One Month Ahead Temperature Prediction for Mid-term Load Forecasting

Temperature is the most important factor in load forecasting of power system analysis. Particularly in short-term load forecasting, it plays an important role in increase/decrease of energy consumption. For instance, according to the Japanese Power Industry announcement in 2001, increase of 1 degree Celsius will cause about 5GW increase in electric power consumption at the summer peak. This amount is as same as consumption power used by 1.6 million general households or amount of generated power by 5 large-scale utility power plants. On the other side, 1 degree Celsius of temperature change had caused about 1.85GW increase in power consumption in the winter of year 2000. Basically, the short-term temperature (hourly up to 1 week ahead) is researched and predicted by environment agencies of every country. Therefore, it is easy to obtain the forecasted temperature data from those agencies, newspapers, TV news and so on. However, it is difficult to obtain the hourly temperature beyond 1 week. Although Japan Meteorological Agency (JMA), which uses the Numerical Weather Predictions (NWP), announces the forecasting data up to 1 or 2 months ahead, but they are expressed only as “high” or “low” which is compared with normal years. This means, we can only know that the temperature may goes up or comes down every day. In addition, super-computer processes it with lots of complex meteorological formulations. The applied data the ones which have observed by weather satellite all over the world. However, if the temperature could be predicted for a longer period, it becomes even a useful factor for projecting a better resolution for the long-term load forecasting, prediction of fuel amount necessary for next couple months of power plants and soforth. In this paper, some intelligent methodologies such as artificial neural network and a combined neuro-genetic algorithm have been used to predict the temperature up to one month ahead

کلیدواژه ها:

Weather Conditions ، Temperature Prediction ، Mid-Term Load Forecasting ، Artificial Neural Network (ANN) ، Neuro-Genetic Learnin