الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 168 |  |  |
| | | from | 168 | | |
المستخلص
Forecasting techniques are evaluated in this thesis for short, medium and long term where good satisfied results can be implemented. Forecasting techniques are provided by using Artificial Neural Network (ANN) System that is the largest applicable type of Artificial Intelligent. Good achievements can be gained as ANN techniques have the robust linear and nonlinear system that has the ability to overcome forecasting problems. For the three types (short, medium and long terms) load forecasting, ANN system is denoted by case dependent since its designs are based mainly on the used data that is gained from Egyptian Electric Holding Company (EEHC). Proposed techniques gain the validity when their results can be examined with real data and compared with similar outputs calculated by Egyptian Electric Holding Company (EEHC) models. All proposed techniques provided in previous chapters, deduces significant comments for studied input data because this data characterizes the ANN structure. More investigation and more training are evaluated by using definite data that may be concluded as the following:
1- Used input data has many problems such as the shortage of the input parameters numbers, influencing on load forecasting, may produce incorrect outputs. Also, the change of daily load curve or annual load curve and sudden hourly peaks, at certain time of working day in summer seasons, make the prediction process became very difficult. Also, unstable trend annual load curve in recent years increased the difficulty of forecasting process where some factors affect on load consumption such that population and consumers traditions.
2- Load forecasting method depends mainly on residential, climatic and geographic environments, tradition and activities of consumers that must be studied accurately just before the predicted intervals. Climatic conditions are the most sensitive parameters that have the most significant parameter on Egyptian load curve shape such as the temperature and humidity weather conditions. Weather data measurements are based on inaccurate programs especially on the predicted intervals when forecasting the load. This may lead to poor outputs.
3- Other parameters must be included to ANN structure to increase the network reliability such that growth rate of consumed load for certain interval, equating the inputs to give new parameters that may advance ANN training
4- Predicted input data is the significant parameter for load forecasting process. Therefore, accurate programs for estimating the predicted parameters must be prepared to gain suitable results. In this thesis, predicted parameters of input data are not available and then, actual data are used as forecasted parameters.