Impact on Balancing

Thermal plant incur some costs when they are run in the frequency control mode. It can be shown, [ 5, 9-11] that if variations in a variable source and in demand occur roughly

Подпись: +Подпись: (3.2)

Impact on Balancing Подпись: ' Total variability^2 of variable v source у

independently, the total resulting variation in the net load to be met by the thermal plant is approximately a ‘sum-of-squares’ addition of the components:

Thus, for example, when the average power variation of the added source equals that of the demand itself, the total variability is not doubled but increased by 40%. This has some important implications. The impact of fluctuations in variable sources at low penetrations can be taken to be practically zero; in other words this impact is just noise added to demand fluctuations.

Over longer timescales, the level of operating reserve required at any given time depends on two key factors: uncertainties in demand prediction and the probability of loss of the largest generation plant on the network. When wind power plant is introduced into the system, an additional source of variation is added to the already variable nature of demand. To analyse the additional variation caused by the wind plant it is important to appreciate that the require­ment is that the entire system must be balanced instead of balancing each individual load or resource. The operator has to ensure that the average system reliability is maintained at the same level it would have been without the wind resource.

Подпись:Подпись: (3.3)

Impact on Balancing Подпись: ' Average error in ^2 predicting variable v input у

However, the crucial question is by how much does wind generation increase the balancing uncertainties? Intuitively it is known that minute-to-minute fluctuations in wind output are largely uncorrelated to load. This implies that the additional uncertainty introduced by wind power does not add tinearly to the uncertainty of predicting the load. As for the issue of variability dealt with above, it can be shown that when errors in predicting the output from variable sources occurs independently of those in predicting demand, the combined error is again a sum-of-squares addition [9-12]:

Demand prediction techniques are constantly being refined but there will always be occa­sions when unforeseen circumstances push up or depress the load. Equation (3.3) indicates that for small penetrations of variable sources the prediction errors are lost among load fluc­tuations. However, since demand is fairly predictable, forecasting errors from substantial penetration of wind will incur some penalty.

Analysis of the combined uncertainties of wind, demand and conventional generation based on the sum-of-squares calculation of Equation (3.2) make use of the standard error in predicting the generation/demand balance. On typical developed country networks, one hour ahead, this averages at around 1% of the demand. For four hours ahead, this figure rises to 3%.

10

Impact on Balancing

Wind capacity/peak demand %

□ Persistence ■ Perfect

Figure 3.15 Additional balancing power for wind (Published in Windpower Monthly News Magazine, December 2003 [12])

The need to schedule reserve to cover for possible trips of conventional thermal plant emphazises the point that no generation is 100% reliable. While the loss of a typical 1000 MW of thermal plant is a real risk, it is almost inconceivable that 1000 MW of wind plant would be suddenly lost. It is also assumed that dispersed wind plant is not sensitive to a common mode disturbance; i. e. the plant rides through voltage dips caused by faults on the transmis­sion system. The more wind that is installed, the more widely it is spread, and sudden changes of wind output across a whole country simply will not in practice occur [13].

Calculations can be made over various timescales to determine the need for extra reserve. Figure 3.15 shows the estimated additional balancing power needed (expressed as a percent­age of installed wind power) as a function of wind power penetration [12]. At 20% penetra­tion, 7% of extra operating reserves are required if persistence[9] in the wind is assumed. For perfect forecasting only 2% of additional capacity would be needed.

Clearly these back-up figures are modest, but what about the associated costs? In the UK, the TNO has estimated that 10% of wind penetration would increase balancing costs by £40 m a year, which is equivalent to £0.002/kWh. A 20% penetration will increase the cost to £0.003/kWh. These figures should be viewed in terms of the retail cost of electricity, which at the time of writing is in the region of £0.10/kWh [14] . Hence a 20% penetration in the UK would incur a 3% additional cost on electricity at present prices.

For relatively little expenditure the predictability of wind could be greatly improved. This could be accomplished partly through the installation of extra weather data monitoring sta­tions (e. g. anemometry towers a few tens of kilometres from major wind farms) and partly through sophisticated computational techniques. Programmes to provide enhanced predict­ability are being developed in several countries [15].

Figure 3.16 illustrates this by showing a typical one hour wind forecast using sophisticated techniques against actual output for one wind farm over a period of a week. Such techniques could provide considerable cost benefits in operating reserve.

Impact on Balancing

Figure 3.16 Wind farm forecast (+1 hour) versus actual output, Ireland, 2004. (Reproduced with permission of Garrad Hassan and Partners Ltd)

Updated: September 27, 2015 — 5:06 am