MP in Action‎ > ‎Other Applications‎ > ‎

Regulating Electricity - Preventing Another Own GOAL

This article was prepared as a curtain-raiser to the Blackett Memorial Lecture delivered by Professor Stephen Littlechild, Director General of Electricity Supply on 30th November 1995.

Electricity Prices and GOAL

As domestic consumers, we pay about 8p a unit for electricity (or 3p a unit for that consumed during 7 hours each night). As well as this price there is a wholesale price at which electricity is traded in what is known as "the Pool". This price varies every half hour and is published every day in the Financial Times. Typically this price varies between about 1p a unit during off-peak periods and 5p a unit during peak periods. However, it may rise much higher, up to 20p or 30p a unit, and, in exceptional circumstances, even beyond this.

This wholesale price for electricity (the Pool price) is set in a complicated way which involves:

  • offers by the generators of the prices at which they will generate electricity at each of their power stations for each half-hour;
  • forecasts of demand for electricity for each half-hour;
  • a computer program, GOAL, which schedules the power stations to meet forecast demand, supposedly in the most cost-effective way.

The problem of scheduling the power stations is highly complex, as will be explained below. It is, however, crucial, because the Pool price is set as the highest offer price by a power station which is accepted and included in the schedule. Thus if GOAL accepts a higher offer price than it needs to, all the generators receive that higher price for the electricity which they generate in that half-hour.

What does this have to do with Operational Research and Regulation, the subject of Professor Littlechild's Blackett lecture? Two things:

  • GOAL is one of the largest Operational Research models in use in the U.K. today;
  • Professor Littlechild's remit is to police the electricity supply industry and ensure fair play.

It is widely accepted that GOAL is capable of being manipulated. A successor is under development. The Operational Research community is uniquely well placed to assist Professor Littlechild in ensuring that GOAL's successor truly is fair.

Electricity Privatisation

Before privatisation the electricity industry in England and Wales was owned and run by a single organisation, the Central Electricity Generating Board (CEGB). This encompassed the power stations, the National Grid and the regional distribution systems. After privatisation these became respectively the generating companies (Power Gen, National Power and Nuclear Electric), the National Grid Company and the regional electricity companies (RECs).

The CEGB's primary responsibility was to ensure that "the lights remained on", i.e. sufficient electricity was generated at all times and was distributed around the country to ensure that demand was met. To do this it used a computer program, GOAL, to work out how much electricity the power stations should generate and how the power should be distributed around the country.

When the electricity industry was privatised, GOAL passed from the CEGB to the National Grid Company (NGC) where it has continued to be used for scheduling the power stations. But a critical change had occurred.

Formerly the power stations were all owned by the CEGB, so it did not much matter if the schedule which GOAL produced was less than optimal. The CEGB's costs were only affected very marginally by less-than-optimal solutions and these costs were not noticed by consumers.

Now the various parts of the industry are independent, so the costs and profits of the separate participants are directly affected by the quality of GOAL's schedules. Because the price set for electricity in the Pool (the wholesale market for electricity) is the marginal price, the quality of GOAL's schedules directly affects everyone.

In practice, the RECs have hedging agreements with the generators which stabilise the price which they pay and render them largely indifferent to Pool prices. But the large industrial consumers do pay Pool prices and so they are directly affected. Nor can they so readily afford the hedging agreements which the RECs have. Many of the RECs' agreements were written at the time of privatisation. Others were bought secure in the knowledge that their costs could be passed on to consumers under the RECs' pricing formula.

The Scheduling Problem

Every day NGC draws up a forecast of demand for electricity for each of the half-hours in the day ahead (i.e. on Monday it draws up the forecasts for Tuesday). These are based on a combination of:

  • the historic pattern of demand through the day;
  • the weather forecast for the day ahead;
  • any special factors which may affect demand (e.g. as I write this, the final episode of Pride and Prejudice).

Each of the generators submits to NGC a computer file of its offers to generate electricity. These give for each half-hour:

  • the amount of generating capacity offered at each power station;
  • the price for generating that electricity (possibly several prices for separate tranches);
  • constraints on its availability (e.g. minimum and maximum rates of ramping up and running down generation).

