OPTIMAL CONTROL OF A BOILING WATER REACTOR IN LOAD FOLLOWING VIA MULTILEVEL METHODS

A multilevel method is applied to the load following control of a BWR using a nodal reactor model to provide detailed spatial information and with practical operational constraints and thermal limits. Issues of tractability relating to the very large size of the problem are resolved by decomposing the problem through hierarchical control techniques. The optimization of subproblems is performed by using the feasible direction method.;The problem is decomposed along the time axis, and a form of decomposition-coordination method called the non-feasible or goal coodination method is applied. The resultant subproblems are optimized by using the feasible direction method (a nonlinear programming method). A gradient type algorithm is used for coordination.;A simplified model of a boiling water nuclear reactor core containing 4 x 4 assemblies with one control rod at the center is considered. Plant data from the Oyster Creek reactor are used in the calculation, and a daily load following cycle is prescribed. The capacity loss which occurs when power rises is due to redistribution of the xenon concentration. A power ramp rate limitation restricts the speed of the power increase. The deviation of the generated power from the desired is within 0.5 percent of the rated power which is within the accuracy of the nodal calculation.;The analysis uses a strong weighting of the power and control rod deviation terms in the objective functional. The initial Lagrange multiplier and the K value of the gradient type coordination algorithm influences the path of convergence; however, the final control strategy is not affected. The methods described in the dissertation can be applied to a full scale reactor with possible addition of a control rod weighting dependent on position, and a power shape derivation term in the objective function.;An objective function of quadratic form is defined to reflect the control objective, namely to achieve a desired thermal power (tracking) with minimum effort returning to the initial xenon and iodine concentration as closely as possible. Nodal source equations and discretized Xe-I dynamic equations are formulated as equality constraints. The linear heat generation rate and the rate of power increase are formulated as inequality constraints. Core flow and position of control rods are the control variables.