Research On Affine Scaling Algorithm And Its Application To Optimal Schedule Of Hydro System

More and more hydro utilities have focused on improvement of operational potentialbecause of affluent economic benefit derived from optimal reservoir dispatch. On the otherhand, some characteristics in specific hydropower plants often hinder usage of optimizationsoftware developed for dispatch. To develop a practical and reliable model and algorithmfor scheduling is still a challenging task. The dissertation investigates optimal short-termand deterministic long-term dispatch models and algorithms against the state key project—The Three Gorges Digital Cascades Decision Support System and The Three GorgesCascaded Scheduling Automatic Generation Control. Some valuable results obtained areoutlined as follows: (1) The principle and implementation technique of affine scaling algorithm (ASA) areintroduced in detail and its iterative characteristics are described. The way to choose a stepof iteration is also introduced, which guarantees the monotonic decrease in objectivefunction value. Frequently computing an inverse matrix in ASA may lead to overriddencomputational time. Therefore, QR factorization is used to translate the solution of inversematrix into tackling a set of linear equations. Initial interior point for iteration is computedby the big M method. Cases about three kinds of solution in linear programming areanalyzed respectively and the results show that ASA is effective. (2) The implementation strategies for two linearization methods are researched. Insuccessive linear approximation, the dissertation employs bound control parameters tosafeguard the boundary of linear programming and proposes an efficient method ofparameter correction. A way of variable expansion ensures the compatibility betweenapproximated programming and ASA. Chaos optimization algorithm based on linear searchis introduced to determine an initial interior point for nonlinear programming, whichaccelerates the convergence of algorithm. In piecewise linear approximation, the promisingapplication of two modes in separable programming and intrinsic limitations are discussed.Detailed solution to a case is provided so that the nuance of the two modes can bediscriminated. (3) The short-term operational characteristics of cascaded hydroelectric system areinvestigated. A modeling of optimal short-term scheduling is based on these characteristics.According to the actual conditions, the dissertation makes three assumptions that enable IIIsuccessive linear programming to settle the scheduling model. Factors in the first orderTailor series expansion are derived and an equation of linearly system is constructed. Thehydropower plant cost functions and reservoir features are approximated by piecewiselinear function. Successive linear approximation simplifies the size of original dispatchproblem by a transformation for bi-directional equations of power output. (4) The successive linear programming and piecewise linear approximation algorithmis used to optimize short-term scheduling of Three Gorges cascades, respectively. Theinitializations both use the cost function with nominal head. Their difference is thatcontinuous function is adopted in successive linear programming while piecewiseapproximation employs the two piecewise linear segment function. Simulations performedreveal that using the cost function with nominal head to calculate initial point increasesconvergence of the model. Especially initial point obtained by this kind of method isusually close to optimum considerably in successive linear programming. Calculation speedin two algorithms shows that ASA is applicable and reliable. (5) A long term scheduling model in Three Gorges cascades is building by maximizingthe stored potential energy at the end of planning horizon. The principle of decompositionand coordination for the large-scale system is applied to relax the coupling constraint ofgeneration. Numeric test is conducted with actual hydra...