Depending on their trading strategies and skills, suppliers can gain or lose advantage in the prices they pay for wholesale supplies
Click here to enlarge image

The implementation of the New Electricity Trading Arrangements (NETA) on 27 March 2001 marked a further step in the development of a fully competitive market for electricity supply in England and Wales. The new arrangements affect the wholesale market for electricity and complement earlier reforms in the retail market.

In reflecting on their performance so far under NETA, suppliers should consider how to develop their trading functions and should not overlook the improvements that can be made to demand forecasting.

NETA introduced three major changes for suppliers. Firstly, wholesaling activity changed from being passive to active. Under the previous pool arrangements, suppliers did not make wholesale purchasing decisions. The physical balance between supply and demand was applied logically to each supply company. Each supplier’s purchases in a 30-minute settlement period were automatically set equal to its customers’ demands in the same period. Under NETA, suppliers must now actively trade before-the-event to secure wholesale supplies.

Secondly, wholesale prices have become much more of a differentiator between supply companies. The pool was an extension of the generation merit-order system used prior to privatization. Generators made offers to sell into the pool and all suppliers paid at the same rate. Although financial instruments, such as Contracts for Difference were in use, these were largely designed to give some security against pool price volatility. Under NETA, everything changes. Prices are set through trading. Depending on their trading strategies and skills, suppliers can gain or lose advantage in the prices they pay for wholesale supplies.

A balancing act

The structural relationship between generators, suppliers and the system operators
Click here to enlarge image

Supply companies also gained a commercial responsibility for balancing supply and demand. Before-the-event trading means each supplier is no longer automatically in balance. Furthermore, trading alone does not guarantee a balance between supply and demand for the system as a whole. The system operator uses the NETA Balancing Mechanism to resolve these issues.

The mechanism provides the operator with the flexibility to maintain a physical balance and creates the transactions that clear each participant’s imbalance. The mechanism is market based. As a result, a supplier whose trading activity results in a short position will tend to be sold the additional energy required at a premium. Similarly, a supplier who is long will tend to be cleared by selling its surplus at a loss. In effect, the Balancing Mechanism provides a financial incentive for suppliers to minimize their levels of imbalance.

The changes introduced by NETA bring new opportunities and risks. These must be managed carefully to ensure profitability and survival. The change from passive to active trading and the introduction of price as a differentiator means effective trading has become essential for all participants.

Trading development

Initially, the need for trading arises from the requirement to balance supply and demand. Trading facilitates effective price discovery and highlights new opportunities that should be exploited to ensure competitiveness. A supplier reducing purchase costs through effective trading can undercut competitors relying on own generation and long-term contracts.

Understanding and mitigating the risks inherent in energy trading moves the trading function forward. There will always be a degree of uncertainty in expected consumption and it is important to understand the scale and likelihood of this, plus the financial impact of last-minute sales and purchases to balance the physical position. A deliberate long or short position may allow the greatest flexibility in short-term balancing, but increases exposure to price risk. If the portfolio of contracts contains a mixture of fixed and market-linked prices this will create further price risk.

It is essential to value the portion of the portfolio that is exposed to price risk against pricing scenarios to understand the impact on the bottom line. Hedging strategies can subsequently be identified and their effectiveness measured to protect margins. The trading function must understand the tools and techniques for monitoring risks and know how to apply the various traded products required for hedging. Arbitrage opportunities should also be identified and rapidly exploited.

Speculation is the final stage in the development of a trading function. This involves deliberately maintaining an open position to capitalize on favourable price movements. This is often regarded as gambling, but can provide a successful source of additional profitability if the odds are understood, risks are controlled and the risk/reward profile is clearly set.

Culture is a barrier to the creation of an effective trading function. There is a common misconception that trading equals risk. In reality, effective trading can reduce suppliers’ existing risks, for example, the risks associated with loss of market share.

Balancing: a risky area

For suppliers, the penalties for failing to balance supply and demand present considerable risk. In the period following the implementation of NETA, the costs of being out of balance were considerable. The System Buy price, the price at which a supplier’s shortfall is cleared through the Balancing Mechanism averaged around £60/MWh ($85/MWh). The System Sell price, the price paid to clear suppliers’ surpluses averaged less than £1/MWh. These prices compare with a typical price of around £20/MWh for traded electricity.

Minimizing levels of imbalance is clearly important if suppliers are to avoid being exposed to Balancing Mechanism prices. Except, perhaps, for a supplier with a highly specialized customer portfolio, it is impossible to eliminate balancing risk, since this would require perfect forecasting of customer demand. However, suppliers can make some inroads by improving the quality of their demand forecasts.

A supplier’s forecast error, the difference between its forecast of demand for each half-hourly settlement period and the subsequent actual demand, will fluctuate from one period to the next. For a supplier who is trading to their forecasts, forecast error translates directly into imbalance and its associated costs. Unlike some applications of forecasting such as stock control, an over-forecast in one period is not compensated by an under-forecast in another. All forecast errors result in cost and the greater the error the greater the costs involved.

Typically, a supplier’s forecast error averages around six to eight per cent of its demand in each settlement period. However, the best forecasters are in the two to three per cent range. Based on the average prices quoted, this difference translates into an additional cost that could be as high as £40 million per year for a typical supply company.

