A multi-time-scale power prediction model of hydropower station considering multiple uncertainties

Area of Science:

• Renewable Energy Systems
• Environmental Engineering
• Computational Hydrodynamics

Background:

• Hydropower is a key renewable energy source globally.
• Uncertainties in reservoir inflow and electricity prices impact hydropower output and benefits.
• Accurate power prediction considering these uncertainties is crucial for efficient hydropower operation.

Purpose of the Study:

• To develop a multi-time-scale power prediction model for hydropower stations.
• To incorporate uncertainties from reservoir inflow, electricity price, and consumption rate.
• To enable probability-based decision-making for hydropower generation.

Main Methods:

• A multi-time-scale coupling operation (MCO) model was developed to generate training data.
• A dynamic Bayesian network (DBN) model was established based on expert knowledge and uncertainty relationships.
• A probability-based prediction (PBP) model was used for decision-making and risk assessment.

Main Results:

• The proposed model quantitatively predicts multi-time-scale output power and benefits under uncertainty.
• The model successfully identified risks associated with power generation and output deficiency.
• Application to the Tankeng hydropower station in China demonstrated model efficacy.

Conclusions:

• The dynamic Bayesian network approach effectively predicts hydropower output considering multiple uncertainties.
• The model provides valuable insights into potential generation risks, aiding in operational decision-making.
• This research contributes to more robust and reliable hydropower management strategies.