ELECTRICITY THEFT DETECTION USING SELF-SUPERVISED PATTERN RECOGNITION ALGORITHM

Abstract

ABSTRACT “ELECTRICITY THEFT DETECTION USING SELF-SUPERVISED PATTERN RECOGNITION ALGORITHM” The present invention provides an electricity theft detection using self-supervised pattern recognition algorithm. The system employs a hybrid deep learning model combining Long Short-Term Memory (LSTM) networks and Autoencoders to process sequential time-series data from Zero Sequence signals in electrical transmission systems. This approach enables accurate detection of anomalies caused by illegal tapping of transmission lines without requiring labeled data for training. The hybrid model optimizes performance using Adaptive Momentum Optimization, achieving a high detection accuracy of up to 97.93% with minimal false positives. The method also reduces the need for extensive data preprocessing and supervision, providing a cost-effective solution for identifying non-technical losses and segregating illegal users in power grids. Figure 1

Specification

Description:TECHNICAL FIELD
[0001] The present invention relates to the field of electricity, and more particularly, the present invention relates to the electricity theft detection using self-supervised pattern recognition algorithm.
BACKGROUND ART
[0002] The following discussion of the background of the invention is intended to facilitate an understanding of the present invention. However, it should be appreciated that the discussion is not an acknowledgment or admission that any of the material referred to was published, known, or part of the common general knowledge in any jurisdiction as of the application's priority date. The details provided herein the background if belongs to any publication is taken only as a reference for describing the problems, in general terminologies or principles or both of science and technology in the associated prior art.
[0003] The electrification of households and villages in rural areas in developing countries like India has seen much of the pace in recent years. Owing to the rise in demand for electricity also necessitates the requirement of newer power plants to be set up to fulfill the demand. These power plants not only require huge investments but they also have a long term irreparable impact on the environment. The non-technical losses in electrical systems in the form of theft due to illegal tapping on transmission lines are rampant in rural India.
[0004] India incurs the loss to the tune of nearly $4.5 billion each year due to the Non-technical losses major portion of which include illegal theft of electricity due to tapping and tampering of energy meters. In rural India, it becomes even more difficult to trace the location of theft manually because the inaccessible terrains present challenges to man those areas. In this context, the critical practices that make use of the classification of power consumption patterns can be used for the extraction of non-technical losses. The sources of likely vulnerable points of attacks are recognized through the upswing in power consumption.
[0005] This work presents a Self- Supervised Pattern Recognition Approach using Hybrid Models of Stacked Long Short Term Memory Networks with Autoencoders for robust Real-time Anomaly Detection in Electrical Transmission Lines for Electricity Theft Detection. The previous approaches used Supervised Machine Learning Algorithms and other Statistical Approaches which failed to handle large volumes of data in real-time. The Hybrid Model of LSTM-Autoencoder is a Self-Supervised Algorithm hence; unlike Supervised Machine Learning Algorithms it doesn't require the involvement of labeled data for training purposes. The Self-Supervised Algorithms are capable of handling and dealing with real-time data input and cater to the purpose efficiently. This Hybrid model has the capability of preserving Sequential and Temporal Information because of the use of Long Short Term Memory Networks. The detailed and continuous changes and feature maps are preserved in the process and help to perform the task of Anomaly Detection with utmost accuracy. Unlike the conventional algorithms which persistently suffer from the problem of False Anomaly Detection. Moreover, the algorithm is robust enough to be used in accordance with the requirements of the industry, i.e., the working of the algorithm is not affected by the change in the process of preparation of data or feeding of data into the system. Thus, the system performs better than all other available approaches in handling similar situations and problems. Taking into consideration the application of the algorithm in industrial applications, the most important aspect to be considered for handling the real-time situations is the Time Complexity of the Hybrid Model which is very less and the Computation Power required is also less compared to other existing models.
[0006] Rouzbeh Razavi et al. presented the Finite Mixture Model clustering for customer segregation and a Genetic Programming algorithm for spotting new features suitable for prediction. The study investigates some important practical aspects to establish theft detection including the detection delay and the required size of historical demand data. The model presents insights of accuracy in detecting thefts of various types and intensity; detecting irregular and unseen attacks along with the computational complexity of the algorithm. Tanveer Ahmad et al. presented a review of different modeling techniques for pinpointing and avoidance of non-technical losses. The data mining based models, support vector machine model, and Optimum-path forest clustering process were discussed to find legal and abnormal profiles of the industry as well as commercial customers to find out the theft of electricity. The support vector machine with a genetic algorithm extends a hybrid method for the NTL investigation and provides automated assistance to dominate electricity theft.
