Environmental targets set by governments around the world are leading to high penetrations of small to medium-scale renewable distributed generation (DG). High penetration of DG in distribution networks, however, can result in voltage and thermal issues among other technical problems. The traditional 'Fit & Forget' approach that refers to the passive use of assets with limited or no control, in the context of distribution network planning, is used to meet maximum demand or generation requirements. However, to ensure that more renewable generation is cost-effectively connected to distribution networks, it is imperative to adopt a more active control of network elements and participants. The active control of future distribution networks requires understanding the corresponding dependencies between voltage magnitudes and DG active/reactive power outputs to mitigate voltage issues. One classical method to calculate these dependencies is to use sensitivity approaches such as those based on the Jacobian matrix. However, during operation, updating the Jacobian matrix requires the network to be fully observable making it unfeasible for decentralized control approaches. Therefore, it is critical to develop a sensitivity approach only requiring local real-time information. This thesis proposes a novel approach to produce voltage sensitivity coefficients using the surface fitting technique based solely on knowledge of network characteristics and, therefore, no remote monitoring is required. To assess the performance of the proposed voltage sensitivity approach, a decentralized (local) voltage control algorithm that simultaneously caters for both the active and reactive power outputs of a single DG plant is adopted. Comparisons with classical sensitivity approaches are carried out using the 16-bus UK GDS test network, 1-min resolution demand and wind generation data. Persistence forecasting (i.e., assuming no changes in demand and wind in a short time period) is considered in this case. The lower Mean Squared Error (MSE) shows that the coefficients of the proposed sensitivity approach are close to those of the Jacobian matrix and better than the perturb-and-observe approach. In the context of voltage management, results highlight that the proposed sensitivity approach is more effective than the Jacobian matrix inverse and perturb-and-observe, resulting in better voltage compliance and energy harvesting (better capacity factor). It should be highlighted that this performance is achieved without the need of full network observability. Furthermore, to cater for the more realistic and complex case of multiple DG plants, this thesis proposes a time-delay based decentralized control algorithm. A comparison with an ideal AC Optimal Power Flow (OPF) is carried out using the same 16-bus UK GDS network but with seven DG plants. The results demonstrate that the proposed sensitivity approach and time delays are very effective when compared to the AC OPF. This, in turn, proves that the combined use of the proposed voltage sensitivity approach and the decentralized controller is an implementable, cost-effective solution to manage DG plants in distribution networks without the need of further communication infrastructure. Finally, a decentralized DG control logic with the capability of using wind forecasting techniques is proposed to tackle the unpredictable nature of wind power. In this work, a time-series based forecasting technique is incorporated to the proposed decentralized controller. The results confirm that the use of more advanced forecasting technique can further improve the management of renewable DG plants.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:728064 |
| Date | January 2017 |
| Creators | Zhang, Zedong |
| Contributors | Ochoa, Luis |
| Publisher | University of Manchester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://www.research.manchester.ac.uk/portal/en/theses/decentralized-control-of-distributed-generation-in-future-distribution-networks(aaffa96a-284f-4b04-a528-ee2bd98dba59).html |
Page generated in 0.0019 seconds