Advertising Console

    FORSE : Finance Optimization Research on Solar Electricity - PPE 2012

    Reposter
    63 vues

    2 commentaires

    ECE Paris Ecole d'Ingénieur
    In this paper, we present an optimization model to make profitable a molten salt solar power plant, by choosing optimal periods of electricity delivery. The problem of provisioning and dynamically optimizing such industrial projects requires implementing numerical methods that yield approximate solutions. The core of the problem is to solve a Hamilton-Jacobi-Bellman (HJB) partial derivatives equation in an infinite horizon dynamic optimal control problem, combined with a stochastic representation of solar incoming energy intensity.
    ECE Paris Ecole d'Ingénieur
    The system behavior is modeled by two sets of equations. The first set links the plant’s specifications: the molten salt tank capacity, the total surface of mirrors, the power generator and their unit costs. The second set links the variables modeling sunlight random intensity, the price of the KWh of a given period with the molten salt tank level as a time dependent control variable u(t). Solving HJB will compute the optimal u*(k) level through a dynamic backward induction algorithm that combines the sunlight intensity with the expected u(k) level at the end of period k estimated from the one on period k+1.

    The application of this technique, inspired from American derivative option pricing, can be extended to other areas where control problems are exposed to time dependent sets of price and return. It is a work-around for the curse of dimensionality in HJB problems.