Electricity demand management is most crucial part for Nepal Electricity Authority (NEA) in case of Nepal. Electricity demand is mainly affected by time of day, week of day, monthly season and ambient temperature of power substation customer, their population trained and their life style according to time period. Mostly in winter most of electricity consumer use heating appliances and in summer cooling appliances. Also, in daily pattern electricity demand is high during morning and evening time due to peoples used cooking, lighting and entertainment appliances. That varies the load on power grid. By leveraging historical electricity demand data, time and meteorological records, we have to identify correlations and seasonal and time series patterns using statistical and machine learning techniques and predict short time electricity demand on power substation (hourly). In this system we use multivariable LSTM based-short term electricity demand forecasting, which can predict with 99% maximum accuracy.