Abstarct: Maintaining the balance between supply and demand for electrical energy is important to increase the welfare of society. The excess of electricity consumption (demand) over its production (supply) in a day/hour causes blackouts and the resulting damage. Electrical energy cannot be stored in large quantities and its excessive production/distribution in the grid imposes economic and environmental costs on the country. Predicting peak electricity consumption and identifying its determinants are helpful in managing and maintaining the balance between electricity supply and demand. Therefore, in this study, the peak load of electricity consumption in Iran was modeled and forecast using different approaches for the period from March 20, 2017 to June 6, 2022. 90% of the 1872 observations were used to build the model and the remaining 10% of the observations were used to predict the peak load. In peak load modeling using the neural network (ANN) approach, the number of optimal neurons baxsed on the minimum prediction error criterion was estimated to be 11 in the first laxyer and 8 in the second laxyer, respectively. The results of the descxriptive statistics of the study show that the frequency of the first consumption peak at 11:00 a.m. and the frequency of the second consumption peak at 9:00 p.m. are greater than in the other hours. The results of peak load modeling using the linear regression model (GLM) approach show that the variables public holidays, hours of energy consumption, the number of new corona patients, the country's average air temperature (the average of the two cities of Mashhad and Shiraz) and its differences in different parts of the country, the relative humidity of the weather and its variation in different locations and populations have a significant impact on peak power consumption. But the effects of the day of the week variable, the cryptocurrency mining index (Bitcoin) and per capita income are not significant. In addition, holidays and air temperature have a greater influence than other variables. A comparison of the peak power consumption prediction results shows that the prediction accuracy of different models and approaches is not the same. The average percentage of prediction error of the GLM, ANN, and ARIMA models over 187 days (12/04/2021 to 06/06/2022) are 0.0799, 0.0754, and 0.0714, respectively. Therefore, the ARIMA model with minimum average prediction error is better suited for peak load forecasting.