It is important for Bangladesh to balance energy development and sustainable growth in light of the country’s need for net-zero emissions. In this study, we develop three hybrid models to forecast the growth of renewable energy and the level of CO2 emissions for target dates set by the country. This paper presents a prediction analysis of CO2 emissions and the renewable energy industry in terms of the necessary rate of mandated emissions based in probabilistic deep learning. We designed the first model, which combines Long Short-Term Memory (LSTM) with autoregressive integrated moving average (ARIMA) and XGBoost. The second hybrid model consists of deep neural networks (DNNs) and Gaussian process regression (GPR). The third model is a blend of random forest and LSTM with a Bayesian neural network (BNN). By exhaustive analysis of the results, we examine the values obtained from the generation of predictive models for the renewable energy sector along with progressive enhancement of energy production and decarbonization. The results demonstrate the promise of the hybrid methodologies for the improvement of energy efficiency and energy policy systems with the objective o net-zero in Bangladesh. The outcome of the prediction method reveals that hybrid model of deep neural networks (DNNs) with a Bayesian neural network (BNN) achieve the best accuracy with the highest performance parameters.