Replacing fossil fuels with renewable energy is one must-do work for policymakers and scientific communities to give a sustainable environment for future generations. Solar power is one of the significant contributors to the renewable energy sector and must injected into the central energy grid to use it efficiently. Advanced solar power forecasting of a specific plant is essential when we transmit the generated energy to the grid. Photovoltaic (PV) production is very weather-dependent and volatile to operate the plant appropriately we need a forecasting system. In this work, we introduce a forecasting model for solar power (PV) plants using the concept of machine learning. More specifically, we develop a lightweight ensemble model using Bi-directional Long Short-Term Memory (Bi-LSTM), which is able to produce very accurate predictions compared to other available state-of-the-art models. This study curates a dataset that combines information from the Indian Institute of Engineering, Science and Technology (IIEST), Shibpur, on power generation with data from the Central Pollution Control Board of India (CPCBI). The outcomes are contrasted with those of a few other standalone deep learning models, such as Recurrent Neural Networks(RNNs), Support Vector Regressor (SVR), and Long Short-Term Memory (LSTM) proposed model outperforms others.