Dynamic pricing for load shifting: Reducing electric vehicle charging impacts on the grid through machine learning-based demand response

•Forecasting the feeder-wise power consumption using machine learning techniques.•Feeder-wise forecasted electric power consumption-based day-ahead dynamic pricing.•The distribution substation feeder-wise demand response scheme is analysed.•The distribution substation network's peak-to-average ratio has minimized.•IoT devices manage price-sensitive loads during peak or high-price periods. A robust smart grid communication network is a critical technology that enables modernized utilities to change power usage in real-time for optimal supply and demand balance. The utility sector must have access to additional power during times of high demand or crises to fulfil the demands of the wholesale market. This paper proposes a dynamic-pricing technique to manage power fluctuations while considering peak and off-peak electricity consumption. The demand for different feeders, overall distribution networks, and end-user power rates decrease throughout the day's peak-hours using proposed dynamic-pricing scheme. Internet of Things (IoT) devices manage price-sensitive loads during peak periods. This article proposed the decision tree regression (DTR)-XGBoost models to analyze short-term electric power consumption forecasting in a dynamic environment. The highest overall distribution substation electric power consumption forecasting accuracy is achieved by DTR-XGBoost in the one-hour interval, with an RMSE of 0.2616 MW, MSE of 0.0684 MW, MAE of 0.1270 MW, and R2 of 0.9888. Using demand response to minimize peak demand caused by charging electric vehicles and other high-power devices in distribution networks. Results show that the proposed demand response day ahead dynamic pricing minimizes energy costs and enables smart substation operators to stabilize the power system.