Herconfiguratie van elektriciteitsdistributienetten met verspreide energiebronnen

Distribution grid reconfiguration is defined as altering the topological structure of distribution feeders by changing the open/closed states of sectionalizers and tie switches. The objectives achieved by grid reconfiguration typically are loss minimization, load balancing or improved grid reliability. The constraints include grid operation and safety limits. As the distributed energy resource (DER) penetration in a network increases, its variation will affect the power loss incurred by the network along with power quality issues like over voltage. This work addresses this power loss problem under normal operation of the distribution grid. Although grid reconfiguration achieves loss reduction in a distribution network, the switching operation involves cost. Hence, there has to be a balance between the number of switching done in a distribution network and the reduction in loss achieved due to switching. This PhD research uses the concept of grid reconfiguration to minimize power losses and ensure power quality while taking into account DER uncertainty and switching cost. This is achieved by proactively changing the configuration of the network, by taking into consideration the future variations in DERs, in determining the present optimal configuration of the network. In particular, this thesis discusses four major topics: factors affecting grid reconfiguration, methods to find the optimal configuration in the presence of DER uncertainty, methods to ensure robustness of the optimal configuration under DER uncertainty and a quantitative mapping between DER uncertainty indices and grid reconfiguration strategies.First, grid reconfiguration which is formulated as an optimization problem characterized by objectives and constraints is analyzed to identify the various factors that affect its benefits. It is found that the objectives, the constraints, the period of reconfiguration (PRC), the location of components of a network such as load and generation, the availability and location of switches etc. affect the benefits derived out of reconfiguring a network.Grid reconfiguration presents challenging issues due to the non-convex optimization needed because of non-convex objectives and integer constraints. It is conventionally considered a mixed-integer non-linear programming problem. This thesis uses genetic algorithm (GA) for solving single objective reconfiguration problems applied to small networks. Multi-objective reconfiguration problems on small networks are handled with the help of non-dominated sorting genetic algorithm (NSGA-II). For solving the optimization problem involved in reconfiguration of large networks, a deterministic approach based on mixed integer conic programming (MICP) is made use of. This takes advantage of the commercially available solvers and software tools in solving the optimization problem. Following the preliminary study to understand the factors affecting grid reconfiguration, the thesis focuses on identifying and accounting for the effect of photovoltaic (PV) uncertainty in the optimal configuration of the distribution grid. A heuristic method to find the optimal configuration for two days that differ only in their level of PV penetration is developed. Secondly, a scenario analysis based method is developed that helps to account for PV generation uncertainty in grid reconfiguration.Three reconfiguration methods namely Most probable scenario (MPS) method, conservative robust grid reconfiguration (CRGR) method and non-conservative robust grid reconfiguration (NCRGR) based on the concepts of scenario analysis and receding horizon control are developed to find the optimal configuration of the distribution network in the presence of DER uncertainty. Two of the developed methods resulted in robust configuration of the network, that is, the resulting configuration did not violate the constraints under any of the predicted DER variations. One of the developed methods resulted in a non-conservative robust configuration of the distribution network by taking into account the probabilities of all the possible DER variation scenarios.Finally, a general methodology to determine the PRC of a distribution network considering the uncertainty of DERs is developed. PRC is the periodicity at which the configuration of a distribution network is updated. The accuracy of forecast and the sensitivity of the network to the variation in predicted variable were identified as two key factors that influence the PRC. Using the developed methodology, the distribution system operator (DSO) can decide whether an hourly, daily, monthly, seasonal or yearly reconfiguration is to be done for a particular distribution network under a particular forecasting method adopted for predicting the variation of DERs.The following results and conclusions were derived out of the various sections mentioned above. The results show that benefits of grid reconfiguration are dependent on the characteristics of the network under consideration. The optimal configuration of two days that differ only in PV penetration level can be found by evaluating minimum loss reduction parameter. The scenario analysis based method helps in reducing the number of switches to be installed and operated in order to cater for the uncertainty in PV generation. The DSO can adopt MPS method of grid reconfiguration if the probability of the most probable scenario is very high and the probability of worst case scenario is negligible. This helps in extracting maximum benefits out of grid reconfiguration. For a network in which the worst case scenario is having a high probability the DSO can adopt CRGR method which ensures a robust and optimal configuration under the worst case scenario. For all other networks with various predicted DER scenarios and comparable probabilities of occurrence, the DSO can adopt the NCRGR method which guarantees robustness as well as nearly-optimal configuration under all scenarios. A closed loop implementation of the three methods can be adopted instead of the open loop implementation by the DSO if accuracy is paramount. In a network which is less sensitive to a change in the Distributed Generation (DG) output, the DSO can adopt a longer PRC. This low level of sensitivity can be due to the low DG penetration in the network or low value of the DG output. A shorter PRC has to be adopted by the DSO if the forecast is inaccurate or even in the presence of accurate forecast if the network is sensitive to a change in the DG output. In other words, this thesis helps the DSOs, by providing guidelines for adopting grid reconfiguration as an operation tool. To the network planners it helps in deciding the location of switches. This thesis also aims to further the state-of-the-art research in the field, by extending the limits and applications of grid reconfiguration.status: publishe