Transition to Centralized Unit Commitment: An Econometric Analysis of Colombia's Experience.

• Author: Riascos, Alvaro

1. INTRODUCTION

   System operators (SO) in electricity markets have the responsibility of balancing supply and demand of electricity at each moment in time, taking into account all of the constraints in the system. One of the most important elements of this task is the dispatch of generators.

   There are essentially two ways of determining which generators are to be dispatched in restructured electricity markets. In self-committed markets, generators place bids for energy production and the SO chooses the least-cost producers. In centrally committed markets, generators submit their cost of production and their fixed start-up (and possibly no-load) cost. These fixed costs are taken into account in the optimization problem resolved by the SO and are used to calculate an uplift payment to dispatched generators that do not fully cover their fixed costs through their energy revenues. In contrast, in self-committed markets, generators can only recoup their start-up costs directly through their energy bids.

   Of course, efficiency requires that the lowest-cost producers be chosen at each moment in time and that these costs include the generators' start-up costs. Thus, at first glance, centrally committed markets may seem preferable. However, the change in rules also affects the strategic behavior of agents, who may have greater opportunities for misreporting information. Therefore, it is not clear which method is superior.

   Indeed, there has been a debate in the literature about this issue. Some authors, such as Ruff (1994), Hogan (1994), Hogan (1995) and Hunt (2002) prefer centrally committed markets. On the other hand, Oren and Ross (2005) show that generators may have incentives to misreport their costs. Wilson (1997) and Elmaghraby and Oren (1999) suggest that self-committed markets may end up being more efficient when bidders' strategic behavior is taken into consideration. Sioshansi and Nicholson (2011) analyze the equilibrium behavior in both designs and show that there are opportunities to misreport in both. Thus, while all SOs in the United States have adopted a design based on voluntary centralized unit commitment for day-ahead markets, so far the theoretical literature has not been able to determine which method is superior. Thus, this important market design question remains an empirical one.

   In this paper, we shed some light on the foregoing debate by taking advantage of a natural experiment performed in the Colombian electricity market, where the market design was changed in 2009 (1) from a self-committed one to a centrally committed one. We perform a comprehensive analysis of the Colombian market before and after the change and reach two main conclusions. First, the centrally committed market contributed to higher productive efficiency. (2) Second, we find evidence that marginal cost markups and prices after 2009 were also higher than they would have been under the regime before the change, possibly as a result of an increase in exercise of market power by generators. These findings suggest that consumers have not benefitted from efficiency gains and although productive efficiency has increased, the additional strategic flexibility of generators has reduced consumers' surplus; depending on demand elasticity, this could have resulted in reduced social welfare. We show that this is true even if we ignore spot prices and focus only on the average price of bilateral contracts.

   This paper is organized as follows. In Section 2, we describe Colombia's electricity market rules before and after 2009. We also describe the unit commitment problem that the system operator XM (Compama de Expertos en Mercados) solves and how each plant is remunerated. Section 3 contains a description of the data used. The econometric analysis is presented and discussed in Section 4 where we argue that productive efficiency has increased since 2009. Section 5 provides evidence of an increase in market power after 2009 and that efficiency gains were not passed on to consumers through lower prices. Section 6 contains the conclusions.

2. THE PROBLEM

   In this section we briefly explain Colombia's spot market design before and after the implementation of resolution 051/2009. (3) We focus on the domestic market and exclude international exchanges with Venezuela and Ecuador. (4)

   Beginning in 2001, Colombia operated a day-ahead market where each generator offered a single bid for energy production for the next 24 hours. The system operator (SO) used these bids to determine which generators would produce. For the spot market and energy dispatch prior to Regulation 051 (i.e., before 2009), (5) there are three relevant points in time: day ahead (economic dispatch), real-time dispatch (real dispatch) and day after (ideal dispatch). The main features of the economic dispatch are:

   1. Plants submit two-part bids: a minimum price at which they are willing to generate during the next 24 hours along with their maximum generation capacity for each of the next 24 hours.

   2. Plants inform the system operator (SO) on what fuel and plants configuration should be used for solving the unit commitment problem.

   3. The system operator estimates the following 24 hours total demand for each hour.

   4. Basic technical characteristics of plants are taken into account: a ramp model for thermal plants (minimum uptime, minimum downtime (6), etc.), minimum energy operating restrictions for hydro plants, etc.

