Stochastic techno-economic assessment of future renewable energy networks based on integrated deep-learning framework: A case study of South Korea

Abstract

This research aims to develop comprehensive deep-learning frameworks that integrate both stochastic and deterministic algorithms, with a focus on proposing a stochastic techno-economic assessment (TEA) for evaluating national energy policy in a specific case study. The approach involves employing a variational autoencoder, a form of stochastic deep-learning to produce renewable energy samples, and various deterministic deep-learning techniques such as recurrent neural networks to train these samples. The optimal forecasting model is identified through a synthesis of evaluation metrics and information criteria. The practicality of this model is showcased in a case study on South Korea's energy policy for 2030. Probability distribution functions-derived stochastic TEA calculates the total costs and the levelized cost of energy (LCOE) for three separate sub-case studies. Each of these is defined by a certain level of energy shortfall, calculated as the difference between planned energy production and actual energy realization. Results suggest that to meet the targets of South Korea's energy policy for 2030, additional implementation of solar power systems is inevitable. This study provides an estimation of the LCOE within the range of 0.025 to 0.051 USD/kWh. The annual total cost for expanding the energy plant with a photovoltaic capacity of 100 %, based on the best-case scenario, is estimated to fall between 6.1 and 9.5 × 108 USD/year. This research is anticipated to make a significant contribution to the formulation and development of various energy policies. It is expected to assist decision-makers in the field of energy policy in crafting and executing strategic plans.