A Computational Framework for Large-Scale Optimisation of Nigeria’s National Water Resources Infrastructure Network

Nigeria's national water resources infrastructure, comprising 339 dams, 12 major river basin development authorities, and thousands of kilometres of irrigation canal networks, is managed without a coherent systems-level optimisation framework, resulting in suboptimal reservoir operation, inter-basin allocation conflicts, and investment inefficiency. This dissertation develops an original large-scale computational optimisation framework for the integrated planning and operation of Nigeria's national water resources infrastructure network. The theoretical contribution is a decomposition methodology for solving the high-dimensional multi-reservoir, multi-objective optimisation problem that is computationally intractable as a monolithic formulation. An original Lagrangian relaxation and Benders decomposition hybrid is developed, with a proved convergence theorem for the proposed iterative algorithm under stated regularity conditions. The multi-objective formulation simultaneously optimises hydropower generation, irrigation water delivery, flood risk reduction, and environmental flow maintenance across all 12 basin authorities. Climate projections from three CORDEX Africa models provide inflow uncertainty scenarios for robust optimisation. The computational framework, implemented on a parallel computing architecture, solves the full national network problem in 4.2 hours versus an estimated 340 hours for the monolithic formulation. Optimal release policies for a representative drought scenario increase national hydropower generation by 18 percent and irrigation service area by 12 percent compared to current operating rules. Keywords: large-scale optimisation, water resources network, Benders decomposition, multi-objective programming, Nigeria.