A N-dimensional Stochastic Control Algorithm for Electricity Asset Management on PC cluster and Blue Gene Supercomputer

Management of French electricity production to control cost while satisfying demand, leads to solve a stochastic optimization problem where the main sources of uncertainty are the demand load, the electricity and fuel market prices, the hydraulicity, and the availability of the thermal production assets. A stochastic dynamic programming method is an interesting solution, but is both CPU and memory consuming. It requires parallelization to achieve speedup and size up, and to deal with a big number of stocks (N) and a big number of uncertainty factors. This paper introduces a distribution of a N-dimension stochastic dynamic programming application, on PC clusters and IBM Blue Gene/L supercomputer. It has needed to parallelize input and output file accesses from thousands of processors, to load balance a N-dimension cube of data and computation evolving at each time step, and to compute Monte-Carlo simulations requiring data spread in many separate files managed by different processors. Finally, a successful experiment of a 7-stock problem using up to 8192 processors validates this distribution strategy. 1 Main objectives and challenges EDF (Electricité de France) is an electricity producer that needs to manage its production assets in order to satisfy the consumer demand. It leads to a stochastic optimization problem where the main sources of uncertainty are the demand load, the electricity and fuel market prices, the hydraulicity, and the availability of the thermal production assets. The goal of the optimization is to save energetic resources and money by managing the production cost while satisfying demand. The main levers of this optimization are the commands that can be used to manage the different stocks involved in the problem and the thermal production levels. The necessity to deal with a lot of stocks (mainly water stocks used by hydraulic power plant, and fuel stocks) gives a very big optimization problem. Due to the non convexity of the problem, the stochastic dynamic programming method is a method of choice used to find a solution. Unfortunately, it is unable to deal with a big number of stocks and a big number of uncertainty factors. The goal of this work is to be able to deal with at least three or four uncertainty factors, and at least six or seven stocks in optimization while being able to efficiently use in simulation the commands calculated. ha l-0 02 91 81 4, v er si on 1 30 J un 2 00 8 Author manuscript, published in "PARA 2008, Trondheim : Norvège (2008)"