This paper presents a sequential disaggregation of residential energy consumption using a Random Forest-based regression model. The primary aim is to address the challenge of accurately disaggregating individual appliance-level consumption from aggregate consumption data. The proposed method leverages a multi-step load disaggregation process in which each appliance’s consumption is sequentially predicted and subtracted from the aggregate data, reducing noise and improving the accuracy of subsequent appliance predictions. The training data was supplemented with additional engineered features such as the outside temperature, temperature difference, time-based features, and holiday data. Individual models were developed for five key appliances, including a fan coil unit (FCU), washing machine, water heater, dishwasher and refrigerator using real-world data from two residential households in Dubai, UAE. The proposed sequential energy disaggregation algorithm demonstrated high predictive performance, achieving an average MAE of 3.79 W, with R2 scores of 0.92 for the two households. The method’s robustness was confirmed through results showing an average estimation accuracy of 98% across all appliances. Our results suggest that this sequential disaggregation method is a reliable and scalable solution for non-intrusive load monitoring (NILM) of residential energy consumption, providing detailed insights into appliance-level consumption.