A Decision-Making Auction Algorithm for Demand Response in Microgrids
M.Sc. Student: Chih-Lun Chang | Advisor: Jimmy C.-H. Peng | Project Duration: 2014 -2016
The need to construct a reliable electric power grid to facilitate renewable generations has led to the notions of distributed generations and microgrids in the last decade. Microgrids are localized energy systems that can disconnect from the traditional grids and operate autonomously to mitigate any network disturbances and strengthen the overall grid resilience. Combined with the renewable technology, the concept of renewable distributed energy resources (DERs) is gaining popularity in microgrid designs (Fig. 1). Popular DERs include solar panels, wind turbines, and energy storage. The resultant microgrids can also participate in the demand response, where the electricity consumers can adjust their generations or loads to support the operating stability of the power utilities during peak periods or in the presence of faults. Those residential houses with DERs can be considered as microgrids, and are the key focus of this work.
Fig. 1. An illustrative example of a distribution system consiting of mulitple VPPs with various types of DERs.
Coming from the power utility perspective, the research is motivated to develop a novel decision-making mechanism for the demand response without the need of a dedicated communication infrastructure. The objective is to move into a decentralized energy management between utilities and their participating virtual power plants (VPPs). Note that a single microgrid can also be regarded as a VPP. The proposed topology will identify a group of VPPs that can supply the requested power to the utility while meeting some designed constraints.As a result, the auction decision process, which is based on the algorithms introduced by Bertsekas, has been incorporated into this work.
To avoid the privacy issues, we assumed the utilities will only know the available surplus power, line impedances, and the present network status of their systems. Furthermore, a distributed communication infrastructure consisting of short ranged ZigBees are used.
Fig. 2. A single-line diagram of IEEE 33-Bus System used in validating the proposed algorithm.
Fig. 3. Visualisation of nodal interactions among VPPs within a power distribution system.
Smart meters are assumed to be connected to residential houses to collect real-time information and coordinate the demand response. Since distribution systems are radial networks in nature (Fig. 2), the demand response is assumed to be a sequential operation. The notion of the distributed communication is built on sending the necessary information over short geographical distances to avoid the use of a centralized communication network. Each VPP will provide its contributions before broadcasting the remaining peak-shaving demands with its neighboring clients (Fig. 3). Such action is similar to the game of "Chinese Whispers", where the demand response information is trickle-down from the utility node to the subsequent connecting nodes. In residential regions, the distance between the participants will mostly fall within a radius of 100 meters. Therefore, ZigBee is recommended as the ideal communication medium to be installed at each house. It adopts the IEEE 802.15.4 standard and can only transmit in a smaller range of 10 to 100 meters. This is sufficient to operate the proposed decision-making mechanism.
Fig. 4. Examples of possible failures within a distribution system: (a) communication failures, and (b) electrical outages.
Table 1: Selection of participating VPP nodes using the proposed decision-making algorithm at various operating conditions. Bold text refers to different selected nodes due to the encountered communication or electrical failures.
In this research, the decision-making algorithm has been developed for demand response. The decision process of each node is based on a limited private information transmitted from the neighboring VPPs through a short-range ZigBee communication. Such approach established a distributed communication infrastructure of which minimizes the risk of privacy intrusion from the utility compared to the conventional centralized implementation. The designed algorithm has been validated in an IEEE 33-Bus System under several operating conditions. Furthermore, both electrical and communication failures have been simulated (Fig. 4. and Table 1). The results demonstrated the proposed decision-making algorithm is able to perform the demand response by meeting the requested power demand from the utility while satisfying the constraints of minimizing the transmission losses. Thus, in summary a comprehensive proof-of-concept analysis has been achieved in this work. The next step is to enhance the ability of the technique to meet the utility power demands while considering dynamic electricity prices. This is part of the ongoing future research.