Electrical load forecasting is one of the important issues in maintenance and planning of the electrical load distribution. There are many techniques employed for electrical load forecasting in which Artificial Neural Networks are the very efficient and easy technique. Neural networks are trained with the previous datasets of inputs and outputs. Researchers have invented many algorithms to train the neural network for giving the best results. Since, electrical load forecasting is a very important task which should be done very accurately. Therefore, it is very necessary to decide the best training algorithm which will give the best result among all the training algorithms. In this paper, many training algorithms are used to train the network but only 3 overall best training algorithm results are compared with each other. These training algorithms are Conjugate Gradient, Bayesian Regulation and Resilient Backpropagation algorithm. Results are in terms of error, speed of training and accuracy of prediction of electrical which are compared for each training algorithm. In this, we have found that Resilient Backpropagation algorithm is the overall best training algorithm providing best results among all training algorithms. © Research India Publications.