In deciding which power stations to schedule, NGC must consider not only the aggregate demand and supply but the locations at which the demand and supply occur and the capacity of the transmission system to transfer power around the country. Thus where limits on transmission capacity exist, NGC may be obliged to schedule an expensive local power station regardless of the price quoted.

Modern gas-fired power stations can largely be turned on and off at will. Not so coal-fired power stations, which take several hours to ramp up generating capacity, and even less so nuclear stations, which generate continuously. It is these time dependencies which render the scheduling problem enormously difficult. Without them, the problem would be straightforward.

Scheduling Technology

Scheduling is difficult. There is no technical panacea. There are various Operational Research techniques which address part of the problem, but none which could guarantee to find the optimal schedule for electricity generation.

It is worth pausing for a moment to consider the word "optimal". NGC's aim is to find that schedule which minimises the total cost of generating electricity while meeting demand and observing operational constraints. "Optimal" means lowest cost among those schedules which meet demand and satisfy the constraints. NGC's problem is thus one of constrained optimization.

If one disregards for the moment the time dependencies, (i.e. one considers each half-hour separately) the electricity scheduling problem is entirely linear. The appropriate technology for solving that problem is linear programming (LP). It is robust, well-proven and perfectly capable of solving the problem to optimality.

If one reintroduces the time dependencies, the problem remains primarily linear but has logical interconnections. It can be posed as a massive integer programming problem. Unfortunately, despite many recent advances, integer programming is unlikely to be able to solve the problem to proven optimality. That is not to say that it would not do well, indeed might find the optimal solution. But it would be unlikely to be able to prove that it had done so. What it would do, however, would be to provide upper and lower bounds on the value of the optimal solution, so that one knew how far one was from optimality.

Other techniques which could be used are based on enumeration, e.g. dynamic programming, constraint programming. For a problem such as this, they rely more than integer programming on 'black art' heuristics to obtain solutions. The existing GOAL algorithms are largely heuristic.

Quis Custodiet...?

If one were setting out to establish a competitive market for electricity, one would not start from here. It is all very well saying that economic theory implies that the clearing price of a commodity is the price charged by the marginal seller. That is to ignore the dirigiste nature of the scheduling process and the imperative to keep the lights burning.

But given that we are here, what can be done? The main aim must be to make the scheduling process as close to optimal as possible. Even then, it would still be subject to manipulation by the exploitation of market power. That problem is, however, diminishing as the share of the market controlled by individual generators is reduced. It will be further reduced as sections of the National Grid are upgraded.

How can we make the scheduling process as close to optimal as possible? NGC has been granted a monopoly in scheduling the power stations. It has proved itself capable of maintaining supplies but not of doing so at minimum cost. When it was owned by the RECs its shareholders had an interest in minimising the cost of supplies, although in practice this interest was not large. With its imminent flotation NGC will not even have that interest. It will be in a position of power without economic responsibility.

NGC needs to be exposed to commercial risk or to competition. It could be made liable to being sued if it scheduled power stations unfairly or at higher than necessary cost. Although it has a natural monopoly of the wires and a wealth of experience, it is not clear that NGC is the only body capable of solving the scheduling problem presented to it each day. If a university or a software house can do better, why should it not be allowed to do so, and charge a hefty fee?

As a start, NGC should be obliged to open its scheduling algorithms to public scrutiny. If there are weaknesses in the algorithms, they are bound to be found by some of the market participants. Better to let everyone see what is going on and remedy the defects sooner rather than later.

Finally, it is worth considering whether the problems with Pool prices reflect a fundamental misconception. Does it make sense to seek to define a single price for electricity supplied in each half hour? Is electricity from a power station which takes six hours to ramp up and down the same as that from a station which can be turned on and off at will? There would be much merit in separating prices for instantaneously available electricity from that where producer and consumer commit for the next six hours or six days.

To find other articles, refer to the MP in Action page.