OPTIMIZING MULTI OBJECTIVE INSPECTION ROUTES WITH NON-DOMINATED SORTING GENETIC ALGORITHM II

• Main Author: Ayu Idiara, Ike
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/87104
• Institution: Institut Teknologi Bandung

Electricity theft, leading to non-technical losses, is a major challenge faced by electric utilities, particularly in enhancing the effectiveness of electricity theft inspection. In optimizing inspection routes, an approach is needed that not only considers efficient travel distances but also the potential revenue from penalties. Previous research has shown that the Self-Organizing Map (SOM) method, adapted for the Traveling Salesman Problem (TSP), can maximize economic revenue from electricity theft penalties at a low cost. However, this approach has the drawback of ignoring travel distances and only focusing on a subset of customers with the highest potential for penalties. This research aims to optimize inspection routes to minimize travel distance, maximize estimated penalty revenue, and maximize the number of inspections per route simultaneously using the Non-dominated Sorting Genetic Algorithm II (NSGA-II). Modifications are made to the initial population formation and crossover operator to meet the minimum number of inspections constraint. Electricity theft inspection optimization in this research is categorized as a Vehicle Routing Problem (VRP) with specific characteristics, namely VRP with time windows (daily working time limits), VRP with minimum capacity (minimum number of customers must be inspected), and Selective VRP (customer selection based on inspection priority) as well as multi-objective VRP (VRP that optimizes more than one objective simultaneously). One of the main challenges for PLN is to ensure that electricity theft inspections are carried out effectively to achieve the minimum inspection target. Route optimization is needed to reduce random inspections, which are often ineffective and only aim to meet the target number of inspections. The results of the study show that the modified NSGA-II approach is superior in maximizing penalty revenue per inspection and consistently produces valid solutions that meet minimum inspection constraints, especially in densely populated areas with short distances.