Properly forecasting the electrical load helps in better electricity system management and more savings. In this work, an improved neural network based is proposed which uses electrical load forecasting system that utilizes historical load data, data on weather and similar factors to help predict the future energy used in a region. The proposed system is trained on a large dataset using also having weather parameters, and scheduling cost is also included. These features are often neglected in most of the research works, however having large effect of the models performance. Moreover, various deep learning algorithms are also evaluated on the dataset for all inclusive assessment. It has been demonstrated that the proposed system improves the accuracy of electrical load forecasting much more than typical methods can. The developed system could make power systems run more efficiently and aid in making improved energy sector decisions.