Best performance

NETA, like the earlier reforms, required considerable investment and preparation by supply companies. New systems and processes were required for trading and demand forecasting. Interfaces were required with the central NETA systems. Back office systems such as settlement will have needed replacement or major overhaul. Many, no doubt, breathed large sighs of relief when the announcements came, postponing implementation until the end of March. In such an atmosphere, the focus will have been on installing functionality and putting ticks into boxes. The impact of demand forecasting accuracy may well have been overlooked. Companies lacked the benchmarks and, perhaps, the expertise to do anything else.

Now that suppliers have had some experience of operating under NETA, many will be reviewing their performance and making adjustments for the forthcoming winter season. In doing this, suppliers would be well advised to pay attention to demand forecasting and the management of balancing risk.

Unless forecasting performance is at the levels of the best, suppliers will suffer from increased costs, reduced competitiveness and loss of profitability. Forecasting can always be improved and, experience so far indicates that, for many, there are considerable improvements to be made. For some, failure to improve will put survival at risk.

Risk factors

There are many different ways of producing demand forecasts, but all involve two distinct phases. In the first phase, data from the past is used to develop a model relating energy demand to factors such as weather, time of year, time of day and market share. The model is a mathematical equation which, for any given set of factors, calculates an estimate of the demand expected in those conditions. The quality of the model is measured by comparing the expected demands in past conditions with the demands that actually occurred. In the second phase, forecast conditions, particularly weather forecasts, are applied to the equation to calculate forecasts of demand.

When developing models, forecasting teams face a number of choices. Segmenting the market allows different models to be developed for different customer segments. This is a useful technique for dealing with the risk that past demand is not representative of the future because of changes in customer numbers and mix. An extreme form of segmentation involves forecasting the demand of individual customers separately. This is appropriate when the supplier’s portfolio contains very large customers with atypical demand patterns.

There is a wide choice of factors that may be included within the forecasting models. In the UK, temperature is an obvious factor to include for most segments of electricity demand. More difficult is the representation of the ‘misery factor’ that might boost demand on a grey, wet afternoon in November. Factors should always be chosen with care. Throwing additional factors into a model will appear to improve its quality but increases the risk of spurious accuracy.

High quality forecasting requires high quality data; without it forecast errors are increased. Consider, for example, the demand data used in the modelling phase. These are generated from meter readings which, as every utility customer knows, are prone to error. The usual response in a billing application is a corrective transaction. This ensures the customer gets the correct bill. For demand forecasting this compounds the problem: the original error remains in the demand series and the correcting transaction simply adds a further error of equal magnitude and opposite sign.

It will pay to review existing data within forecasting systems and how the data is processed. Particular areas for attention should include consistency of units, BST/GMT clock changes and tracking of the customer base. Errors when detected should be corrected and action taken to prevent recurrence. This may require new data validation and cleansing routines or a tightening of business processes to ensure, for example, that the forecasting system always contains the latest weather forecasts.

Externally supplied data should not be overlooked. Weather forecast data is of vital importance. Periodic analyses to assess the value of new weather variables and alternative weather stations can contribute to improvements in forecast accuracy, particularly if sourcing decisions have been based largely on cost. Contracts with data providers such as meter reading agents may need reviewing to enforce or introduce conditions related to quality and timeliness.

Limiting factors

The settlement processes introduced to support the reform of the retail market, classify customers as half hourly (HH) or non-half hourly (NHH) on the basis of their metering arrangements. HH customers have meters which measure consumption over each 30-minute settlement period. NHH customers have accumulation meters and the consumption of such customers over each settlement period is deemed by calculation. – Depending on their trading strategies and skills, suppliers can gain or lose advantage in the prices they pay for wholesale supplies

One of the most useful factors that can be included in a forecast model is recent demand, particularly demand in the corresponding period of the previous week. Under current processes, the demand deemed for each supplier’s NHH metered customers is not determined until around four weeks after the event. Worse, the attributed value is subject to revision with a final value only being determined some 14 months after the event. These time lags severely limit the value of a very important source of data for demand forecasting. Reducing the time lag so that day-ahead forecasts can be based on last week’s demands should reduce levels of forecast error by at least one percentage point.

Secondly, the deeming calculation involves the use of a scaling factor. The factor reconciles total supply and demand in each settlement period. The factor is applied to each NHH customer’s demand to ensure that the total HH and NHH demands across all suppliers balance to the total generated supply. In effect, the scaling factor corrects imperfections in the deeming calculation. If each supplier’s NHH portfolio is representative of the market as a whole, the effect of this scaling is neutral. However, where suppliers have unrepresentative portfolios, often the case for new entrants, scaling is not neutral. The demands of some suppliers are overstated while others are understated. Particularly damaging is volatility in scaling factors. One supplier has estimated that the abnormally high values experienced in April increased its forecast error by over five per cent of demand, resulting in considerable exposure to the Balancing Mechanism.

The lessons so far

Weather, time of year and other factors effect demand. The dots reflect historic data while the line represents the theoretical model
Click here to enlarge image

Speed, accuracy and quality matter for both trading and forecasting functions. Companies with new trading functions should be considering how to develop these and all companies should be reviewing their demand forecasting activities. Forecasting performance is a key to success, the levels of forecast error should be of as much concern to CEOs as sales revenue and wholesale costs.

There is a need to improve understanding of large customer demand; predictability now matters as well as volume. The new wholesale market has also highlighted the shortcomings of the industry’s settlement processes, which served the retail market so well in the past. The lights have stayed on, NETA works and looks like it is here to stay but can the same be said for all today’s supply companies?