Electricity Load forecasting has consistently been the fundamental part of expanding productivity and income of intensity framework planning and activity such as generation and distribution of the organizations. However, since the connection between load power and factors affecting power control is nonlinear and dynamic uncertainties, it is difficult to distinguish its nonlinearity by utilizing conventional strategies. This paper proposes a multilayered/deep neural network-based predictive models, that forecast yearly and season wise monthly loads. Mean absolute percentage error (MAPE) is used to yield models' accuracy. A Deep-Feed forward Neural Network (Deep-FNN) with a sigmoid transfer function, resilient backpropagation training algorithm and Deep-FNN with Rectified Linear Unit (ReLu) activation function, Levenberg-Marquardt training algorithm for short term and long term, respectively, exhibit accurate and robust forecast compared to other forecasting models. The performance of the models are compared with the state-of-the-art methods for the peak loads depending on the weather conditions.