Utility system integration and optimization models for nuclear power management

Abstract

A nuclear power management model suitable for nuclear utility systems optimization has been developed for use in multi-reactor fuel management planning over periods of up to ten years. The overall utility planning model consists of four sub-models: (1) Refueling and Maintenance Model (RAMM), (2) System Integration Model (SIM), (3) System Optimization Model (SOM), and (4) CORE Simulation and Optimization Models (CORSOM's). The SIM and SOM sub-models were developed in this study and are discussed in detail; full-scale computerized versions of each (SYSINT and SYSOPT, respectively) are evaluated as part of the methods development research. The RAMM generates feasible, mutually exclusive nuclear refueling-fossil maintenance schedules. These are evaluated in detail by the rest of the model. Using the Booth-Baleriaux probabilistic utility system model, the SIM integrates the characteristics of the utility's plants into a representation which meets the necessary operating constraints.

Scheduling of system nuclear production and detailed fossil production is done for each time period (few weeks) making up the multi-year planning horizon. Utilizing a network programming model, the SOM optimizes the detailed production schedules of the nuclear units so as to produce the required system nuclear energy at minimum system cost. CORSOM's are utilized to optimize reload parameters (batch size and enrichment) and to generate the individual reactor fuel costs and nuclear incremental costs. These incremental costs are then used by the SOM's iterative gradient optimization technique known as the method of convex combinations. The SYSINT model is shown to be remarkably fast, performing the Booth-Baleriaux simulation for a single time period on a system with over 45 generating units in less than 2.5 seconds on an IBM-370 model 155 computer. SYSOPT converged to optimum solutions in roughly ten iterations.

Immediate reduction of iterations by roughly half is estimated by merely increasing piecewise-linearization of the network objective function. Overall model computational requirements are limited by available CORSOM's, which require 99% of the computational effort (over 3 minutes per reactor per SOM iteration). Nuclear incremental costs (~ 0.8-1.6 $/MWH) are shown to be less than fossil incremental costs (> 2.0 $/MWH) for the foreseeable future. Thus, nuclear power should always be operated so as to supply customer demands with a minimum use of the more expensive fossil energy. For the same reason, the lengthening of nuclear irradiation cycles (in terms of both energy and time) more than pays for itself by reducing the total cost of fossil replacement energy. Idealized nuclear production schedules yield constant nuclear incremental costs regardless of reactor unit and time. One of the key input parameters is the fossil thermal energy cost.