Uncertainty-Aware Transactive Operation Decisions for Grid-Friendly Building Clusters
MetadataShow full item record
In this thesis, the emerging local energy transaction of prosumer (building) at distribution level is focused. Decision models and efficient algorithms are developed to study the collaborative energy transaction decisions of building clusters in three research phases. Number of research has demonstrated that building clusters can achieve more benefits like lower total energy cost, however, some buildings have to make sacrifices of their own interests for collective interests for the clusters. To motivate individual buildings, we propose four different transactive energy management models in first phase where each building is allowed to have energy transaction with others while individual requirement has to be satisfied. The first model focuses on maximizing collective interests and this model is appropriate when all the buildings are operated by one manager, both collective and individual interests are considered in second model which is suitable when different buildings have heterogeneous individual interests. The third and fourth models aim to maximize both collective and individual interests, this two models are preferred when buildings have homogeneous individual interests (e.g. same saving percentage or absolute saving amount). Then next, in second phase, large scale (e.g. community level) building clusters is studied. To enable more efficient transactive operations among prosumers, we propose a swarm intelligence based bi-level distributed decision approach. Particle swarm optimizer is employed at system level to coordinate all the buildings to dispatch shared energy while each building at sub-system level will employ a mixed integer operating model to obtain operation decisions for its energy systems, such as distributed generators and storage systems. For the purpose of accelerating convergence of swarm algorithm, a marginal price based feedback strategy is proposed. During each iteration, each building will solve its local decision model, the marginal prices for exchanged energy will be collected and fed back to system level to guide velocity and position updating of particle swarm. Proposed distributed approach is applied on distributed control for building-charging station integration as a case study, and then it is evaluated in terms of accuracy, scalability and robustness. It is demonstrated that proposed approach is very computationally efficient, scalable and robust, and the computational complexity if O(n) where n is the number of buildings in the cluster. To deal with uncertain information about electricity load and solar radiation, scenario-based centralized two-stage stochastic operation model is firstly established at third research phase, where electric storage and power generating unit are assumed to provide different kinds of operating reserves in ancillary market. Proposed swarm intelligence based distributed decision framework and coordination algorithm from previous phase are extended to incorporate with stochastic programming. In order to further decrease model complexity of planning optimization and utilize updated information, model prediction control approach is embedded in proposed energy transaction process to make online decisions. In summary, this thesis has proposed a swarm intelligence based methodology of coordinating buildings' transactive operation at distribution level. The main idea is to utilize marginal information from individual optimization to allocate resources more effectively for collective optimality. This methodology could be adopted for more applications, such as robots swarm coordination, etc. There are, however, several issues that could be addressed in future investigations. For example, only electricity transaction is allowed in research phase II and III, multiple transacted energy resources (heating, cooling and electricity) will be considered, and the correlation between different kinds of energy resources will be emphasized. In addition, the energy transaction price of local transaction market is assumed based on transparent information in research phase I. Pricing negotiation mechanism will be worth developed based on game theory to optimally determine local energy transaction price. More broadly, from system perspective, uncertainty coupling and propagation from different sources may have great impacts on the algorithm performance, also communication between system level and subsystem agents may be delayed and missing, therefore distributed coordination algorithm should be robust when facing with such unexpected conditions in practice.
Marginal Price Guided Particle Swarm
Model Predictive Control