Adaptive Neural-Based Fuzzy Inference System and Cooperation Search Algorithm for Simulating and Predicting Discharge Time Series Under Hydropower Reservoir Operation

Reservoir is regarded as one of the most important engineering measures in promoting the high-efficiency utilization of the limited water resources, like water supply, peak operation, power generation and environment protection. Accurate discharge data simulation and prediction information is an essential factor to achieve the expected goals. With the booming development of computer technologies, machine learning is becoming increasingly popular in water resource field. As a classical machine learning approach, adaptive neuro-fuzzy inference system (ANFIS) may fail to effectively capture the nonstationary features of discharge time series in practice. In order to alleviate this problem, this paper develops a hybrid discharge time series simulation method, where the emerging cooperative search algorithm (CSA) is used to find the satisfying parameter combinations of the ANFIS model for the first time. To prove its feasibility and effectiveness, the proposed method is used to simulate multiple-time-scale discharge data of a huge reservoir in China. Based on several statistical indicators, the experiment results indicate that the developed method yields better simulation results than the conventional ANFIS model. Thus, the utilization of swarm intelligence tools can effectively improve the performances of machine learning models in simulating discharge data under hydropower reservoir operation.

Automated summary

Reservoir is an important engineering measure for efficient utilization of water resources, and accurate simulation and prediction of discharge data is crucial. This paper proposes a hybrid simulation method using cooperative search algorithm and adaptive neuro-fuzzy inference system, which shows better performance in simulating reservoir discharge data.