[0007] These techniques require human resources and huge investment for setting up labs due to their assisted nature, making them a costly affair. The encoding process is required to streamline and reshape customer energy consumption data for easier analysis without arbitrating the quality or singularity of the data. The study of parallelized algorithms has shown comprehensive results in this context. Time series forecasting (TSF) is a method of speculating upcoming values for a given sequence using recorded data. Recently, it has attracted a lot of attention from researchers in the area of machine learning to avoid the impediment of traditional forecasting techniques. Which are time taking and full of intricacies often required human assistance for data preparations. The deep long-short term memory (DLSTM) architecture is an augmentation of the conventional recurrent neural network. A genetic algorithm is used in order to appropriately configure DLSTM optimum architecture. The empirical results showcase that the DLSTM model outperforms other standard approaches.
[0008] Multivariate time series classification models are proving to be a popular approach over the Univariate counterpart. The Long Short Term Memory Fully Convolutional Network (LSTM-FCN) and Attention LSTM-FCN (ALSTM-FCN), introduced into a multivariate time series classification model by augmenting the fully convolutional block. With a squeeze-and excitation block to further improve accuracy. These models are highly efficient due to lesser test time and are small enough to be deployed on memory-constrained systems.
[0009] In light of the foregoing, there is a need for an electricity theft detection using self-supervised pattern recognition algorithm that overcomes problems prevalent in the prior art associated with the traditionally available method or system, of the above-mentioned inventions that can be used with the presented disclosed technique with or without modification.
[0010] All publications herein are incorporated by reference to the same extent as if each individual publication or patent application were specifically and individually indicated to be incorporated by reference. Where a definition or use of a term in an incorporated reference is inconsistent or contrary to the definition of that term provided herein, the definition of that term provided herein applies, and the definition of that term in the reference does not apply.
OBJECTS OF THE INVENTION
[0011] The principal object of the present invention is to overcome the disadvantages of the prior art by providing Electricity theft detection using self-supervised pattern recognition algorithm.
[0012] Another object of the present invention is to provide an electricity theft detection using self-supervised pattern recognition algorithm that minimizes the occurrence of false positives in anomaly detection, ensuring reliable and precise identification of theft-related patterns.
[0013] Another object of the present invention is to provide an electricity theft detection using self-supervised pattern recognition algorithm that utilizes a hybrid model combining Long Short-Term Memory (LSTM) networks and Autoencoders to process sequential time-series data, such as Zero Sequence signals, without the need for labeled data during training.
[0014] Another object of the present invention is to provide Electricity theft detection using self-supervised pattern recognition algorithm that reduces the time, cost, and human effort required for data preprocessing and manual supervision of the detection process.
[0015] Another object of the present invention is to provide Electricity theft detection using self-supervised pattern recognition algorithm that provides a cost-effective and computationally efficient solution for electricity theft detection, achieving high accuracy (up to 97.93%) and fast processing times.
[0016] Another object of the present invention is to provide Electricity theft detection using self-supervised pattern recognition algorithm that optimizes performance using advanced algorithms like Adaptive Momentum Optimization to enhance system accuracy and reliability
[0017] The foregoing and other objects of the present invention will become readily apparent upon further review of the following detailed description of the embodiments as illustrated in the accompanying drawings.
SUMMARY OF THE INVENTION
[0018] The present invention relates to Electricity theft detection using self-supervised pattern recognition algorithm.
[0019] This work focuses on Self-Supervised Approaches which not only saves time but also helps in reducing expenditure on manpower required for supervision of events and data preparations. Since the real-time data often doesn't offer labeled data, the Self-Supervised Pattern Recognition Algorithm has been proposed. The algorithm is represented in the form of a hybrid model comprised of Stacked Long-Short-Term-Memory Networks along with Autoencoder Network. which appropriately identifies several different patterns of upsurges caused by illegal tapping of transmission lines in electrical systems. Unlike conventional Supervised Machine Learning Algorithms, The Hybrid Model of LSTM-Autoencoder doesn't need the association of labeled data for training purposes. It helps us to perform the task of Anomaly Detection with utmost accuracy and low computational complexity, unlike the conventional algorithms which persistently suffer from the problem of False Anomaly Detection.
[0020] The proposed hybrid model consists of a Long Short Term Memory Network & an Autoencoder Architecture working combinedly to get the desired output as shown in Fig.1. Since Long Short Term Memory Networks are efficient in dealing with Sequential Time Series data. Extracting fine details and temporal features from the sequential data of the Zero Sequence ( Electrical Transmission System based feature ) Signals, so LSTM is used for getting the Original Zero Sequence Signal Data as Input Layer. The Autoencoders are used in combination with Stacked LSTM Architecture for encoding the temporal features in a latent space vector with the help of the encoding part. Further, with the help of the decoder part, the time-series signal data is reconstructed to get the output.
[0021] While the invention has been described and shown with reference to the preferred embodiment, it will be apparent that variations might be possible that would fall within the scope of the present invention.
