(45a) Power Capacity Expansion Planning Considering Endogenous Technology Cost Learning | AIChE

(45a) Power Capacity Expansion Planning Considering Endogenous Technology Cost Learning


Heuberger, C. F. - Presenter, Imperial College London
Rubin, E. S., Carnegie Mellon University
Staffell, I., Imperial College London
Shah, N., Imperial College London
Mac Dowell, N., Imperial College London
Integrated modelling can guide structural and operational decision-making in energy systems. Long-term models can assess pathways complaint with future demands and environmental targets and optimal investment timing for power generation, energy storage, and transmission capacity. Short-term operational models can elucidate the role and value of specific technologies within the power system.

We develop a mixed-integer linear formulation of a hybrid power generation capacity expansion and unit commitment model with a high level of technical detail and granular time discretisation [1]. The model considers up to 2000 units of 16 different power generation and storage technologies, including international interconnectors for electricity import and export, and grid-level energy storage. The national-scale single-node formulation includes the consideration of endogenous learning for technology capital cost. In line with previous approaches, we determine a piece-wise linear formulation of the exponential learning curve model [2, 3]. A reformulation of the non-convex term of specific unit cost installed capacity is implemented in form of the cumulative investment cost as a function of cumulative installed capacity.

The model is applied to the power system of the United Kingdom for the years 2015 to 2050, subject to electricity demand and ancillary service requirements, CO2 emission targets, technology specific performance parameters, and learning rates [4]. Wefind that the consideration of cost learning for an individual technology moves the optimal investment timing to earlier planning years and increases the economic amount of capacity deployment of the respective technology. Other unsupported technologies, however, might experience deployment delays and become less competitive. These systemic implications on the profitability of technologies is again observed in scenarios where all technology types experience cost learning. We compare a case of local (national) and global technology learning curves, andfind more pronounced effect in the case of global technology diffusion assumptions.

Specifically in the case of offshore wind capacity, wefind that the omission of cost reduction effects leads to an underestimation of the optimal economic deployment level in 2050 by 32 % to 50 %. Similarly, coal-fired power generation with Carbon Capture and Storage capacity which is crowded out by less costly low-carbon technologies under static cost assumptions, becomes an attractive investment when cost reduction is taken into account. System-wide effects prove that early investments are able to drive technology learning and reduce total system cost by mid-century. In addition, wefind that the model results are highly sensitive to the build rate parameter which limits the maximum annual capacity additions. A high build rate can reduce cumulative total system cost by a multiple of the cost learning effects.

[1] C. F. Heuberger, I. Staffell, N. Shah, and N. Mac Dowell. Levelised Value of Electricity - A Systemic Approach to Technology Valuation. In 26th European Symposium on Computer Aided Process Engineering, volume 38, pages 721-726, 2016.
[2] S. Messner. Endogenized technological learning in an energy systems model. Journal of Evolutionary Economics, 7(3):291-313, 1997.
[3] A. Gruebler and S. Messner. Technological change and the timing of mitigation measures. Energy Economics, 20(5-6):495-512, 1998.
[4] E. S. Rubin, I. M. L. Azevedo, P. Jaramillo, and S. Yeh. A review of learning rates for electricity supply technologies. Energy Policy, 86:198-218, 2015.