Determinants of profitability in the Dutch positive balancing power market
A case study at AkzoNobel
More Info
expand_more
Abstract
Intermittent generation behaviour of wind and solar power units requires Transmission System Operators (TSOs) to activate balancing power to maintain a stable grid frequency. With the current strong growth in installed renewable energy capacity in The Netherlands, it is therefore expected that the volumes of activated balancing power will grow accordingly. In the Dutch balancing power market system, the activation of balancing power is remunerated against the highest activated balancing power bid in the respective time-block. Without a proportional increase in affordable balancing power bids, the economic costs of grid balancing in The Netherlands are expected to increase as a result. Imbalance power prices for positive balancing power, the counterbalancing of net shortages of power supply, may see a particularly strong rise, as the opportunity to quickly increase grid frequency places influential restrictions on the operation status of suitable power assets.
The financial opportunities inherent to this trend have gained the attention of non-traditional suppliers of technical positive balancing power potential, amongst which the energy-intensive industry and greenhouse farmers. However, limited insight into the exact value potential of market participation, and contradictory developments in the German market are currently preventing these parties to invest in unlocking their technical balancing power potential. A limited supply of new balancing power bids will further increase the economic costs of grid balancing. This study therefore sets out to identify a method to quantitatively determine the profitability potential of providing positive balancing power to the Dutch balancing power market to an individual market participant. A case study is conducted at AkzoNobel, whose sizeable chlorine plant possesses significant balancing power potential.
It is found that a Unit Commitment Model based on a Mixed Integer Linear Programming solver can be used to successfully replicate the market environment. To simulate market profitability to AkzoNobel over time, it should be supplied with input data of the future development of balancing power activation, the merit order of balancing power bids, and balancing power capabilities of the AkzoNobel plant.
Whilst the balancing power capabilities of the plant can be relatively simply and accurately modelled into a Unit Commitment Model, the same does not hold for the future activation of balancing power and respective prices in the merit order of balancing power bids. On the one hand, the future development of the various factors that comprise these sub-systems is characterized by significant uncertainty, as power system dynamics are notoriously hard to predict. On the other, both are the emergent behaviour from complex systems of interacting power system agents and assets. Their actions result from a large number of factors that may vary strongly on a short time scale. This causes difficulties in obtaining accurate projections of their behaviour on the required minute-by-minute time scale.
The activation of a high balancing power bid price in general also implies a strong grid frequency deviation. This implies that the profitability of providing balancing power is particularly sensitive to deviations in high imbalance power prices. As imbalance power prices are primarily determined by the specific height, volatility and frequency of high balancing power activation peaks relative to the height of prices in the merit order for those respective fifteen-minute time blocks, the choice of a method to obtain the input data of balancing power activation and the merit order must be based on a clear substantiation of their accuracy.
Various approaches to obtain data projections with the required accuracy have been assessed, for which the following conclusions are drawn. To determine the development of the future activation of balancing power bids it is advised to evaluate a combined method of statistical convolution of probability distributions and hidden Markov chains. However, this method must be preceded by significant additional research on critical complexities and uncertain developments in the balancing power market, for which it may not be possible to draw definite conclusions.
To obtain scenarios of the various potential developments in the volume and price elements of balancing power bids in the merit order it is suggested to use a stochastic modelling approach, such as Monte Carlo simulation, to evaluate the range of different compositions the merit order may take over time. It will, however, be difficult to strategize on the results, as the potential range of states of bid prices and volumes placed by individual agents with balancing power technologies are very wide.
These methods must be synthesized into an approach to determine balancing power market profitability based on the above combination of methods. Projections of the development of the key determinants of profitability in the balancing power market will inherently be associated with strong uncertainty. Considering the sensitivity of model output to inaccuracies it is very questionable whether the effort of further research weighs up against the benefits. It is concluded that it is not deemed feasible to obtain quantitative long-term indications of balancing power market profitability to an individual market participant. As a result, it is not advised to proceed with the execution of the suggested methods. It appears that alternative approaches to ensure the future affordability of balancing the power grid are more worthwhile pursuing.
It is instead advised to focus continued research on public incentive schemes to attract non-conventional suppliers of positive balancing power to provide their technical potential to the market if the market fails to find an economic optimum. Continued research could also focus on enhanced regulatory designs to reduce total balancing power requirements.