Load forecasting predicts future energy demand using historical data and algorithms, helping utilities plan generation and distribution to ensure reliable power supply and optimize operations. Accurate load forecasting is vital for efficient energy management, resource allocation, and grid stability. This research explores advanced forecasting methodologies to predict electricity demand, focusing on the integration of machine learning techniques and optimization strategies. This study introduces a 24-h ahead load forecasting approach employing the CatBoost algorithm, specifically tailored to handle categorical variables and missing data. In this research study, we are developing a hybrid forecasting model which combines the CatBoost Regressor and Wavelet Transform so that accuracy is enhanced to predict 24-hour ahead loads in the electric sector. This model will use wavelet-based feature extraction and noise reduction to increase model accuracy. The Mean Absolute Percentage Error (MAPE) shows that the newly developed WT-CatBoost model outperforms existing models with a MAPE of 14.04% whereas the existing CatBoost and XGBRegressor models show a 17.98% and 18.85%, respectively. Historical load and weather data is used for the model training and evaluation, which shows better accuracy when compared to the linear regression and support vector regression models. This newly proposed WT-CatBoost hybrid model provides a better framework for better load forecasting which will aid in effective power management and prediction.