Prevention, detection and penalization of electricity thefts in smart utility networks by rule-based and long short-term memory techniques

Simple item page
dc.contributor.authorOtuoze, AbdulRahaman Okino
dc.date.accessioned2023-12-11T05:05:58Z
dc.date.available2023-12-11T05:05:58Z
dc.date.issued2020
dc.descriptionThesis (PhD. (Electrical Engineering))
dc.description.abstractThe successful deployment of Smart Grids (SG) clearly hinges on energy efficiency relying majorly on the operations of the Advanced Metering Infrastructure (AMI) with Smart Electricity Meters (SEM) as its key aspect. Like every Cyber-Physical System (CPS), it is threatened by cyber-attacks and electricity theft is a notable motive of these attacks. Nonetheless, SEM offer adequate data being leveraged upon for analytical inferences. However, various research efforts mainly utilising artificial intelligence and machine learning are aimed at generating suspicious customers’ lists rather than a holistic approach to curbing the various aspects of the menace. In this thesis, a proactive scheme for preventing, detecting and penalizing electricity thefts is proposed. To achieve the prevention phase, a cyber security layer based on a novel Monkey-Banana Deceptive Algorithm (MBDA) for intrusion detection is introduced. This algorithm is developed from the popular 5 or 8-monkey theory by first presenting each of the stages to scenarios and then formulated to a probability assignment model. MBDA probability assignment is then applied to develop the algorithm for detecting intrusion in SEM’s communication gateway. To strengthen the prevention phase, selected factors indicative of electricity thefts is then modelled by defining a set of rules to infer security risk level using Fuzzy Inference System (FIS). The detection phase utilises a Long Short-Term Memory (LSTM) network based on time series prediction of the energy consumption data. The forecast values of the energy consumption are compared with the observed values to detect suspicious consumers based on defined anomaly detection model. To confirm true fraudulent consumers, a confirmation model is introduced based on selected monitoring parameters using FIS model. In the penalization phase, a cost estimation-based model by an analytical approach is introduced to deduce the penalty fine on confirmed fraudulent customers with considerations to energy consumed during reported theft period and modifications of some existent Electricity Theft Acts. A self-generated attack was used to implement the MBDA while the results of the FIS model determines the prevention status. The detection phase was implemented using the SEM energy consumption data of four selected consumers of different profiles to build consumer-dependent LSTM models. The anomaly and confirmation models are used to justify true fraudulent customers based on the states of the monitored parameters. The results of the cost estimation-based model implemented on twenty randomly selected electricity fraudulent consumers for the penalization phase indicate fraudulent customers reported at second and third attempts incurring 42% and 60% increase in the imposed fines, respectively. Implementation of this proactive scheme will enhance real-time protection of the SEM, reduces over reliance on energy consumption data analytics, reduces false positive rates, eliminates the usual practice of bogus financial sanctions, drastically reduce the need for the complicated on-site customer-to-customer inspections thereby saving manpower, stress, cost and time. In addition, the penalization phase also helps shift electricity theft burden from honest consumers. This proposed scheme is a suitable deployment for electricity theft prevention, detection and penalization in a smart utility network.
dc.description.sponsorshipFaculty of Engineering - School of Electrical Engineering
dc.identifier.urihttp://openscience.utm.my/handle/123456789/877
dc.language.isoen
dc.publisherUniversiti Teknologi Malaysia
dc.subjectSmart power grids—Security measures
dc.subjectElectricity—Problems, exercises, etc
dc.subjectTheft—Prevention
dc.titlePrevention, detection and penalization of electricity thefts in smart utility networks by rule-based and long short-term memory techniques
dc.typeThesis
dc.typeDataset
Files
Original bundle
Now showing 1 - 2 of 2
Thumbnail Image
Name:
AbdulRahamanOkinoPSKE2020_A.pdf
Size:
311.83 KB
Format:
Adobe Portable Document Format
Description:
Developed MBDA Code for Implementing MBDA
Download
Thumbnail Image
Name:
AbdulRahamanOkinoPSKE2020_B.pdf
Size:
714.28 KB
Format:
Adobe Portable Document Format
Description:
Developed codes for the LSTM Deep learning forecasting and Anomaly Detection
Download
License bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed to upon submission
Description:
Download
Collections
Physics, Engineering and Material Science