AN INTELLIGENT DECISION SUPPORT FRAMEWORK FOR FLOOD FORECASTING AND STRUCTURAL SAFETY MONITORING IN HYDROPOWER SYSTEMS: INTEGRATION WITH POWER SYSTEM PROTECTION AND CONTROL

Abstract

Most hydropower stations still run flood forecasting and dam safety monitoring as two separate operations, with no shared decision logic between them — even though a hydrological risk can become a structural risk within hours. This paper proposes a four-layer protection and control framework — Data Acquisition, Hydrological Forecasting, Structural Risk Assessment, and Application/Control — that couples hydrological forecast uncertainty with physical dam integrity in a single SCADA-interfaced decision pipeline.

A calibrated HBV (Hydrologiska Byråns Vattenbalansavdelning) rainfall-runoff model is coupled with an LSTM network for residual error correction, producing 6–72 hour streamflow forecasts. A Fuzzy-Bayesian engine then maps these forecasts together with concurrent structural signals — seepage, deformation, and reservoir level — onto a four-tier risk scheme consistent with SL 314-2018, with risk-level triggers passed to SCADA/EMS platforms over IEC 61850.

Applied to the Kunhar River Basin in Pakistan, the HBV-LSTM model reaches NSE = 0.91 and RMSE = 14.8 m³/s, a 16.7-percentage-point NSE gain over standalone HBV, with peak-flow error falling from 14.1% to 7.2%. Under the RCP8.5 scenario, the 100-year design flood rises from 6,240 to 8,150 m³/s, pushing the baseline risk classification from Level II to Level III in 38% of ensemble members. The 6–12 hour forecast horizon enables advance generation dispatch, cutting spillage losses by 5%–8%, while Level IV emergency commands execute within 100 ms via GOOSE messaging.

The framework offers a standards-compliant pathway for coordinating hydropower flood response with power-system protection, and the modular four-layer design should transfer to other climate-sensitive basins without requiring a full SCADA rebuild.