A Two-Stage Swarm Optimizer With Local Search for Water Distribution Network Optimization.

Area of Science:

• Engineering
• Computer Science
• Optimization

Background:

• Evolutionary computation (EC) algorithms are effective for small-scale water distribution network (WDN) optimization.
• Rapid city expansion leads to large-scale WDNs, diminishing the efficacy of current EC algorithms.
• Large-scale WDN optimization presents significant computational challenges due to network size and multimodal search spaces.

Purpose of the Study:

• To propose an effective optimization algorithm for large-scale water distribution networks (WDNs).
• To address the limitations of existing EC algorithms in handling complex and large-scale WDN optimization problems.
• To introduce a novel two-stage swarm optimizer with local search (TSOL) tailored for WDN optimization.

Main Methods:

• A two-stage optimization process is employed, dividing the problem into exploration and exploitation stages.
• An improved level-based learning optimizer is utilized for efficient search in both stages.
• Two novel local search algorithms are incorporated to refine solution quality.

Main Results:

• The proposed Two-Stage Swarm Optimizer with Local Search (TSOL) demonstrated superior performance on synthetic and real-world WDNs.
• TSOL effectively navigated the large-scale and multimodal search spaces characteristic of expanded WDNs.
• Experimental results indicate that TSOL outperforms existing state-of-the-art metaheuristic algorithms.

Conclusions:

• The TSOL algorithm offers a robust and effective solution for optimizing large-scale water distribution networks.
• The two-stage approach combined with enhanced learning and local search significantly improves optimization efficacy.
• This research provides a valuable advancement in applying computational intelligence to critical infrastructure optimization.