Transformer area line loss intelligent supervision system and method

Abstract

The invention relates to the technical field of transformer area line loss supervision, in particular to a transformer area line loss intelligent supervision system and method, and the system comprises a high-frequency data collection module, a data preliminary screening module, a line loss recognition and positioning module, a control center, and a load balancing module. According to the invention, the intelligent electric meters and the sensors are installed at each power consumption point, each distribution transformer and each branch line, so that full-coverage power data acquisition is realized, and the data acquisition frequency is improved to a millisecond level; the system can obtain the operation data of the power system in real time; meanwhile, an edge computing device is integrated in the high-frequency acquisition module, data can be processed and analyzed in real time, real-time monitoring of the power system is achieved, and the line loss problem is found and processed in time; denoising and smoothing processing are carried out on data by adopting wavelet transform, main features are extracted by utilizing principal component analysis (PCA), a Gaussian mixture model is trained according to historical data, modeling is carried out on feature vectors, probability density is calculated, and the depth and precision of data processing are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of line loss supervision in a transformer area, and in particular to an intelligent supervision system and method for line loss in a transformer area. Background Art

[0002] The line loss rate of the substation is an important indicator of the stable, safe and efficient operation of the residential power system, and is also the core reflection of the operating efficiency of the distribution network.

[0003] Chinese Patent Publication No.: CN104730393B discloses a method for detecting line loss rate in a substation, including: determining low-voltage users in the substation according to the connection relationship between the public transformer and the low-voltage users in the distribution network system; obtaining the power consumption of the low-voltage users in the substation; determining the low-voltage power sales of the substation according to the power consumption of the low-voltage users in the substation; determining the power supply of the public transformer in the substation; and determining the line loss rate of the substation by using the low-voltage power sales of the substation and the power supply of the public transformer in the substation when preset conditions are met; the method for detecting line loss rate in the substation disclosed by the invention can accurately determine the low-voltage users in the substation during the process of detecting the line loss rate in the substation, and thus can accurately detect the line loss rate in the substation.

[0004] However, in the management of line loss in the power system, traditional power systems usually adopt a low-frequency data collection strategy, such as collecting data every 15 minutes or every hour. Low-frequency collection cannot reflect the real-time changes of the power system in a timely manner, especially when the load fluctuates greatly. Low-frequency collection cannot capture instantaneous changes, cannot accurately reflect the real-time operating status of the power system, and is difficult to detect and respond to emergencies in a timely manner. In addition, due to the low frequency of data collection, when abnormalities occur in the power system, such as equipment failure, power theft, and load mutation, these abnormalities cannot be captured and processed in the first time, which will lead to delayed discovery and processing of problems, and faults and abnormalities cannot be processed quickly. In addition, load imbalance will also lead to increased line loss. In a three-phase power system, uneven load will lead to unreasonable resource allocation, some equipment and lines will be overloaded, while other resources will not be fully utilized, reducing the overall efficiency of the system. At present, load balancers are usually installed in the power system to automatically adjust the load of each phase, achieve three-phase load balancing, and reduce the uneven load between phases. However, the initial investment and maintenance costs of load balancers, intelligent load management systems, and energy storage systems are high, which increases the overall cost of the system. Summary of the invention

[0005] To this end, the present invention provides a system and method for intelligent line loss supervision in a substation area, so as to overcome the problems of insufficient data collection frequency and untimely capture of abnormal data in the prior art. Although the prior art has improved the frequency and real-time performance of data collection, problems such as high cost, large amount of data, and high system complexity still exist. For the problem of increased line loss caused by uneven load, although technologies such as load balancers and intelligent load management systems provide solutions, their high cost, technical complexity, and insufficient user participation still need to be solved.

[0006] To achieve the above objectives, on the one hand, the present invention provides an intelligent monitoring system for line loss in a transformer area, comprising:

[0007] A high-frequency data acquisition module is used for data acquisition. The high-frequency data acquisition module includes a smart meter and a sensor, and an edge computing device is integrated in the high-frequency acquisition module.

[0008] The data screening module is used to preliminarily screen out abnormal data and provide immediate feedback; it uses data cleaning technology to remove noise and outliers, and extracts key feature information including voltage, current, and power from the cleaned data;

[0009] The line loss identification and positioning module uses the LSTM model to predict line losses based on historical data and real-time data, capturing the changing patterns of load and line losses. It also uses IsolationForest and Autoencoder for anomaly detection, identification and positioning of abnormal line loss events.

[0010] The control center is responsible for commanding and dispatching abnormal alarms and abnormal processing. When the edge computing device or line loss identification and positioning module detects an abnormality, it will provide real-time feedback to the control center and trigger an alarm mechanism.

