Wind Power Plant Support in Weather-Dependent Active Distribution Network

Distributed weather-dependent generation connected to medium and low voltage networks creates challenges for distribution network operators due to the generation’s variability, uncertainty, non-dispatchable nature, and real-time unobservability in the low voltage networks. Expensive grid reinforcements are usually required to incorporate weather dependent renewable generation while maintaining acceptable network operating conditions. However, the converter-connected weather-dependent renewable generators are controllable and can support the distribution network operators to strengthen the network operation.

This research focuses on optimizing operation in distribution networks with the help of controllable converter-connected renewable generators as local sources of reactive power and legacy devices such as tap-changers. The effect of optimizing medium voltage network assets on the unobservable low voltage distribution network is analyzed by co-simulating a three-level distribution network in a deterministic and stochastic optimization framework. The impact of weather-dependent uncertainty, variability, and unobservability in the medium and low-voltage distribution network is examined.

Due to the need for distribution network data with a large share of weather-dependent generation, an opensource multi-voltage network model is developed based on the Danish distribution network. The distribution network dataset, DTU 7kBus Active Distribution Network (DTUADN), covers rural, semi-urban, and urban Danish distribution network topologies and provides a year of load and generation time series. The optimization methods developed in this research are tested primarily on DTUADN.

An accurate model for a converter-connected renewable generator’s reactive power capabilities is incorporated in two widely used convex optimization models, e.g. semi-definite programming and secondorder conic programming model. A novel strategic optimization framework is developed to optimize the large-scale multi-voltage distribution network by dividing the network into controllable and uncontrollable zones.

The size of the three-level distribution network and the high penetration of weather-dependent generation induces uncertainty via multiple fronts in the network operation. This research considers three sources of uncertainty: generation forecast, weather-dependent load forecast, and modelling errors for load forecast. Latin hypercube sampling method is deployed for sampling the uncertain variable space with a minimum of 17 stochastic variables. The stochastic variable space is reduced using a scenario reduction algorithm with the Wasserstein metric and Energy distance. The strategic optimization framework optimizes the network operation in the reduced stochastic variable space.

Results from the research affirm the effectiveness of converter-connected renewable generation as local reactive power sources in coordination with the tap-changers. The multi voltage network simulates the pitfalls of controlling medium-voltage assets on unobservable and highly stochastic low-voltage networks. Most importantly, the research presents an in-depth study of a highly weather-dependent multi-voltage distribution network that is partially observable and partly controllable and provides insight into future technical and operational challenges and potential solutions.