Dc. Hill et Dg. Infield, MODELED OPERATION OF THE SHETLAND-ISLANDS POWER-SYSTEM COMPARING COMPUTATIONAL AND HUMAN OPERATORS LOAD FORECASTS, IEE proceedings. Generation, transmission and distribution, 142(6), 1995, pp. 555-559
A load forecasting technique, based upon an autoregressive (AR) method
is presented. Its use for short term load forecasting is assessed by
direct comparison with real forecasts made by human operators of the L
erwick power station on the Shetland Islands. A substantial improvemen
t in load prediction, as measured by a reduction of RMS error, is demo
nstrated. Shetland has a total installed capacity of about 68 MW, and
an average load (1990) of around 20 MW. Although the operators could f
orecast the load for a few distinct hours better than the AR method, r
esults from simulations of the scheduling and operation of the generat
ing plant show that the AR forecasts provide increased overall system
performance. A detailed model of the island power system, which includ
es plant scheduling, was run using the AR and Lerwick operators' forec
asts as input to the scheduling routine. A reduction in plant cycling,
underloading and fuel consumption was obtained using the AR forecasts
rather than the operators' forecasts in simulations over a 28 day stu
dy period. It is concluded that the load forecasting method presented
could be of benefit to the operators of such mesoscale power systems.