BRIEF DESCRIPTION OF DRAWINGS
[0022] So that the manner in which the above-recited features of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may have been referred by embodiments, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.
[0023] These and other features, benefits, and advantages of the present invention will become apparent by reference to the following text figure, with like reference numbers referring to like structures across the views, wherein:
[0024] Fig. 1. Proposed Hybrid Model Architecture.
DETAILED DESCRIPTION OF THE INVENTION
[0025] While the present invention is described herein by way of example using embodiments and illustrative drawings, those skilled in the art will recognize that the invention is not limited to the embodiments of drawing or drawings described and are not intended to represent the scale of the various components. Further, some components that may form a part of the invention may not be illustrated in certain figures, for ease of illustration, and such omissions do not limit the embodiments outlined in any way. It should be understood that the drawings and the detailed description thereto are not intended to limit the invention to the particular form disclosed, but on the contrary, the invention is to cover all modifications, equivalents, and alternatives falling within the scope of the present invention as defined by the appended claim.
[0026] As used throughout this description, the word "may" is used in a permissive sense (i.e. meaning having the potential to), rather than the mandatory sense, (i.e. meaning must). Further, the words "a" or "an" mean "at least one" and the word "plurality" means "one or more" unless otherwise mentioned. Furthermore, the terminology and phraseology used herein are solely used for descriptive purposes and should not be construed as limiting in scope. Language such as "including," "comprising," "having," "containing," or "involving," and variations thereof, is intended to be broad and encompass the subject matter listed thereafter, equivalents, and additional subject matter not recited, and is not intended to exclude other additives, components, integers, or steps. Likewise, the term "comprising" is considered synonymous with the terms "including" or "containing" for applicable legal purposes. Any discussion of documents, acts, materials, devices, articles, and the like are included in the specification solely for the purpose of providing a context for the present invention. It is not suggested or represented that any or all these matters form part of the prior art base or were common general knowledge in the field relevant to the present invention.
[0027] In this disclosure, whenever a composition or an element or a group of elements is preceded with the transitional phrase "comprising", it is understood that we also contemplate the same composition, element, or group of elements with transitional phrases "consisting of", "consisting", "selected from the group of consisting of, "including", or "is" preceding the recitation of the composition, element or group of elements and vice versa.
[0028] The present invention is described hereinafter by various embodiments with reference to the accompanying drawing, wherein reference numerals used in the accompanying drawing correspond to the like elements throughout the description. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiment set forth herein. Rather, the embodiment is provided so that this disclosure will be thorough and complete and will fully convey the scope of the invention to those skilled in the art. In the following detailed description, numeric values and ranges are provided for various aspects of the implementations described. These values and ranges are to be treated as examples only and are not intended to limit the scope of the claims. In addition, several materials are identified as suitable for various facets of the implementations. These materials are to be treated as exemplary and are not intended to limit the scope of the invention.
[0029] The present invention relates to Electricity theft detection using self-supervised pattern recognition algorithm.
[0030] Various Machine Learning and Deep Learning Algorithms for identifying and detecting electrical thefts via illegal tapping onto transmission lines. The data mining based models have been at the forefront and can detect an uptick in power consumption. The Support Vector Machine Algorithm, as well as the Random Forest Algorithm, which segregates illegal customers, is a type of conventional mixed machine learning model. These techniques are used to modulate the parts of non-technical losses (NTL) through the radial distribution method. These techniques require an investment of a large sum of money, time, and energy for conducting experimental works related to data pre-processing. This work focuses on Self-Supervised Approaches which not only saves time but also helps in reducing expenditure on manpower required for supervision of events and data preparations. Since the real-time data often doesn't offer labeled data, the Self-Supervised Pattern Recognition Algorithm has been proposed.
[0031] Accuracy is used as a performance metric in order to analyze the performance of the Self-supervised Hybrid Model in the process of Anomaly Detection of the Zero Sequence Signals. A comparative analysis report of accuracies achieved with the help of various Network Optimization Technique is shown in Table. 1.
[0032] From table-1, it is inferred that the Adaptive Momentum Optimization Algorithm (Adam) which is actually a combination of Stochastic Gradient Descent with Momentum (SGDM) and RMSProp Algorithms helps in the best optimization of weights and parameters in order to achieve an accuracy of 96.93 %, which shows the robust nature of the Hybrid Model in the process of Anomaly Detection in Zero Sequence Signals.