[0011] The load balancing module monitors the load conditions of each substation in real time based on the high-frequency data acquisition module; it uses a load balancing algorithm to dynamically adjust the load distribution according to the load demand, power supply capacity and historical data of each substation.

[0012] Furthermore, in the high-frequency data acquisition module, smart meters and sensors are deployed at each power point, distribution transformer and branch line; and high-frequency data acquisition modules are added at key nodes to increase the data acquisition frequency to millisecond level;

[0013] Furthermore, the load balancing module also includes a dynamic line loss optimization unit, which combines real-time data and historical data to perform dynamic optimization, uses a machine learning model to predict future load changes, and dynamically adjusts the power supply plan according to the prediction results to minimize line losses;

[0014] On the other hand, the present invention also provides a method for intelligent monitoring of line losses in a transformer area, comprising:

[0015] Step S1, high-frequency data collection, using smart meters and sensors to collect full coverage of power data;

[0016] In step S1, smart meters and sensors are used to collect power data from all areas to achieve real-time monitoring of key parameters such as voltage, current, and power;

[0017] Step S2, preliminary data screening, denoising and removing outliers from the data collected in the high-frequency data collection step, and extracting key feature information including voltage, current, and power from the cleaned data;

[0018] Perform real-time data processing and analysis on the edge computing device in the high-frequency acquisition module, preliminarily screen out abnormal data and provide timely feedback;

[0019] When edge computing devices detect abnormal data, they immediately feed back to the control center and trigger an alarm mechanism;

[0020] Step S2 is the initial data screening. The collected data is denoised and smoothed using wavelet transform to obtain a smoothed data set, and then the main features are extracted using principal component analysis (PCA). The Gaussian mixture model is trained using historical data to calculate the probability density of the feature vector. The feature vector is then modeled using the Gaussian mixture model to calculate the anomaly score and determine whether it is abnormal data based on the set threshold. Once the edge computing device detects abnormal data, it is immediately fed back to the control center and an alarm mechanism is triggered.

[0021] Step S3, line loss identification and location, based on historical data and real-time data, the LSTM model is used to predict line loss and capture the changing rules of load and line loss; IsolationForest and Autoencoder are used for anomaly detection to identify and locate abnormal line loss events;

[0022] In step S3, line loss identification and location are performed by combining the LSTM model and the anomaly detection model; first, the historical data and real-time data are standardized, and the LSTM model is used for model training and real-time prediction. The architecture includes an input layer, an LSTM unit, and an output layer; the input layer processes the feature vectors of historical data and real-time data, and the LSTM unit captures the temporal relationship of the data through the operation of the input gate, the forget gate, and the output gate, and the output layer predicts the future line loss value. LSTM model training uses the mean square error MSE as the loss function to minimize the error between the predicted value and the true value, and optimizes the model parameters through the gradient descent algorithm; after the real-time data input is standardized, the trained LSTM model is used for prediction output;

[0023] Step S4: real-time monitoring feedback. When the edge computing device or line loss identification and positioning module detects an abnormality, real-time feedback is sent to the control center and an alarm mechanism is triggered;

[0024] Step S4 is real-time monitoring feedback. When the edge computing device or line loss identification and positioning module detects an abnormality, it is fed back to the control center in real time and triggers an alarm mechanism;

[0025] Step S5: load balancing and dynamic line loss optimization, performing load balancing and dynamic line loss optimization;

[0026] Load balancing: Based on the real-time monitoring of the load conditions of each substation by the high-frequency data acquisition module, the load balancing algorithm is used to dynamically adjust the load distribution according to the load demand, power supply capacity and historical data of each substation;

[0027] Dynamic line loss optimization: Combine real-time data with historical data, use machine learning models to predict future load changes, dynamically adjust the power supply plan based on the prediction results, and minimize line losses; the dynamic line loss optimization unit performs continuous optimization;

[0028] Step S5 first monitors the load situation of each substation in real time through the high-frequency data acquisition module, and uses the LSTM model to predict the load to obtain the load demand in the future time period; sets the upper limit of the power supply capacity of each substation, performs load balancing, minimizes the load imbalance of each substation, and satisfies the load demand and power supply capacity constraints at the same time; the optimization goal is defined as minimizing the load imbalance, and by constructing the Lagrangian function and solving the optimal load distribution, the load distribution of each substation is dynamically adjusted according to the calculation results;

[0029] Furthermore, the data initial screening method in step S2 includes:

