مقاله انگلیسی رایگان در مورد مدل سازی تعادل بازار انرژی با عوامل متعدد (الزویر)

Agents in the European energy market have experienced, and may continue to experience, considerable uncertainty: during the last financial crisis demand for energy dropped significantly, and in the future the Paris agreement may trigger radical changes in the energy markets. Such abundant uncertainties can have huge consequences on investment in the energy industry. At the same time, if there is reluctance to invest in some technologies, for example, due to expectations about high fossil fuel prices or high taxes on greenhouse gas emissions, the market may look more promising for other technologies, like renewables. Thus, to fully analyze the impact of uncertainty the interdependences of different technologies, energy goods and agents have to be taken into account – this calls for a multi-dimensional equilibrium model that captures the essential characteristics of the energy industry. It is, however, not trivial to solve, or even formulate, a model where many heterogeneous decision makers face uncertainty. Thus, it is not surprising that most analyses assume full certainty, or, if uncertainty is analyzed, rely on Monte Carlo simulations instead of examining the behavior of agents optimizing under uncertainty. Some examples of Monte Carlo simulations, which are also termed what-if analysis, sensitivity analysis, or scenario analysis, are IPCC (2014) on how short- and long-term factors have impact on mitigation pathways, Egerer et al. (2015) on how alternative scenarios have impact on intermittent renewables in the German electricity system, and Fæhn and Isaksen (2016) on impacts, within a technology-rich CGE model, of regulatory climate policy uncertainty. In order to properly handle uncertainty, the basic ideas of stochastic programming, see, for example, Kall and Wallace (1994), should be built upon. Here, a crucial distinction is made between decisions made before the uncertainty is revealed, and decisions made afterwards. A key insight is that it is not valid to solve the model scenario by scenario – as is the case with Monte Carlo simulations – and then try to extract an overall picture from these solutions.