This paper presents an artificial neural network (ANN) approach to electric load forecasting, aiming to improve the accuracy of load predictions. The ANN is trained to learn the relationship between past, current, and future temperatures and loads. The method is tested on actual utility data, showing average absolute errors of 1.40% for one-hour forecasts and 2.06% for 24-hour forecasts, which are significantly better than the 4.22% error of a currently used forecasting technique. The paper reviews traditional time series and regression approaches, highlighting their limitations such as inability to handle abrupt changes and numerical instabilities. The proposed ANN combines elements of both time series and regression, using a layered perceptron model. The generalized Delta rule is used for training the ANN, and the performance is evaluated using various test cases, including peak load, total load, and hourly load forecasting. The results show that the ANN can effectively predict load patterns, with lower errors compared to conventional methods. The authors also discuss the potential for integrating additional weather variables and the need for further research to address specific challenges in load forecasting.This paper presents an artificial neural network (ANN) approach to electric load forecasting, aiming to improve the accuracy of load predictions. The ANN is trained to learn the relationship between past, current, and future temperatures and loads. The method is tested on actual utility data, showing average absolute errors of 1.40% for one-hour forecasts and 2.06% for 24-hour forecasts, which are significantly better than the 4.22% error of a currently used forecasting technique. The paper reviews traditional time series and regression approaches, highlighting their limitations such as inability to handle abrupt changes and numerical instabilities. The proposed ANN combines elements of both time series and regression, using a layered perceptron model. The generalized Delta rule is used for training the ANN, and the performance is evaluated using various test cases, including peak load, total load, and hourly load forecasting. The results show that the ANN can effectively predict load patterns, with lower errors compared to conventional methods. The authors also discuss the potential for integrating additional weather variables and the need for further research to address specific challenges in load forecasting.