Study of Power System Planning by Two-stage Stochastic Programming

• Main Authors: Chih-WeiChiang, 蔣志偉
• Other Authors: Rong-Hwa Wu
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/75230343491194922816

碩士 === 國立成功大學 === 資源工程學系碩博士班 === 101 === Power plant investments are generally irreversible, lumpy, capital intensive, long lived, and have significant lead times. Investments are also necessarily undertaken in the context of projections of future electricity demand. However, future electricity demand is subjected to a range of uncertainties, particularly the economic growth and patterns of development. Overestimating long-term electricity demand can lead to overinvestment in generation assets and significantly higher industry costs. On the other hand, underestimating electricity demand can lead to insufficient investment in generation capacity, resulting in unserved demand and potentially significant adverse impacts on economic progress. Traditional electricity supply planning model regards the electricity demand as deterministic parameter (i.e. the electricity demand is given exogenously) to select power generation technologies. But in today’s world, the energy planners are facing tremendously complex environments full of uncertainty and risks, and electricity industry is also in an uncertain situation, where the assumption of certain electricity demand is apparently unreasonable. Therefore, the research aims to establish an electricity supply planning model incorporating electricity demand uncertainty. The two-stage stochastic programming methodology is applied in the research to deal with the uncertainty in electricity demand. The model also combines decision tree and Monte Carlo methodology to reduce possible nodes and determine the future electricity demand values and probabilities. Using the model, we implement some simulation scenarios to evaluate the impact on the portfolio of power generation technologies and generating cost. The simulation results obtained by applying the model to Taiwan's electricity sector indicate that the two-stage stochastic programming methodology can explicitly address the electricity demand uncertainty. Monte Carlo methodology can characterize uncertainty by assigning a normal distribution to uncertain electricity demand, which can be estimated from historical data. Consequently, the probabilities of future electricity demand can be determined endogenously instead of given exogenously. In the nuclear-free scenario, the simulation shows that the share of coal-fired power plants will grow from 36.85% in 2010 to 50.4% in 2025 as all nuclear power plants decommission in 2025. In the reduction of carbon dioxide emissions scenario, the proportion of coal-fired power plants reduces and is replaced by LNG-fired and renewable energy power plants. Hence, this contributes significantly to the increase in total generating cost.