AI is Predicting Solar Radiation

Climate change has been a very pressing issue lately. All global leaders have been called to intervene, and alternative energy sources are a major go-to for most countries. The rise in demand for renewable resources is yet to be met with efficient technology. The use of clean, environmentally friendly renewable energies has become essential to combat climate change and air pollution. Currently, solar photovoltaic technology makes up 55 per cent of all renewable-power capacity, according to a 2018 report. Climate change is affecting the performance and efficiencies of Photo Voltaic (PV) panels. Successful integration of increased levels of solar power, while maintaining the same reliability requirements, depends heavily upon the accuracy of the forecasted values of solar irradiance.

In this article, we focus on how predicting the solar radiation patterns and the power output of a plant is helping further cement the position of solar energy as one of the most viable renewable energies. Prediction also helps in estimating the exact requirement of solar panels. Forecasting solar radiation based on historical data analysis and meteorological features using deep neural networks enables an accurate way to predict this. This paper worked on by Arghya Mukherjee, Antara Ain and Prof Pallab Dasgupta, at Indian Institute of Technology Kharagpur, presents three models to determine the same.

To address the planning and operational requirements of transmission and distribution grids, two purposes have to be met. The short-term goals are to make real-time decisions for grid operation, and the long-term goals are of interest to companies for commitment to a particular power grid. Due to the unpredictable nature of cloud motion, and chaotic variation of weather parameters, a robust, high-accuracy solar forecast system is a challenge.

Technical crux:

Deep neural networks performed very well here in prediction-based tasks, due to their excellent learning abilities. The data used are the hourly solar irradiance using the meteorological data, along with the historical trend at that particular location. The predicted solar irradiance output is further used to predict the solar power output of a PV plant at that location within a shallow margin of error. The knowledge of the previous irradiance patterns at a solar Photo Voltaic plant location, along with the meteorological parameters, can significantly reduce the errors in power scheduled from the plant.

Three major prediction models were used:

1. Artificial Neural Networks (ANN) Model with Meteorological Features

The input layer accepts as parameters, the hourly meteorological features of the timeslot. The output of the model is the predicted Global Horizontal Irradiance (GHI).

2. ANN Model with Meteorological Features and Previous Irradiance Trend

Here the input layer accepts the hourly meteorological features of the day similar to the previous model, along with the yearly and monthly history data of irradiance patterns.

3. Long Short-Term Memory (LSTM) network model with Meteorological Features and Previous Irradiance Trend

This model considers the sequential information available in the consecutive times, and the output is a function of the meteorological features, historical trends, as well as the outputs of previous times.

The expansion of the global population and the per-capita demand, along with the rising impacts of climate change are going to push renewable energy into every individual’s life. Knowing the accurate power output that can be generated in a location can provide the suppliers, regional planners and the consumers more information and power to make wiser decisions.

So, cheers to a better climate, hoping AI will drive it.

• Jasmine Jerry A

Solar Irradiance Prediction from Historical Trends Using Deep Neural Networks, A Mukherjee, A Ain and P Dasgupta, IIT Kharagpur.

Jasmine Jerry is a third-year undergraduate pursuing Aerospace Engineering passionate about robotics and unmanned systems. She spends most of her time on soccer-playing robots or aerial vehicles. She also is a member of Lean In Chapter at IIT Kharagpur.