based on electrical characteristic parameters, characterized by: Data acquisition module, which monitors the high-frequency electrical characteristic parameter data of the transmission line in real time and transmits the data to the cloud platform; A data processing module, combining random forest and support vector machine algorithms, performs abnormality detection on the electrical characteristic parameters collected by the data collection module; A fire risk assessment module, which uses a long short-term memory network to process time series data and constructs a decision tree model to assess the fire risk level based on the electrical characteristic parameters processed by the data processing module; The multi-dimensional analysis module uses a method combining K-means clustering and neural network to perform cluster analysis on the electrical characteristic parameters and environmental data processed by the data processing module to identify patterns of different fire causes; The early warning module sets thresholds based on historical data, and sets up an alarm mechanism based on the risk levels assessed by the fire risk assessment module and the fire causes identified by the multi-dimensional analysis module. When the electrical characteristic parameter data exceeds the normal range, a real-time early warning is issued, and the threshold is adjusted according to the real-time input electrical characteristic parameter data.

2. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1 is characterized in that: The data acquisition module collects monitoring data of current sensors, voltage sensors, temperature sensors, humidity sensors, wind speed sensors and light sensors in real time, monitors corresponding electrical characteristic parameters simultaneously, and uses dual data mining algorithms of association rule mining and cluster analysis to mine the potential relationship between electrical characteristic parameters and fire causes; The data acquisition module relies on the following formula when applied: Electrical characteristic parameter monitoring: P(t)=V(t).I(t) Where I(t) is the current value at time t, V(t) is the voltage value at time t, R is the resistance value, and P(t) is the power value at time t; Association rule mining algorithm: The support formula is: Suppose item set A and item set B, The confidence formula is: Cluster analysis algorithm: Where k: the number of tight classes, that is, the number of groups into which the data is to be divided, C i : The i-th cluster is a subset of the data set. x: sample point, that is, a data point in the data set; μ i : The i-th cluster center is a point representing the cluster; x-μ i 2 : Sample point x to cluster center μ i The square of the Euclidean distance.

3. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: The data processing module relies on the following formula when applied: Random Forest: Through the integration of decision trees: Among them, h i (x) is the prediction result of the i-th decision tree. When implementing, first determine the parameters of the random forest, then randomly extract samples and features from the original data to build multiple decision trees; for new input data, each decision tree makes a prediction, and finally obtains the final prediction result of the random forest by voting or averaging; Support Vector Machine: The goal is to find the optimal hyperplane: Among them, w is the normal vector of the hyperplane, and b is the bias; Random forests classify electrical features by building multiple decision trees, and SVMs are used to distinguish between normal and abnormal states and output abnormal alarms; Deep Learning Algorithms: y=f(W x +b) Where x is the input vector, that is, a sample in the data set; W is the weight matrix, which is used to perform linear transformation on the input; b is the bias vector, which is used to adjust the result of the linear transformation; f is the activation function, which is used to introduce nonlinearity; y is the output vector, which is the prediction result of the model for the input sample.

4. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: The fire risk assessment module relies on the following formula when applied: Long Short-Term Memory Network Algorithm: f t = σ(W f · [h t-1 , x t + b f ) where: f t : The output of the forget gate at time step t determines which information is discarded from the cell state; σ: sigmoid activation function, which compresses the value between 0 and 1; W f : The weight matrix of the forget gate; h t-1 : The hidden state of the previous time step; x t : Input of the current time step; b f : Bias term of forget gate. i t = σ(W i · [h t-1 , x t + b i ) where: i t : The output of the input gate at time step t determines what new information will be stored in the cell state; W i : weight matrix of input gate; b i : Bias term for input gate. middle: Output of candidate cell state at time step t; W C : Weight matrix used to calculate candidate cell states; b C : Bias term; middle: C t : The cell state at the current time step; C t-1 : The cell state at the previous time step. o t = σ(W o · [h t-1 , x t + b o ) where: o t : The output of the output gate at time step t determines which information will be output; W o : The weight matrix of the output gate; b o : Bias term for the output gate. h t =o t *tanh(C t ) where h t : The hidden state of the current time step; LSTM processes time series data, outputs fire risk levels, and generates early warning strategies in combination with decision trees. When implementing, first, the LSTM network structure is constructed, the number of layers and the number of hidden units are determined, and the multi-dimensional time series data is used as input for training. During the training process, the network parameters are updated through the back propagation algorithm. For new input data, the trained LSTM model is used for fire risk assessment.

5. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: The multi-dimensional analysis module relies on the following formula when applied: Among them, μ i is the center of the ith class; Principal Component Analysis (PCA): This is achieved through eigenvalue decomposition: Z=XW Where W is the eigenvector matrix.

6. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: The warning module relies on the following formula when applied: T=μ+k·σ Among them, T is the threshold, μ is the mean of historical data, σ is the standard deviation, and k is the safety factor.

7. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: It also includes a multi-source data fusion module for data layer fusion, feature layer fusion, decision layer fusion, time series fusion and space fusion; establishes a data quality assessment model to assess and screen the quality of data before fusion; Data layer fusion uses weighted average fusion: Among them, w i is the weight, D i Is the data source.

8. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: It also includes a post-analysis module for generating a retrospective report of fire events and analyzing the relationship between changes in electrical characteristic parameters and the occurrence of fire; Correlation analysis formula: Where Cov(X,Y) is the covariance, σ X and σ Y is the standard deviation.

9. The power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: It also includes a self-learning mechanism module, which optimizes the algorithm model by updating rules through online learning algorithms. Online learning algorithm update rules: Among them, η is the learning rate, y is the actual value, is the predicted value.

10. The identification method of the power transmission line fire cause identification system based on electrical characteristic parameters according to claim 1, characterized in that: The steps include: Real-time monitoring and collection of high-frequency electrical characteristic parameter data of transmission lines, and transmission of the data to the cloud platform; Combining random forest and support vector machine algorithms, abnormality detection is performed on the electrical characteristic parameters collected by the data acquisition module; Using a long short-term memory network to process time series data, and constructing a decision tree model based on the processed electrical characteristic parameters to evaluate the fire risk level; A method combining K-means clustering and neural network is used to perform cluster analysis on the processed electrical characteristic parameters and environmental data to identify patterns of different fire causes; Thresholds are established based on historical data, and an alarm mechanism is established based on the risk levels of the fire risk assessment and the identified causes of fire. When the electrical characteristic parameter data exceeds the normal range, a real-time warning is issued, and the threshold is adjusted based on the real-time input electrical characteristic parameter data.

11. The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters according to claim 10, characterized in that: In the data collection step, the monitoring data of the current sensor, voltage sensor, temperature sensor, humidity sensor, wind speed sensor and light sensor are collected in real time, and the corresponding electrical characteristic parameters are monitored synchronously. The dual data mining algorithms of association rule mining and cluster analysis are used to mine the potential relationship between the electrical characteristic parameters and the cause of the fire.

12. The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters according to claim 10, characterized in that: In the data processing step, Use correlation analysis to identify electrical characteristic parameter data that affect fire identification; Combined with the random forest algorithm, multiple decision trees are constructed to detect anomalies in the electrical characteristic parameter data, and the prediction results of each decision tree are output to obtain abnormal data; Use support vector machines to classify abnormal data.

13. The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters according to claim 10, characterized in that: In the fire risk assessment step, Construct a long short-term memory network, input multi-dimensional time series data, and determine the number of layers and hidden units of the network; Use historical data to train the long short-term memory network model and update the network parameters through the back-propagation algorithm; The new input data is fed into the trained long short-term memory network model to output the fire risk level.

14. The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters according to claim 10, characterized in that: In the multi-dimensional analysis step, Using the K-means clustering algorithm, the electrical characteristic parameter data and the environmental data are clustered to determine the cluster number K; Through adaptive neural network dynamic adjustment analysis, clustering patterns of different fire causes can be identified; Principal component analysis is introduced to reduce the dimensionality of high-dimensional electrical characteristic parameter data.

15. The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters according to claim 10, characterized in that: It also includes the multi-source data fusion step. Splice or weighted average the data from different sensors to achieve data layer fusion; Use principal component analysis or independent component analysis to extract common features and achieve feature layer fusion; The decision results of different models are integrated by using voting method or weighted average method.

16. The method for identifying the cause of a power transmission line fire based on electrical characteristic parameters according to claim 10, characterized in that: It also includes a post-analysis step, Collect data related to fire events, including changes in electrical characteristics, time and location information; By calculating the covariance and standard deviation, the relationship between the change of electrical characteristics and the occurrence of fire is analyzed; Generate visual retrospective reports, analyze abnormal data and its impact, and store historical fire events in the case database.

Images