[0030] Use wavelet transform to denoise and smooth the data to obtain a smoothed data set in Including voltage V t 、Current I t , Power P t , W represents the wavelet transform operation, D t is the original data set at time t, including voltage V t 、Current I t , Power P t ;

[0031] Using principal component analysis PCA to extract the main features F t ;

[0032] Train the Gaussian mixture model based on historical data and calculate the feature vector F t The probability density p(F t );

[0033] Model the feature vector using a Gaussian mixture model: Where K is the number of components in the Gaussian mixture model, π k is the weight of the kth Gaussian component, N(F t |u k ,Σ k ) represents the probability density function of the kth Gaussian component: u k is the mean vector of the kth Gaussian component, Σ k is the covariance matrix of the kth Gaussian component;

[0034] Calculate the anomaly score A using the probability density of the feature vector t , determine whether it is abnormal data according to the set threshold τ;

[0035] A t = -logp(F t ), A t is the anomaly score at time t;

[0036] Set the abnormal threshold τ, when A t >τ, it is marked as abnormal data;

[0037] When edge computing devices detect abnormal data, they immediately feed back to the control center and trigger an alarm mechanism;

[0038] Furthermore, the line loss prediction method using the LSTM model in step S3 is:

[0039] Standardize historical and real-time data. Historical data includes voltage V, current I, power P and line loss L, and real-time data includes real-time collected voltage V t 、Current I t , Power P t and the current line loss L t , where μ is the data mean, σ is the data standard deviation, X t ′ and X t These are the data before and after standardization;

[0040] Use LSTM model for model training and real-time prediction;

[0041] The LSTM model architecture is:

[0042] Input layer: feature vectors of historical data and real-time data, X t Refers to the input feature vector at time t, including the voltage V t 、Current I t , Power P t and the current line loss Lt ;

[0043] LSTM unit, including input gate, forget gate and output gate:

[0044]

[0045]

[0046]

[0047]

[0048]

[0049] h t =o t ⊙tanh(C t );where f t is the forget gate activation vector, i t is the input gate activation vector, is the candidate memory vector, C t is the memory unit vector, o t is the output gate activation vector, h t represents the hidden state vector, W f , W i , W C , W o is the weight matrix, b f , b i , b C , b o is the bias vector, σ is the Sigmoid activation function, tanh represents the hyperbolic tangent activation function, and ⊙ is the element-wise product;

[0050] Output layer: predict future line loss values, in is the predicted line loss value at future time t+1, W y is the output weight matrix, b y represents the output bias vector;

[0051] Furthermore, the LSTM model training and real-time prediction methods are as follows:

[0052] Use mean square error MSE as the loss function to minimize the error between the predicted value and the true value. Where L is the loss function, N is the number of training samples, is the predicted line loss value, L t is the actual line loss value;

[0053] Use the gradient descent algorithm to optimize the LSTM model parameters. Where θ is the model parameter, η is the learning rate, Represents the gradient of the loss function with respect to the model parameters;

[0054] Input the standardized real-time data and use the trained LSTM model to predict the output

[0055] Furthermore, the abnormality detection method using IsolationForest and Autoencoder in step S3 is:

[0056] Collect historical voltage, current, power and line loss data; collect voltage, current, power and current line loss data in real time;

[0057] Use feature vectors from historical data to train the IsolationForest model;

[0058] Use the trained IsolationForest model to predict real-time data and calculate the anomaly score;

[0059] Set anomaly threshold and determine whether the data is abnormal based on the calculated anomaly score;

[0060] Use feature vectors from historical data to train the Autoencoder model, including encoder and decoder networks;

[0061] Reconstruct the input data through the Autoencoder model and calculate the reconstruction error;

[0062] Reconstruct the real-time data and calculate the reconstruction error;

[0063] Set the reconstruction error threshold and determine whether it is abnormal data based on the calculated reconstruction error;

[0064] If the reconstruction error exceeds the threshold, it is marked as abnormal data;

[0065] Compare the abnormal data detected by IsolationForest and Autoencoder to confirm the location and time of the abnormal event;

[0066] When an abnormal event is detected, the abnormal information is immediately fed back to the control center;

[0067] In terms of anomaly detection, the IsolationForest model is used to predict real-time data, calculate anomaly scores, set anomaly thresholds, and determine whether it is abnormal data. At the same time, the feature vectors in historical data are used to train the Autoencoder model, and the input data is reconstructed through the encoder and decoder network, the reconstruction error is calculated, the reconstruction error threshold is set, and whether it is abnormal data is determined. By comparing the two models, the location and time of the abnormal event are confirmed and fed back to the control center in a timely manner.

