Nowadays, energy consumption prediction has become crucial to address rising energy demand, guide energy investment policies, utilize resources efficiently, and facilitate energy planning. Energy consumption is primarily in the form of electricity. Electricity consumption predictions are made using historical data with several techniques, including artificial intelligence methods, machine learning algorithms, and classical regression and time-series methods. In time series methods, the seasonal effect is a constant, such as monthly, yearly, or quarterly. But the seasons do not have to be in the same period. It is important to capture values that exhibit the same behavior and to determine the transition situation among seasons. So, clustering algorithms can capture these values. In this study, seasons are determined using the Fuzzy C-means clustering method, and transitions are calculated using a Markov transition matrix. In addition, because expressing a country's electricity consumption in crisp numbers is not inherently true, fuzzy numbers are used to classify values. More accurate forecasts can be achieved by using comprehensive clusters and distribution numbers to summarize the reported data within consumption data. This study considered Türkiye's monthly electricity consumption between 2019 and 2024. The MAPE value is used to compare similar studies in the literature. In conclusion, this study presents a relevant and practical approach to estimating electricity consumption using fuzzy C-means clustering and Markov chain-based time series analysis.

Keywords

Electricity consumption, Fuzzy C-means classification, Markov matrix, Time series analysis, Seasonal effect.