The Introduction of NonDIispatchable Technologies a Decision Variables in Long-Term Generation Expansion Models

Non-dispatchable technologies (solar, wind, run-of-the-ri ver hydro, cogeneration) affect the cost of electricity production in a complex manner by modifying the probability distribution of demand for conventional generation. The lack of an appropriate methodology to efficiently derive this modification has prevented inclusion of non-dispatchable technologies as decision variables in capacity expansion models. This paper develops a stochastic approach to load modification which explicitly models two types of interdependencies between load and non-dispatchable generation: through time of day and through weather. If more than one non-dispatchable generation technology is considered, the dependency among them is also modeled. Furthermore, the total as well as the marginal impact of non-dispatchable capacity on system reliability and operating cost is derived. This allows us to model non-dispatchable generation in the context of two broad classes of optimization algorithms. Dynamic programming and mathematical decomposition are considered in this paper as characteristic examples of algorithms in each class. Load modification models already in use derive total cost impact only, and are based on hourly chronological simulation which is a computationally cumbersome method. The methodology developed here provides marginal impact in addition to total impact values, is computationally efficient and is applicable to future demand projections at almost any level of detail. Finally, its accuracy proved very satisfactory when tested on 1975 Miami load and insolation data.