A Box–Jenkins-Based Feedforward Control Approach for Industrial Power Factor Management Using Multiple Electrical Measurements

Power factor (PF) fluctuations in industrial equipment can affect power-system performance and create exposure to tariff surcharges. This study applies Box–Jenkins transfer-function modelling to multiple electrical measurements obtained from a robotic welding machine in a Malaysian factory. Measured industrial time-series data were used in numerical control experiments and simulation; calculated current-control commands were not physically implemented on the machine. Individual factor-specific PF predictions were combined using two weighting schemes: minimisation of historical fitting errors (Method A) and minimisation of one-step-ahead prediction errors (Method B). In the reported rolling-window comparisons, Method B produced fewer adverse negative deviations than Method A across all examined thresholds for multivariable feedforward control. Adding ARMA-based feedback did not further reduce those downside frequencies in the reported comparisons. The findings show the value of prediction-oriented weighting in a Box–Jenkins-based numerical control evaluation under the investigated operating conditions.