Estimating CO2 reductions from renewable energy sources : The impact of power system nonlinearities

Abstract: Replacing conventional generation with renewable generation in power systems is essential for reducing CO2 emissions. It is important to know how effective renewables are in reducing CO2 emissions. Since CO2 reduction cannot be measured directly, different methods have been used to estimate reduction of CO2 emissions. The two most common methods are econometric models and dispatch models. Econometric models apply regression analysis using historical data for CO2 emissions, power production, and electricity demand to estimate CO2 reduction. On the other hand, dispatch models are detailed optimization simulations of power systems where the objective is to calculate the cost-optimal dispatch of the power plants. The dispatch model finds the optimal dispatch for a base case and counterfactual case. In the counterfactual case, the renewable generation in the system is modified. From the difference in CO2 emissions between the two cases, an estimation of CO2 reduction can be made. Recent studies have shown that dispatch models and econometric models can give different estimations of CO2 reduction. However, these studies did not include several factors that can increase CO2 emissions, such as; transmission constraints, security requirements, and non-linear factors. Examples of non-linear factors are; minimum dispatched energy of generating units, start up emissions, minimum up- and downtime for generating units, and energy generated during start-up and shut-down. This thesis examines if there is an agreement between econometric models and dispatch models for estimating CO2 reduction and if the agreement changes when more non-linear factors are considered. To examine these questions a systematic comparison has been done. Two econometric models are constructed, a linear econometric model and a polynomial linear econometric model. The polynomial linear econometric model is constructed to take into account non-linear factors. Eight dispatch models are constructed with increasing modelling complexity. Four model versions do not include any non-linear factors and four include non-linear factors. The results showed that the agreement between econometric and dispatch models is fairly good, except for versions that contain transmission constraints. The simulation is executed in a fictional test system that is not dimensioned for wind power generation at the given buses. Therefore is possible that transmission constraints impacts the reduction of CO2 too heavily. Furthermore, the results show that the non-linear factors contribute to CO2 emission and consequently lower the estimation of CO2 reduction. However, no conclusion can be made if the agreement between econometric and dispatch models divert when more non-linear factors are considered.