Optimization Technique or Algorithm Accuracy Score
Adaptive Momentum 96.93 %
Adaptive Gradient 95.23 %
Adaptive Delta 91.45 %
RMSProp 91.34 %
Stochastic Gradient Descent with Momentum 85.67 %
[0033] Conventional and native Techniques have failed for a long time to robustly and efficiently handle the process of Anomaly Detection of sequential time series based Electrical Transmission Signals at the forefront. The conventional algorithms are unable to deal with the persistent problem of False Anomaly Detection of the Signals, which is a matter of grave concern in the fields of Smart Grid. Therefore, the application of effective and robust Hybrid Deep Learning Models like Long Short Term Memory Networks with Autoencoders (Time-stamp: Minute Wise) was proposed in the fields of Electrical Theft Detection in the Electrical transmission and distribution systems. It promises accurate detection of anomalies (Reconstruction Loss is only 0.30) without the problem of false detection. The computational time taken by the Hybrid LSTM-Autoencoder model is only 240 ms. Besides, the model is capable of handling noise associated with the time series based Zero Sequence signals and the accuracy achieved is 97.93 %.
[0034] Machine Learning and Deep Learning Algorithms for identifying and detecting electrical thefts via illegal tapping onto transmission lines. The data mining based models have been at the forefront and can detect an uptick in power consumption. The Support Vector Machine Algorithm, as well as the Random Forest Algorithm, which segregates illegal customers, is a type of conventional mixed machine learning model. These techniques are used to modulate the parts of non-technical losses (NTL) through the radial distribution method. These techniques require an investment of a large sum of money, time, and energy for conducting experimental works related to data pre-processing. This work focuses on Self-Supervised Approaches which not only saves time but also helps in reducing expenditure on manpower required for supervision of events and data preparations. Since the real-time data often doesn't offer labeled data, the Self-Supervised Pattern Recognition Algorithm has been proposed. The algorithm is represented in the form of a hybrid model comprised of Stacked Long-Short-Term-Memory Networks along with Autoencoder Network. which appropriately identifies several different patterns of upsurges caused by illegal tapping of transmission lines in electrical systems. Unlike conventional Supervised Machine Learning Algorithms, The Hybrid Model of LSTM-Autoencoder doesn't need the association of labeled data for training purposes. It helps us to perform the task of Anomaly Detection with utmost accuracy and low computational complexity, unlike the conventional algorithms which persistently suffer from the problem of False Anomaly Detection.
[0035] The maximum Reconstruction Error Loss obtained in the process is 0.30 which is minimum compared to other models. Moreover, the time taken by the algorithm for the execution of the entire process is 240 ms.
[0036] Various Machine Learning and Deep Learning Algorithms for identifying and detecting electrical thefts via illegal tapping onto transmission lines. The data mining based models have been at the forefront and can detect an uptick in power consumption. The Support Vector Machine Algorithm, as well as the Random Forest Algorithm, which segregates illegal customers, is a type of conventional mixed machine learning model. These techniques are used to modulate the parts of non-technical losses (NTL) through the radial distribution method. These techniques require an investment of a large sum of money, time, and energy for conducting experimental works related to data pre-processing. This work focuses on Self-Supervised Approaches which not only saves time but also helps in reducing expenditure on manpower required for supervision of events and data preparations. Since the real-time data often doesn't offer labeled data, the Self-Supervised Pattern Recognition Algorithm has been proposed.
[0037] Various modifications to these embodiments are apparent to those skilled in the art from the description and the accompanying drawings. The principles associated with the various embodiments described herein may be applied to other embodiments. Therefore, the description is not intended to be limited to the 5 embodiments shown along with the accompanying drawings but is to be providing the broadest scope consistent with the principles and the novel and inventive features disclosed or suggested herein. Accordingly, the invention is anticipated to hold on to all other such alternatives, modifications, and variations that fall within the scope of the present invention and appended claims. , Claims:CLAIMS
We Claim:
1) A system for detecting electricity theft, the system comprising:
- a self-supervised pattern recognition algorithm for identifying electricity theft via illegal tapping of transmission lines, using a hybrid model consisting of Long Short-Term Memory (LSTM) networks and Autoencoders;
- wherein the LSTM network processes sequential time-series data from Zero Sequence signals in an electrical transmission system, and the Autoencoder encodes and reconstructs the temporal features of the data.
2) The system as claimed in claim 1, wherein the self-supervised algorithm does not require labeled data for training and detects anomalies with high accuracy in real-time data.
3) The system as claimed in claim 1, wherein the hybrid LSTM-Autoencoder model achieves an accuracy of at least 97.93% and reduces false anomaly detection.
4) The system as claimed in claim 1, wherein the model uses the Adaptive Momentum Optimization algorithm to achieve a reconstruction error loss as low as 0.30.
5) A method for electricity theft detection, the method comprising:
- collecting real-time Zero Sequence signals from electrical transmission systems;
- applying a hybrid deep learning model consisting of Stacked LSTM networks and autoencoders to process the sequential time-series data;
- detecting anomalies related to illegal power tapping by analyzing the upsurges in power consumption patterns.

Documents