Electric power requirement of household and industries depends on geographical location as climate parameters directly impact on the power requirements. Balancing of power in industries need some power forecasting algorithms to learn pattern of load balancing and calculate the power requirement. This paper has resolved the issue of electric power requirement estimation for big infrastructures, like Industries, company, plants, and so on. This paper proposes a Fish Schooling Genetic algorithm-based neural network for climate feature ratio estimation as per geographical location. Once feature ration for a geographical location was identified, the power is forecasted. Experiment has been done on real dataset of Industries in Ahmednagar, Maharastra, India and the comparison of proposed model was done with the existing algorithms to reveal the effectiveness of the proposed strategy. The result shows that the model thus proposed has improved various evaluation parameter values for different season of the whole year. The proposed Fish Schooling Genetic algorithm-based neural network method achieved an Average Quarterly required power (AQPR) of 7430.333 using MIDC dataset and 6828.667 using Ahmednagar dataset