[0068] Furthermore, the load balancing method is:

[0069] The high-frequency data acquisition module monitors the load conditions of each substation in real time, including the voltage V t 、Current I t , Power P t ;

[0070] Collect historical load data of each substation area t ;

[0071] The LSTM machine learning model is used to predict the load of each substation and obtain the load demand in the future time period, which is expressed as in represents the predicted load demand of the ith substation at the future time t+1; X i,t is the input feature vector of the ith station at time t;

[0072] Set the upper limit of power supply capacity of each area C i , L i,max ≤C i , where L i,max is the maximum load of the ith area, C i is the power supply capacity of the ith substation;

[0073] To balance the load, the optimization objective is defined as: minimizing the load imbalance in each substation while satisfying the load demand and power supply capacity constraints, that is, Here L i is the actual load of the ith substation, is the average load of all substations, N here represents the number of substations, and w i is the weight of the ith station;

[0074] Load Balancing: is the average load of all substations, L i represents the actual load of the ith substation, and N is the number of substations;

[0075] Constraints:

[0076] Construct Lagrangian function and solve the optimal load distribution;

[0077] Dynamically adjust the load distribution of each substation according to the calculation results of the load balancing algorithm;

[0078] Furthermore, the dynamic optimization step includes:

[0079] Step a. Data collection, collecting real-time load data;

[0080] Step b. Load forecasting, using machine learning models to predict short-term and long-term load changes;

[0081] Step c. Optimization calculation, using an optimization algorithm to calculate the optimal load distribution plan based on the prediction results;

[0082] Step d. Adjust the power supply plan and dynamically adjust the load distribution of each substation based on the optimization calculation results to reduce line losses.

[0083] Compared with the prior art, the present invention has the following beneficial effects:

[0084] The present invention installs smart meters and sensors at each power consumption point, distribution transformer and branch line to achieve full coverage of power data collection and increase the data collection frequency to millisecond level; enabling the system to obtain the operation data of the power system in real time; and integrating edge computing equipment in the high-frequency collection module to process and analyze data in real time, achieve real-time monitoring of the power system, and promptly discover and handle line loss problems;

[0085] The present invention uses wavelet transform to denoise and smooth the data, uses principal component analysis PCA to extract the main features, and trains a Gaussian mixture model based on historical data to model the feature vectors and calculate the probability density, thereby improving the depth and accuracy of data processing.

[0086] To solve the problem of poor anomaly detection effect, the present invention uses IsolationForest and Autoencoder anomaly detection algorithms to achieve efficient and accurate anomaly detection by training models, calculating anomaly scores, and reconstructing errors. The dual detection mechanism improves the reliability of anomaly identification and reduces missed reports and false positives. Once the edge computing device detects abnormal data, it immediately feeds back to the control center and triggers an alarm mechanism to ensure that abnormal events can be handled in a timely manner.

[0087] In order to solve the problem of inaccurate line loss prediction, the present invention adopts the LSTM long short-term memory network model to perform training and prediction after standardizing historical data and real-time data, thereby improving the accuracy and reliability of line loss prediction. The high-frequency data acquisition module is used to monitor the load conditions of each substation in real time, and the LSTM model is used to predict future load demand. The load distribution is dynamically adjusted according to the power supply capacity to ensure load balance and stability. The real-time data and historical data are combined to use the machine learning model to predict future load changes, dynamically adjust the power supply plan, and minimize line loss.

[0088] The problems of insufficient data collection frequency and untimely capture of abnormal data in the existing technology have been solved. Although the existing technology has improved the frequency and real-time performance of data collection, problems such as high cost, large data volume and high system complexity still exist. For the problem of increased line loss caused by uneven load, although technologies such as load balancers and intelligent load management systems provide solutions, their high cost, technical complexity and insufficient user participation still need to be solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 It is a schematic diagram of the process flow of the intelligent monitoring method of the transmission line loss in the transformer area of ​​the present invention;

[0090] Figure 2 It is a schematic diagram of the structure of the intelligent line loss supervision system of the substation area of ​​the present invention. DETAILED DESCRIPTION

[0091] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0092] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0093] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0094] Example 1

[0095] See also Figure 1-Figure 2 The present invention provides an intelligent monitoring system for line loss in a transformer area, comprising:

[0096] A high-frequency data acquisition module is used for data acquisition. The high-frequency data acquisition module includes a smart meter and a sensor, and an edge computing device is integrated in the high-frequency acquisition module.

