Long Term Load Forecasting for sustainable Electricity Market and Regulation in Jordan

The analysis was performed under three scenarios: high-forecasted (optimistic case), mediumforecasted, and low-forecasted (pessimistic case), to account for varying economic and population growth trajectories. To achieve this, multiple forecasting methods were employed, including statistical models such as Multiple Linear Regression (MLR) and Exponential

Smoothing (ES), alongside Artificial Neural Networks (ANN). The ANN methods included

Feed Forward Back Propagation Neural Networks (FFBPNN), Elman Neural Network (ENN), and Layer Recurrent Neural Network (LRNN). Historical data used in the analysis encompassed electricity demand, Gross Domestic Product (GDP), and oil prices, obtained from national statistical authorities and publicly available international oil price indices. The proposed methodology was implemented and evaluated using standard numerical simulation and statistical analysis tools.

To evaluate the performance and accuracy of the forecasting models, the performance indices such as Mean Absolute Percentage Error (MAPE), Normalized Root Mean Square Error

(NRMSE), and Mean Absolute Error (MAE) were calculated during the validation phase for both peak load (MW) and energy generation data (GWh). For peak load data, the MAPE and

MAE values were 4.15% and 93.4 MW for the MLR method, 6.54% and 111 MW for ES, 3.51% and 77.7 MW for FFBP, 3.52% and 78.1 MW for ENN, and 3.58% and 78.2 MW for

LRNN. Similarly, for energy generation data , the MAPE and NRMSE values were 4.351% and 0.0482 for the MLR method, 6.12% and 0.0589 for ES, 2.94% and 0.0403 for FFBP, 2.72% and 0.0389 for ENN, and 2.87% and 0.0402 for LRNN. These results indicate that neural network models achieve higher accuracy than statistical models. Specifically, the FFBPNN model was selected for peak load forecasting and the ENN model for energy generation forecasting, as they achieved the lowest errors during the validation phase after the training phase.

These findings provide valuable insights into optimizing long-term load forecasting for Jordan, offering critical guidance to policymakers and engineers. They are expected to support energy resource planning, infrastructure development, and efficient allocation of resources over the next 25 years. This study also reinforces the foundation for sustainable electricity market planning in Jordan through accurate long-term forecasting. It enables strategic investment decisions, ensures resource adequacy, and supports non-market regulatory approaches. By minimizing uncertainty and guiding future infrastructure development, the results contribute directly to a more resilient and future-ready power system.