Stochastic dynamic optimal power flow under the variability of renewable energy with modern heuristic optimization techniques.
Access rightsWorldwide access
Access changed 4/3/18
Bai, Wenlei, 1987-
MetadataShow full item record
With the increasing penetration of renewable energy to power systems, such as wind power, more challenges have been brought to system operations due to the intermittent nature of wind. Such influence can be reflected on ancillary services of systems such as frequency control, scheduling and dispatch, and operating reserves. To tackle those challenges, wind power forecast has become an important tool. Nowadays, forecasters typically have access to information scattered through a huge number of observed wind power time-series data from a large number of wind farms. However traditional multivariate time-series models can only process small number of data and capture only the temporal correlation in wind. In this work we utilized a probabilistic forecast model, dynamic factor model (DFM), to predict wind power. The DFM is able to capture both the spatial and temporal correlation of data, and generate as many scenarios as possible to represent the uncertainty of wind power forecast. This work also focuses on the optimization of the system integrated with wind power and storage devices over 24 hours. Thus we formulate such problem as a stochastic dynamic optimal power flow (DOPF) problem. The essence of solving stochastic problem is to make a decision that performs well on average under almost all possible scenarios. In all, the objective functions are to optimize the expected value over all scenarios generated by DFM. Once the stochastic optimization problem is formulated, a proper methodology is required to solve the problem. Static optimal power flow (OPF) is a highly non-linear, mixed-integer, non-convex and non-smooth problem, and traditional techniques such as nonlinear programming, quadratic programming, interior point method simplifies the problem which sacrifices the accuracy of the solution, and fails to consider the non-smooth, non-differentiable and non-convex objective functions. Therefore, to circumvent these downsides we proposed a novel heuristic method called artificial bee colony (ABC) to tackle the static OPF without approximation. In this study, the ABC has been tested on small, medium and large power system for OPF (IEEE-30, IEEE-57 and IEEE-118 buses) and then it was modified and extended to solve a dynamic optimization problem recursively.