[0097] In the high-frequency data acquisition module, smart meters and sensors are deployed at each power point, distribution transformer and branch line; and high-frequency data acquisition modules are added at key nodes to increase the data acquisition frequency to millisecond level;

[0098] The data screening module is used to preliminarily screen out abnormal data and provide immediate feedback; it uses data cleaning technology to remove noise and outliers, and extracts key feature information including voltage, current, and power from the cleaned data;

[0099] The line loss identification and positioning module uses the LSTM model to predict line losses based on historical data and real-time data, capturing the changing patterns of load and line losses. It also uses IsolationForest and Autoencoder for anomaly detection, identification and positioning of abnormal line loss events.

[0100] The control center is responsible for commanding and dispatching abnormal alarms and abnormal processing. When the edge computing device or line loss identification and positioning module detects an abnormality, it will provide real-time feedback to the control center and trigger an alarm mechanism.

[0101] The load balancing module monitors the load conditions of each substation in real time based on the high-frequency data acquisition module; the load balancing algorithm is used to dynamically adjust the load distribution according to the load demand, power supply capacity and historical data of each substation; the load balancing module also includes a dynamic line loss optimization unit, which combines real-time data and historical data to perform dynamic optimization, uses a machine learning model to predict future load changes, and dynamically adjusts the power supply plan according to the prediction results to minimize line losses;

[0102] The present invention also provides a method for intelligent monitoring of line losses in a transformer area, comprising:

[0103] Step S1, high-frequency data collection, using smart meters and sensors to collect full coverage of power data;

[0104] Step S2, preliminary data screening, denoising and removing outliers from the data collected in the high-frequency data collection step, and extracting key feature information including voltage, current, and power from the cleaned data;

[0105] Perform real-time data processing and analysis on the edge computing device in the high-frequency acquisition module, preliminarily screen out abnormal data and provide timely feedback;

[0106] When edge computing devices detect abnormal data, they immediately feed back to the control center and trigger an alarm mechanism;

[0107] The data initial screening method in step S2 includes:

[0108] Use wavelet transform to denoise and smooth the data to obtain a smoothed data set in Including voltage V t 、Current I t , Power P t , W represents the wavelet transform operation, D t is the original data set at time t, including voltage V t 、Current I t , Power P t ;

[0109] Using principal component analysis PCA to extract the main features F t ;

[0110] Train the Gaussian mixture model based on historical data and calculate the feature vector F t The probability density p(F t );

[0111] Model the feature vector using a Gaussian mixture model: Where K is the number of components in the Gaussian mixture model, π k is the weight of the kth Gaussian component, N(F t |u k ,Σ k ) represents the probability density function of the kth Gaussian component: u k is the mean vector of the kth Gaussian component, Σ k is the covariance matrix of the kth Gaussian component;

[0112] Calculate the anomaly score A using the probability density of the feature vector t , determine whether it is abnormal data according to the set threshold τ;

[0113] A t = -logp(F t ), A t is the anomaly score at time t;

[0114] Set the abnormal threshold τ, when A t >τ, it is marked as abnormal data;

[0115] When edge computing devices detect abnormal data, they immediately feed back to the control center and trigger an alarm mechanism;

[0116] Specifically: In terms of data processing, wavelet transform, principal component analysis PCA and Gaussian mixture model are used for comprehensive data cleaning and feature extraction; real-time processing by edge computing devices can achieve preliminary screening and immediate feedback of abnormal data, improving the efficiency and accuracy of data processing; the anomaly detection module enhances the reliability of abnormal event identification through dual detection of IsolationForest and Autoencoder models, and timely feedback of abnormal information helps to quickly locate and solve problems in the power system;

[0117] Step S3, line loss identification and location, based on historical data and real-time data, the LSTM model is used to predict line loss and capture the changing rules of load and line loss; IsolationForest and Autoencoder are used for anomaly detection to identify and locate abnormal line loss events;

[0118] The line loss prediction method using the LSTM model in step S3 is:

[0119] Standardize historical and real-time data. Historical data includes voltage V, current I, power P and line loss L, and real-time data includes real-time collected voltage V t 、Current I t , Power P t and the current line loss L t , where μ is the data mean, σ is the data standard deviation, X t ′ and X t These are the data before and after standardization;

[0120] Use LSTM model for model training and real-time prediction;

[0121] The LSTM model architecture is:

[0122] Input layer: feature vectors of historical data and real-time data, Xt refers to the input feature vector at time t, including the voltage V t 、Current I t , Power P t and the current line loss L t ;

[0123] LSTM unit, including input gate, forget gate and output gate:

[0124]

[0125]

[0126]

[0127]

[0128]

