Dealing with electricity demand fluctuations throughout peak and off-peak periods is challenging for electricity companies. During peak demand times, the grid should be able to match the high consumer needs. Conversely, minimal usage during off-peak periods leads to underutilization of generation capacity. This imbalance challenges utilities to ensure sufficient capacity and devise fair pricing models. The Time-of-Use (ToU) pricing model has emerged as a viable solution in many countries, encouraging consumers to shift their energy consumption from expensive peak hours to more affordable off-peak periods. To this end, this paper proposes unsupervised machine learning methods for designing ToU tariffs using only energy consumption time series data. Additionally, a new metric is introduced to evaluate the adaptability of the ToU methods to fluctuations in energy consumption. To validate the implemented techniques, public datasets from different countries were used.
