Abstract This paper presents the first statistical modeling of peak times in maximum daily energy consumption for major industrial and service-sector companies in Tanzania, representing the first such analysis for any African country. Using high-resolution automated meter reading data from six of Tanzania’s largest energy consumers collected between 2020 and 2023, we model the time of peak demand as a circular variable. Due to the observed multimodality in peak times, a finite mixture of von Mises distributions is applied, with model parameters estimated via Markov chain Monte Carlo. The optimal number of mixture components varies considerably across companies-ranging from 4 to 10-reflecting diverse and company-specific demand patterns. Key circular statistics, including mean resultant, variance, skewness, and kurtosis, reveal substantial differences in peak concentration and temporal distribution: industrial firms exhibit more predictable, clustered peaks, while utility companies show highly variable and dispersed demand. Likelihood ratio tests confirm that peak-time distributions are stable across years, months, and days, indicating structurally embedded operational rhythms.
Nadarajah et al. (Tue,) studied this question.