[0129] h t =o t ⊙tanh(C t );where f t is the forget gate activation vector, i t is the input gate activation vector, is the candidate memory vector, C t is the memory unit vector, o t is the output gate activation vector, h t represents the hidden state vector, W f , W i , W C , W o is the weight matrix, b f , b i , b C , b o is the bias vector, σ is the Sigmoid activation function, tanh represents the hyperbolic tangent activation function, and ⊙ is the element-wise product;

[0130] Output layer: predict future line loss values, in is the predicted line loss value at future time t+1, W y is the output weight matrix, b y represents the output bias vector;

[0131] The LSTM model training and real-time prediction methods are as follows:

[0132] Use mean square error MSE as the loss function to minimize the error between the predicted value and the true value. Where L is the loss function, N is the number of training samples, is the predicted line loss value, L t is the actual line loss value;

[0133] Use the gradient descent algorithm to optimize the LSTM model parameters. Where θ is the model parameter, η is the learning rate, Represents the gradient of the loss function with respect to the model parameters;

[0134] Input the standardized real-time data and use the trained LSTM model to predict the output

[0135] In step S3, the method of using IsolationForest and Autoencoder to perform anomaly detection is as follows:

[0136] Collect historical voltage, current, power and line loss data; collect voltage, current, power and current line loss data in real time;

[0137] Use feature vectors from historical data to train the IsolationForest model;

[0138] Use the trained IsolationForest model to predict real-time data and calculate the anomaly score;

[0139] Set anomaly threshold and determine whether the data is abnormal based on the calculated anomaly score;

[0140] Use feature vectors from historical data to train the Autoencoder model, including encoder and decoder networks;

[0141] Reconstruct the input data through the Autoencoder model and calculate the reconstruction error;

[0142] Reconstruct the real-time data and calculate the reconstruction error;

[0143] Set the reconstruction error threshold and determine whether it is abnormal data based on the calculated reconstruction error;

[0144] If the reconstruction error exceeds the threshold, it is marked as abnormal data;

[0145] Compare the abnormal data detected by IsolationForest and Autoencoder to confirm the location and time of the abnormal event;

[0146] When an abnormal event is detected, the abnormal information is immediately fed back to the control center;

[0147] Specifically: In terms of line loss prediction and load balancing, the LSTM model makes line loss prediction more accurate and effectively captures the changing patterns of load and line loss; by combining historical data with real-time data, the LSTM model can dynamically predict future line loss conditions; in addition, the load balancing algorithm optimizes load distribution, reduces load imbalance, and ensures the stable operation of the power system; based on the load demand, power supply capacity, and historical data of each substation, the load distribution is dynamically adjusted to effectively reduce line losses;

[0148] Step S4: real-time monitoring feedback. When the edge computing device or line loss identification and positioning module detects an abnormality, real-time feedback is sent to the control center and an alarm mechanism is triggered;

[0149] Step S5: load balancing and dynamic line loss optimization, performing load balancing and dynamic line loss optimization;

[0150] Load balancing: Based on the real-time monitoring of the load conditions of each substation by the high-frequency data acquisition module, the load balancing algorithm is used to dynamically adjust the load distribution according to the load demand, power supply capacity and historical data of each substation;

[0151] Dynamic line loss optimization: Combine real-time data with historical data, use machine learning models to predict future load changes, dynamically adjust the power supply plan based on the prediction results, and minimize line losses; the dynamic line loss optimization unit performs continuous optimization;

[0152] The load balancing method is:

[0153] The high-frequency data acquisition module monitors the load conditions of each substation in real time, including the voltage V t 、Current I t , Power P t ;

[0154] Collect historical load data of each substation area t ;

[0155] The LSTM machine learning model is used to predict the load of each substation and obtain the load demand in the future time period, which is expressed as in represents the predicted load demand of the ith substation at the future time t+1; X i,t is the input feature vector of the ith station at time t;

[0156] Set the upper limit of power supply capacity of each area C i , L i,max ≤C i , where L i,max is the maximum load of the ith area, C i is the power supply capacity of the ith substation;

[0157] To balance the load, the optimization objective is defined as: minimizing the load imbalance in each substation while satisfying the load demand and power supply capacity constraints, that is, Here L i is the actual load of the ith substation, is the average load of all substations, N here represents the number of substations, and w i is the weight of the ith station;

[0158] Load Balancing: is the average load of all substations, L i represents the actual load of the ith substation, and N is the number of substations;

[0159] Constraints:

[0160] Construct Lagrangian function and solve the optimal load distribution;

[0161] Dynamically adjust the load distribution of each substation according to the calculation results of the load balancing algorithm;

[0162] The dynamic optimization steps include:

[0163] Step a. Data collection, collecting real-time load data;

[0164] Step b. Load forecasting, using machine learning models to predict short-term and long-term load changes;

[0165] Step c. Optimization calculation, using an optimization algorithm to calculate the optimal load distribution plan based on the prediction results;

[0166] Step d. Adjust the power supply plan, dynamically adjust the load distribution of each substation according to the optimization calculation results, and reduce line losses;

[0167] Specifically: The dynamic line loss optimization unit uses a machine learning model to predict future load changes based on real-time data and historical data, and dynamically adjusts the power supply plan according to the prediction results to minimize line losses; this process ensures the efficient operation of the power system under different load conditions through continuous optimization; by optimizing calculations and adjusting the power supply plan, line losses can be further reduced, thereby improving the economy and reliability of the power system.

