Utilizing Optimization Approaches to Assess Total Network Losses in Türki̇ye

Abstract

The accurate forecasting of energy losses in the interconnected grids of Türkiye assumes crucial importance in facilitating strategic planning, optimizing resource allocation, and fostering infrastructure development within the energy sector. A limited number of studies in the existing literature address the estimation of network losses, with investigations into such losses have predominantly relied on conventional machine learning (ML) methodologies. This study proposes an arithmetic optimization algorithm (AOA) approach to the total network losses (TNL) estimation of Türkiye. Firstly, the models are generated with AOA, IAOA, and LevyAOA methods, and then long-term TNL projections were conducted for three scenarios between 2021 and 2050. TNL is modelled as a linear regression model, and for this model, import, export, population, and gross domestic product (GDP) indicators are used as input parameters, and the TNL indicator is used as the output parameter. In the experiments, the historical data records of Türkiye from 1979 to 2020 are used to create the estimation model. Then, long-term TNL estimations with different scenarios are realized for Türkiye up to 2050. The TNL are expressed in gigawatt-hours (GWh) on an annual basis. According to the experimental results and comparisons, the IAOA method has shown quality and robust performance for estimating the TNL compared to other methods. Compared to the standard AOA, the proposed IAOA reduced the total error by 42.57% and the total relative error by 55.49% on the 1979–2020 dataset.