Sei sulla pagina 1di 2

SINAUT Spectrum

Functional Specification
System Overview

3.12.3 Short Term Load Forecast


☞ Refer to Functional Specification F-SA05.
The principal objective of the Short Term Load Forecast function (STLF) is to calculate the power system load for
the future hours and days. This power system load is the quantity for planning sufficient generation, spinning re-
serve and standby reserve.
Timeliness and accuracy of load forecasts have significant effects on operation and production costs.
Reserve levels are set to mitigate the impacts in the availability of generating units and of the uncertainty in the
load forecast and are defined by the desired level of reliability.
Forecast errors in load predictions can increase operating costs:
■ underprediction:
too small reserve margin may cause failure to provide sufficient cheap reserve
Copyright © Siemens AG 1997-2007. All rights reserved.

■ overprediction:
may cause too many expensive units started up
Power system load is the sum of all individual demands. Each demand or usage pattern is random from the point
of view of the computer system. Due to the unpredictability and the diversity of the individual demands, power sys-
tem load cannot be calculated exactly by extrapolating the estimated individual demand usage patterns. But the
totality of the individual loads results in a distinct consumption pattern which is predicted with statistical methods.
The system load behavior is influenced by:
■ economic factors
– long time trending
■ time factors
– seasonal trending
– day type (Monday, ... Sunday)
– state and religious holidays
■ weather effects
– temperature (the most significant single factor)
– light intensity, number of daylight hours
– wind speed
– other weather factors (humidity, wind direction, ...)
■ special events
– strikes
– shutdown of facilities
– TV programs
■ unexpected events
– load reduction due to failure of network part

F-GD02-E-1.0.0.1 Version 1.0.0.1, 06/07 Page 63


SINAUT Spectrum
Functional Specification
System Overview

Short Term Load Forecast


STLF is carried out by Multiple Regression Analysis, where the historical power system load is assumed to be
dependent upon weather data (linear, quadratic, cubic degree), day types and time.
Further an adaptive regression analysis using a Kalman filter is implemented.
An off-line analysis and prediction based on a particular historical data set decides which regression method will
be activated.
Very Short Term (Tracking) Load Forecast
A moving average model is implemented to minimize the forecast error of the next few hours.
Manual Load Forecast
The operator has several possibilities to make a manual load forecast, such as
■ Single manual entries
Copyright © Siemens AG 1997-2007. All rights reserved.

■ Rescaling of daily forecasts to assumed peak load and energy demand


■ Shifting for percentage or absolute value
■ Weighted copy of forecasts or historical loads
■ Using predefined similar day load patterns (weather sensitive), delivered by customer
■ Using a weather sensitive pattern matching algorithm

Functional Overview
Short term load forecast is basically structured into the following features:
■ correction of historical load
■ rescaling of weather data
■ data analysis
■ automatic load forecast
■ very short term load forecast
■ manual load forecast
■ manual adaptation of the load forecast
■ after the fact error analysis

F-GD02-E-1.0.0.1 Version 1.0.0.1, 06/07 Page 64

Potrebbero piacerti anche