Case Study Conclusions

To get insights about the impact of demand flexibility, we compare two cases: one with highly flexible demands in term of their ability to shift consumption across periods in the day-ahead market without considering the demand ability of reserve provision for real-time deployment; and the case with inflexible demands including inelastic demands following a traditional consumption pattern with peak consumption in day hours and off-peak consumption in night hours. The common feature of the two cases is that demands have the same total energy consumption over the entire clearing horizon. Therefore, we are able to explore producers and consumers expenses for the same amount of energy consumption. Based on the observations in these case studies, we conclude the followings:

1. Demand flexibility is beneficial to the system, as it shifts consumption from day hours to night hours, when cheap wind production is available. In other words, demand flexibility adapts its consumption pattern to the production pattern of the wind unit. Specifically, the scheduled production of the wind unit (i.e., the contribution of wind production in energy supply) is almost twice in the case with flexibility. This leads to a cost reduction of about 25%, as compared to the cost obtained from the case with inflexible demands.

2. Due to the notable shift in consumption in the case with demand flexibility, a price shift occurs: the prices from the flexible demand case are comparatively higher over the off-peak periods and comparatively lower over the peak-periods with respect to the prices from the case with inflexible demands. Network congestion and ramping limits of the conventional units do not change these conclusions.

3. Higher off-peak prices result in a higher consumer payment in the flexible demand case than that of the case with inflexible demands. The increase in consumer payment is 9% in the base case and 7.5% in the case with congestion and limited ramping capability. 4. The observations from the case studies show an inherent conflict in incorporating

demand flexibility to an electricity market: on one hand, the system benefits from the reduced operation cost caused by demand flexibility, but on the other hand, the resulting prices increases the demand expenses. Thus, demands might be better off being inflexible.

4.7. Summary and Conclusion of the Chapter

4.7 Summary and Conclusion of the Chapter

This chapter is devoted to analyze the economic consequences resulting from the actions of flexible demands in a common future market consisting of a significant amount of renew- able power production with comparatively low marginal cost and fast-ramping units with comparatively high marginal cost, such as combined-cycle gas turbines.

On one hand, demand flexibility (the ability of some demands to move load from peak periods to off-peak periods) is beneficial for the system as a whole since it decreases the expected operation cost, but on the other hand, demand flexibility can result in price increases that in turn increase demand expenses. Therefore, demands might be better off being inflexible in systems with a generation mix dominated by comparatively cheap renewable units and comparatively expensive fast-ramping units.

We should note that if pricing schemes other than the one adopted [46] are used, e.g., a convex- hull pricing [24], the conclusions derived in our study are likely to remain valid provided that the final prices do not deviate significantly from marginal prices.

We should also note that the use of a stochastic clearing model is for the purpose of obtaining optimal outcomes in a market with a high penetration of renewable generation, and the choice of clearing model, i.e., a deterministic or a stochastic one, does not change the conclusion above.

The observations in this chapter call for new settlement approaches seeking to encourage demand flexibility.

5

Two-Stage Stochastic Clearing Model

for the Reserve Market

5.1 Introduction

The system operator is responsible to ensure system security in power systems. A mechanism to do so is reserve; in real-time operation, there is a need to compensate mismatches between supply and demand in order to preserve the power balance in the system. For this purpose, reserves are scheduled in a market prior to real time to be eventually deployed in real-time operation. The structure of this market (e.g., gate closure, type of offers, etc.) depends on the market organization, i.e., a centralized market organization and a decentralized one.

In a centralized market organization, such as electricity markets in the US, reserves are sched- uled in the day-ahead market co-optimizing energy and reserves (this is similar to the clearing models in Chapter 2), whereas in a decentralized market organization, such as electricity markets in Europe, reserves are procured in reserve markets separately from energy markets. In a centralized market, the system operator and market operator are generally the same entity. However, a decentralized market separates energy transactions and system operations to a large extent; the former is done by the market operator, whereas the latter is the responsibility of the Transmission System Operator (TSO) [3].

The common practice in many European countries, as examples of the decentralized market organization, is first to determine fixed amounts of reserves (of different types) using technical (security) criteria, and then, to procure them in reserve markets. We challenge this practice as it decouples technical criteria from market aspects, which may result in economic inefficiencies. The particular focus of this chapter is the reserve market in Switzerland, as an example of a decentralized market organization. We use two-stage stochastic clearing models to show the advantages of our proposed model with respect to deterministic one, not only through simulated case studies, but also through the outcomes of the actual implementation of the

proposed two-stage clearing model.

The lay-out of this chapter is as follows. We first describe the Swiss reserve market in Section 5.2. Next, the decision-making process and scenarios are described in Sections 5.3 and 5.4, respectively. Section 5.5 provides the assumptions that we use to model the reserve market. Section 5.6 provides the mathematical descriptions of the proposed risk-neutral two-stage stochastic model (in Section 5.6.1), of the proposed risk-averse counterpart (in Section 5.6.2), and of a common deterministic model (in Section 5.6.3). Also, we formulate how to obtain a perfect information solution as well as the actual cost of the two-stage model in Section 5.6.4. The proposed models are showcased through real cases from the Swiss reserve market in Section 5.7. Finally, relevant remarks are concluded in Section 5.8.