Project Abstract:
Fossil fuels are an unsustainable source of energy, and their combustion harms human health, according to the World Health Organization. There is an urgent need to expand renewable energy resources. While many experts agree that wind power is the most viable alternative, resources are wasted due to inaccurate predictions of turbine power output. If power plants are wasting money and energy, they have no incentive to use wind power. To address this problem, I created a way to make wind power more feasible using machine learning. Machine learning is a type of artificial intelligence that teaches the computer to learn from existing data and predict outcomes for future data. My innovation -- Hybrid Algorithms for Wind-power Computation (HAWC) implements several types of algorithms that predict turbine power output: 2-layer neural network, 3-layer neural network, linear regression, and polynomial regression. My hypothesis that the 3-layer neural network is the most accurate for this particular problem was proven correct, since more layers provide a more detailed iterative prediction method. Layers refer to the steps a neural network executes to process data, train the algorithm, and produce output. An ensemble of the four algorithms I tested has not previously been implemented, and is a contribution to the field because of its high accuracy. My results suggest that HAWC has the potential to make wind power more feasible by allowing power plants to integrate wind energy into the power grid.