   5. Automatic generation control (AGC) restrictions are taken into account. (7)

   6. Transmission restrictions are given.

   7. Every day, the economic dispatch optimizes the following function:

   [summation over (t=0,...,23)] [summation over (i)] x [q.sub.it]

   where [b.sub.i] is the price bid by plant i for the next 24 hours and [q.sub.it] is the production of plant i in hour t subject to hourly AGC, transmission, demand and technical constraints (ramps), environmental restrictions, etc.

   This optimization defines the economic dispatch for every hour and provides a scheduling plan for energy generation for the next 24 hours.

   Real-time generation sometimes deviates from the planned economic dispatch for a variety of reasons: demand turns out to be slightly different than the demand estimated on the previous day, energy losses, overloaded lines, etc. Therefore, the system operator has to fine-tune the actual dispatch in real time.

   Once the real generation for the 24 hours has occurred, the system operator calculates the ideal dispatch, which is an ex-post calculation used for settlement purposes. The optimization problem solved is the following:

   [mathematical expression not reproducible]

   where [b.sub.i] is the price bid by plant i for the next 24 hours, [q.sub.i,t] is the production of plant i in hour t and [D.sub.t] is actual demand at time t. Notice that the ideal dispatch is determined through an hour-by-hour optimization problem.

   The ideal dispatch forms the basis for calculating the spot price. (8) Once the optimization problem of the ideal dispatch is solved for every hour the market clearing price is calculated as the price bid by the marginal plant that is not saturated and which is needed to meet demand. (9) We denote this equilibrium price as [b.sup.m.sub.t]. The hourly spot price [P.sub.t] is defined as this equilibrium price, [P.sub.t], = [b.sup.m.sub.t] (since 2009, the spot price is modified by an uplift as explained below).

   Since the real dispatch turns out to be different than the ideal dispatch, side payments are implemented to pay for any differences. (10)

   After the regulation of 2009, the ideal dispatch solves a centralized unit commitment problem. Rather than minimizing the hourly costs of generation, the objective function was set as equal to the objective function of the economic dispatch (24-hour optimization problem), generators submit complex bids and side payments were introduced. The bids specify an energy offer price for the next 24 hours, start-up costs for the next three months and maximum generation capacity for each hour in the next 24 hours.

   Once the optimization problem of the ideal dispatch is solved for the 24 hours, the equilibrium price [b.sup.m.sub.t] is calculated as the price bid by the marginal plant that is not saturated. The hourly spot price [P.sub.t] is defined as this equilibrium price plus an uplift, [DELTA]I, where the uplift is defined in the following way.

   Let

   [I.sub.i] = [24.summation over (i=1)][q.sub.i,t] x [b.sup.m.sub.t]

   be the plant's i income according to the ideal dispatch and:

   [C.sub.i] = [24.summation over (t=1)][q.sub.i,t] x [b.sub.i]+[24.summation over (t=1)][S.sub.i][u.sub.i,t]

   be the plant's i generating cost (assuming truthful bidding) where si is plant's i start-up costs and [u.sub.i,t] is a binary variable that is 1 if the plant is operating in period t and 0 otherwise.

   Now let [q.sup.s.sub.i,t] be plant i energy production at the time when it is saturated (0 otherwise) and P[R.sub.i] the positive reconciliation price. (11) Then the uplift is defined as:

   [DELTA]I= [[[SIGMA].sub.i]{0,[C.sub.i]-[I.sub.i]}+D[I.sub.i]]/[[SIGMA].sup.24.sub.t=1][D.sub.t]

   Panel A (upper) shows thermal vs. hydro generation as a proportion of total generation. Panel B (lower) shows the spot price in Colombian pesos per KWh.

   where:

   D[I.sub.i][24.summation over (t=1)][q.sup.s.sub.i,t]x(max{[b.sup.m.sub.t],P[R.sub.i]}-[b.sup.m.sub.t])

   The hourly spot price is defined as:

   [P.sub.t] = [b.sup.m.sub.t] + [DELTA]I

   Therefore, the spot price guarantees that demand will pay for the start-up of dispatched plants and energy production by saturated plants. Having defined the spot prices, we now explain the settlements for the various agents. Agents are paid the spot price for any unit of energy produced (regardless of whether the plant is saturated or not) and hydro plants reimburse the [DELTA]I component of the price for each unit of energy produced, while thermal plants for which [C.sub.] [less than or equal to] [I.sub.i] also reimburse the...