Security-Constrained Stochastic Economic Dispatch With Wind Farm And Pumped-storage Station Integration

Recently, wind power generation, as a kind of renewable energy source, is more and more integrated to the power grid to cope with the problem of energy depletion and environmental pollution. However, the randomness of wind power generation has made the optimization of dispatch difficult. Hydro pumped storage station is a kind of vast storage in the power grid and it has the advantages of mature technology, low operating cost and fast response, which can help a lot to reduce the passive influence of the randomness on the power system and make full use of renewable energy. This dissertation focuses on the model and algorithm to solve the problem of security-constrained stochastic economic dispatch with wind farm and pumped-storage station integration.For stochastic economic dispatch problem with single wind farm and single pumped storage station, by treating the outputs of wind farm as random variable and the storage quantity of pumped storage station as storage variable, a simple stochastic storage model considering the upper/lower limits and ramping up/down limits of generating units is constructed.There exist expectation computation in the above model, and the state space and decision space of actual problems are high dimensional, which leads to the curse of dimensionality and intractability of accurate solution. Approximate dynamic programming(ADP) is adopted to get the near optimal solution by approximate the value function using piecewise linear function. Firstly, based on the forecast outputs of wind farm, sufficient error scenarios are generated by Monte Carlo sampling according to a certain probability distribution. Then, approximate value functions are properly initialized according to the information inside the power system. The value functions are trained until convergence by scanning the error scenarios with SPAR(successive projective approximation routine) algorithm, which is intentionally proposed for stochastic storage model. Finally, the near optimal day-ahead scheme for stochastic economic dispatch could be obtained by solving the model expressed by the converged approximate value function. Moreover, the approximate value function can be used as a policy to help make decisions for stochastic economic dispatch under error scenarios rather than the forecast scenario.The transmission lines have their transmitting limits in the process of operation due to thermal factor, which should be considered in economic dispatch. Therefore, each randomness source and storage device plays different part in the power grid. They must be considered separately rather than easily combined. What’s more, transition constraints from the forecast scenario to the sampling scenarios could be considered while dispatching. Model and algo-rithm are properly improved in accordance with these situations, to make them applicable to security-constrained stochastic economic dispatch with multiple wind farms and pumped storage integration.The results on test system and real provincial system show that, the model constructed and the algorithm selected in this dissertation are effective to obtain near optimal solution to the question, which shows practical value.