The research in the present work deals with modelling of electricity consumption by machines in manufacturing industry. Artificial Neural Networks (ANNs) such as NNARX, NNARMAX and NNOE nonlinear neural network was utilized to predict electricity based on real data collected. The measured data related to electricity consumed and production rate were collected for a period that ranges between three months up to six months. The Value Stream Line (VLS) under examination is composed by four machines that performs works such as turning, milling, robot and lapping. Present research is focused on one Machine (that perform turning operation) and the analysis was conducted following a methodology developed previously by the present researchers. The model developed were validated utilizing four criteria such as Mean Squared Error (MSE), Mean Absolute Error (MAE), Goodness of Fit (G) and Coefficient of Determination (r2). Analysis was conducted for different step ahead prediction that ranges between 1 h up to 4 h. Selection of NNARX, NNARMAX and NNOE models that better predict electricity consumption is based on the four criteria of validations for different step ahead prediction. Results analysis demonstrated that NNARX [3 2 1], NNARMAX [2 2 2 1 and NNOE [3 2 1] are the selected models which present good prediction capabilities related to electricity consumption from one hour to up four hours. Finally, these models can be combined with programmer logic controllers (PLCs) which will be the focus of the next research on machine’s fault detection and diagnose.