IWRA Proceedings

< Return to abstract list

A LSTMs AI Based Model to Forecast Outflow-Inflow From-To SMBA Reservoir Rabat-Morocco - Medium and Long Run Forecasting

IWRA World Water Congress 2025 Marrakech Morocco
Towards Innovation and a Smart Water Future
Author(s): Dr. Mustapha Hajji - IEA-ONEE
Oral: PDF

Abstract

Context and Problem Statement

Critical Resource: Reservoirs are vital for shortage prevention and flood mitigation in the Rabat-Casablanca corridor.

Challenge: Climate change and drought frequency create massive hydrological uncertainties.

Gap: Physical models effective for short-term (1 month) but fail at medium-to-long term strategic planning.

Objective: Build a dynamic Deep Learning (LSTM) model to forecast Inflow (Nature) and Outflow (Demand).


Conclusions

No single model fits all variables. A robust system should use Bi-LSTM for Inflow (nature) and SARIMA for Outflow (demand).

Wavelet denoising is critical for improving the accuracy of noisy inflow data before feeding it into AI models