Demand Forecasting

Accurate forecasts of demand are required because:

• Electrical energy cannot yet be stored economically.

• The largest proportion of generating plant is thermal in nature. An unfortunate feature of this plant is the considerable delays involved in preparing the cold generators for connec­tion to the power system (several hours) and the restrictions in the rate at which a steam driven turbogenerator can be loaded after connection. These operational delays are dictated by the thermal/mechanical safety requirements of massive boilers and of turbogenerator sets.

• Thermal generators using steam turbines have an upper limit of power generation equal to their nameplate rating, but also a lower limit dictated by cavitation problems in the turbine blades at low throughputs of steam. Consequently, when a turbogenerator is connected to the network it should be loaded to a level at least equal to the minimum recommended by the manufacturers (from 30 to 50% of rated power).

Figure 3.7 shows that there are periods during the day (e. g. 6 to 7 am) when the rate of demand growth is considerable. To maintain system frequency, the injected power must closely track the trajectory of the demand curve. Unfortunately, because of the sluggishness of the thermal plant, this tracking cannot be done unless preparative action is taken some hours before the event.

It may be concluded that there is an absolute necessity to carry out a demand forecasting activity in order to prepare and progressively load plant as required. Utilities have invested considerable effort in forecasting the daily pattern of demand. Through years of experience they have evolved sophisticated mathematical techniques to correlate demand to the aggregate of the national habits and to other factors such as weather. All methods are essentially based on the fact that demand exhibits regular patterns. Forecasting techniques adjust past demand to present weather and other conditions. Meteorological data on temperature, wind speed, humidity, cloud cover and visibility are used as variables because such factors have an important bearing on heating and lighting demand. The art of load forecasting has been refined to such an extent that estimates are rarely in error by more than ±3% and on average in the UK system they are accurate to within ±1.3%.

Demand prediction techniques are constantly being refined but there will always be occa­sions when unforeseen circumstances increase or depress the load. The average daily errors in demand in a typical month on the English system are shown in Figure 3.8 ; during this period the maximum error in prediction was just under 4% and on average it was less than 1.53%. The standard error during this month was 1.6% which, as the average demand was about 32 GW, corresponds to about 300 MW. The figure shows that on 11 November the forecast was adrift by 3.5%, representing a maximum error of over 1 GW.

Day of month (November 1995)

Source: Electricity Pool Standard deviation: 1.6%

30

Figure 3.8 Typical scheduling errors on the network in England and Wales. (Reproduced from Milborrow, D., ‘Wind power on the grid’, in: Boyle, G. (ed.), Renewable electricity and the grid – the challenge of variability, with permission of Earthscan, 2007)

Figure 3.9 Operational and statutory limits of system frequency for the UK power network. (Source: National Grid plc)

In conclusion, even if all generation were 100% reliable, a substantial reserve would still be required because it is simply not possible to predict the demand on a power system exactly.

Updated: September 26, 2015 — 11:32 am