Hourly load forecasting using artificial neural networks. Final report
An artificial neural network short-term load forecaster (ANNSTLF) and an artificial neural network (ANN) based temperature forecaster have been developed by Southern Methodist University under contracts RP2473-44 and RP3573-4. ANNSTLF can produce hourly load forecasts for one to 168 hours ahead (one to seven days ahead) with errors ranging from 2 to 4% depending on utility size and characteristics. Implementation of ANNSTLF requires an initial training with historical hourly load and weather data. Two weather parameters, temperature and relative humidity, from either one or multiple locations can be utilized. In the operational phase, the previous day`s load and weather data and hourly weather forecasts are needed. The temperature forecaster can generate hourly temperature forecasts from the predicted values for high and low temperatures of future days. Both forecasters run on a PC platform under the MS-DOS operating system. The development of ANNSTLF is based on decomposition of the load-weather relationship into three distinct trends: Weekly, daily, and hourly. Each trend is modeled by a separate module containing several multi-layer feed-forward ANNs trained by the back-propagation learning rule. The forecasts produced by each module are combined by adaptive filters to arrive at the final forecast. During the forecasting phase, the parameters of the ANNs within each module are adoptively changed according to the latest forecast accuracy. The temperature forecaster consists of a single ANN that requires the previous day`s hourly temperatures and the next day`s predicted high and low temperatures as inputs. The resulting hourly forecasts are adoptively scaled to assure that the high and low temperatures match their respective predictions. The system is capable of forecasting up to seven days ahead. ANNSTLF has been implemented at twenty utilities across the nation and is being used on-Ene by several of them.
- Research Organization:
- Electric Power Research Inst. (EPRI), Palo Alto, CA (United States); Southern Methodist Univ., Dallas, TX (United States). Dept. of Electrical Engineering
- Sponsoring Organization:
- Electric Power Research Inst., Palo Alto, CA (United States)
- OSTI ID:
- 117796
- Report Number(s):
- EPRI-TR-105278
- Resource Relation:
- Other Information: PBD: Sep 1995
- Country of Publication:
- United States
- Language:
- English
Similar Records
Hourly load forecasting by neural networks
An artificial neutral network hourly temperature forecaster with applications in load forecasting