[0168] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. An intelligent monitoring system for line loss in a transformer area, characterized in that: include: A high-frequency data acquisition module is used for data acquisition. The high-frequency data acquisition module includes a smart meter and a sensor, and an edge computing device is integrated in the high-frequency acquisition module. The data screening module is used to preliminarily screen out abnormal data and provide immediate feedback; it uses data cleaning technology to remove noise and outliers, and extracts key feature information including voltage, current, and power from the cleaned data; The line loss identification and positioning module uses the LSTM model to predict line losses based on historical data and real-time data, capturing the changing patterns of load and line losses. It also uses IsolationForest and Autoencoder for anomaly detection, identification and positioning of abnormal line loss events. The control center is responsible for commanding and dispatching abnormal alarms and abnormal processing. When the edge computing device or line loss identification and positioning module detects an abnormality, it will provide real-time feedback to the control center and trigger an alarm mechanism. The load balancing module monitors the load conditions of each substation in real time based on the high-frequency data acquisition module; it uses a load balancing algorithm to dynamically adjust the load distribution according to the load demand, power supply capacity and historical data of each substation.

2. According to claim 1, the intelligent monitoring system for line loss in a transformer area is characterized in that: In the high-frequency data acquisition module, smart meters and sensors are deployed at each power consumption point, distribution transformer and branch line; and high-frequency data acquisition modules are added at key nodes to increase the data acquisition frequency to millisecond level.

3. According to claim 2, the intelligent monitoring system for line loss in a transformer area is characterized in that: The load balancing module also includes a dynamic line loss optimization unit, which combines real-time data and historical data to perform dynamic optimization, uses a machine learning model to predict future load changes, and dynamically adjusts the power supply plan based on the prediction results.

4. A method for applying to a transformer area line loss intelligent monitoring system as claimed in any one of claims 1 to 3, characterized in that: The intelligent monitoring method for line loss in a transformer area comprises: Step S1, high-frequency data collection, using smart meters and sensors to collect full coverage of power data; Step S2, preliminary data screening, denoising and removing outliers from the data collected in the high-frequency data collection step, and extracting key feature information including voltage, current, and power from the cleaned data; Perform real-time data processing and analysis on the edge computing device in the high-frequency acquisition module, preliminarily screen out abnormal data and provide timely feedback; When edge computing devices detect abnormal data, they immediately feed back to the control center and trigger an alarm mechanism; Step S3, line loss identification and location, based on historical data and real-time data, the LSTM model is used to predict line loss and capture the changing patterns of load and line loss; IsolationForest and Autoencoder are used for anomaly detection to identify and locate abnormal line loss events; Step S4: real-time monitoring feedback. When the edge computing device or line loss identification and positioning module detects an abnormality, real-time feedback is sent to the control center and an alarm mechanism is triggered; Step S5, load balancing and dynamic line loss optimization, performs load balancing and dynamic line loss optimization.

5. The method for intelligent monitoring of line loss in a transformer area according to claim 4, characterized in that: The data initial screening method in step S2 includes: Use wavelet transform to denoise and smooth the data to obtain a smoothed data set in Including voltage V t 、Current I t , power P t , W represents the wavelet transform operation, D t is the original data set at time t, including voltage V t 、Current I t , power P t ; Using principal component analysis PCA to extract the main features F t ; Train the Gaussian mixture model based on historical data and calculate the feature vector F t The probability density p(F t ); Model the feature vector using a Gaussian mixture model: Where K is the number of components in the Gaussian mixture model, π k is the weight of the kth Gaussian component, N(F t |u k ,Σ k ) represents the probability density function of the kth Gaussian component: u k is the mean vector of the kth Gaussian component, Σ k is the covariance matrix of the kth Gaussian component; Calculate the anomaly score A using the probability density of the feature vector t , determine whether it is abnormal data according to the set threshold τ; A t = -logp(F t ), A t is the anomaly score at time t; Set the abnormal threshold τ, when A t >τ, it is marked as abnormal data; When the edge computing device detects abnormal data, it immediately feeds back to the control center and triggers the alarm mechanism.

