Basically the active power demands at various load buses need to be estimated ahead of time in order to plan the generation and distribution schedules, for contingency analysis and also for monitoring the system security. A great deal of attention has been given in recent years to the question of setting up the demand models for the individual appliances and their impact on the aggregated demand. For the allocation of spinning reserve it would be necessary to predict the load demands at least an hour ahead. A method using ANN based technique is developed for short-term load forecast. The technique is tested on real time data collected from a 220 KV / 132 KV / 33 KV / 11 KV Renigunta Sub-Station, A.P, India. Calculations are done based on the hourly data of active power variations obtained over a period of one month. The active powers were used as input quantities for training the ANN and obtained the respective output active powers for the corresponding day.