Methods for automated fuel cycle facility deployment

Developing nuclear fuel cycle scenarios requires knowledge of what facilities will be deployed as the simulation evolves. This can be done manually by a researcher, or it can be done computationally. Manual deployment requires a researcher to calculate the deployment of each facility in the nuclear fuel cycle. This can be simple for once-through fuel cycles; however, it is much more difficult for closed fuel cycles. These more advanced fuel cycles have complex interactions between different reactors types. Automated deployment can rely on several methods, from complicated logic to mathematical models (such as time series methods). Time series methods have been used historically in other energy-based utilities to predict solar radiation and wind behavior. By using these methods to predict the quantities and quality of nuclear material, it is possible to determine when nuclear facilities need to be built to match the demands of a nuclear fuel cycle. This work expands upon previous work done using two specific types of time series methods; autoregressive moving average (ARMA) and autoregressive conditional heteroskedasticity (ARCH). Like the previous work, this also utilizes the D3ploy module for the Cyclus Fuel Cycle Simulator to do the automated deployment using predictive time series methods. This new work explores a much wider range of predictive techniques and demonstrates their unique capabilities with regards to the nuclear fuel cycle. These new methods include autoregressive integrated moving average (ARIMA), fast Fourier transforms (FFT), polynomial regression fitting, exponential smoothing, and the Holt Winters method. The implementation of each of these methods will be explained in this work, as well as their advantages and short comings regarding the nuclear fuel cycle.