6. The method for intelligent monitoring of line loss in a transformer area according to claim 5, characterized in that: The line loss prediction method using the LSTM model in step S3 is: Standardize historical and real-time data. Historical data includes voltage V, current I, power P and line loss L, and real-time data includes real-time collected voltage V t 、Current I t , power P t and the current line loss L t , where μ is the data mean, σ is the data standard deviation, X′ t and X t These are the data before and after standardization; Use LSTM model for model training and real-time prediction; The LSTM model architecture is: Input layer: feature vectors of historical data and real-time data, Xt refers to the input feature vector at time t, including the voltage V t 、Current I t , power P t and the current line loss L t ; LSTM unit, including input gate, forget gate and output gate: h t =o t ⊙tanh(C t );where f t is the forget gate activation vector, i t is the input gate activation vector, is the candidate memory vector, C t is the memory unit vector, o t is the output gate activation vector, h t represents the hidden state vector, W f , W i , W C , W o is the weight matrix, b f , b i , b C , b o is the bias vector, σ is the Sigmoid activation function, tanh represents the hyperbolic tangent activation function, and ⊙ is the element-wise product; Output layer: predict future line loss values, in 1 is the predicted line loss value at future time t+1, W y is the output weight matrix, b y Represents the output bias vector.

7. The method for intelligent monitoring of line loss in a transformer area according to claim 6, characterized in that: The LSTM model training and real-time prediction methods are as follows: Use mean square error MSE as the loss function to minimize the error between the predicted value and the true value. Where L is the loss function, N is the number of training samples, is the predicted line loss value, L t is the actual line loss value; Use the gradient descent algorithm to optimize the LSTM model parameters. Where θ is the model parameter, η is the learning rate, Represents the gradient of the loss function with respect to the model parameters; Input the standardized real-time data and use the trained LSTM model to predict the output 8. The method for intelligent monitoring of line loss in a transformer area according to claim 7, characterized in that: In step S3, the method of using IsolationForest and Autoencoder to perform anomaly detection is as follows: Collect historical voltage, current, power and line loss data; collect voltage, current, power and current line loss data in real time; Use feature vectors from historical data to train the IsolationForest model; Use the trained IsolationForest model to predict real-time data and calculate the anomaly score; Set anomaly threshold and determine whether the data is abnormal based on the calculated anomaly score; Use feature vectors from historical data to train the Autoencoder model, including encoder and decoder networks; Reconstruct the input data through the Autoencoder model and calculate the reconstruction error; Reconstruct the real-time data and calculate the reconstruction error; Set the reconstruction error threshold and determine whether it is abnormal data based on the calculated reconstruction error; If the reconstruction error exceeds the threshold, it is marked as abnormal data; Compare the abnormal data detected by IsolationForest and Autoencoder to confirm the location and time of the abnormal event; When an abnormal event is detected, the abnormal information is immediately fed back to the control center.

9. The method for intelligent monitoring of line loss in a transformer area according to claim 8, characterized in that: The load balancing method is: The high-frequency data acquisition module monitors the load conditions of each substation in real time, including the voltage V t 、Current I t , power P t ; Collect historical load data of each substation area t ; The LSTM machine learning model is used to predict the load of each substation and obtain the load demand in the future time period, which is expressed as in represents the predicted load demand of the ith substation at the future time t+1; X i,t is the input feature vector of the ith station at time t; Set the upper limit of power supply capacity of each area C i , L i,max ≤C i , where L i,max is the maximum load of the ith area, C i is the power supply capacity of the ith substation; To balance the load, the optimization objective is defined as: minimizing the load imbalance in each substation while satisfying the load demand and power supply capacity constraints, that is, Here L i is the actual load of the ith substation, is the average load of all substations, N here represents the number of substations, and w i is the weight of the ith station; Load Balancing: is the average load of all substations, L i represents the actual load of the ith substation, and N is the number of substations; Constraints: Construct Lagrangian function and solve the optimal load distribution; According to the calculation results of the load balancing algorithm, the load distribution of each substation is dynamically adjusted.

10. The method for intelligent monitoring of line loss in a transformer area according to claim 9, characterized in that: The dynamic optimization steps include: Step a. Data collection, collecting real-time load data; Step b. Load forecasting, using machine learning models to predict short-term and long-term load changes; Step c. Optimization calculation, using an optimization algorithm to calculate the optimal load distribution plan based on the prediction results; Step d. Adjust the power supply plan and dynamically adjust the load distribution of each substation based on the optimization calculation results.

Images