200 PART | II Smart Supply: Integrating Renewable & Distributed Generation Inverters The use of renewables will almost certainly necessitate the inclusion of inverters, both to convert from direct to alternating currents where necessary and to provide some level of frequency control. The key to the successful integration of inverters into the microgrid is to facilitate parallel operation of inverters serving potentially heterogeneous sources without loss of synchronization, propagation of harmonics, or loss of stability in general. In the case where a small number of inverters are used, this presents a challenging, though well-studied, dilemma. If the inverters use centralized or master-slave control, current sharing is enabled but high-bandwidth links are required to, at the least, facilitate the distribution of error signals [18]. In contrast, distributed, on-board control reduces the bandwidth but at the cost of synchronization difficulties [18]. For larger microgrids with numerous inverters, the available literature is less comprehensive, and doubts remain as to the scalability of techniques pro- posed for small-scale systems. In particular, given the complex interactions that may occur between inverters [18], there is an increasing potential for proble- matic emergent phenomena to appear as the size grows. Predicting and control- ling such behavior are difficult and have not been sufficiently explored in preceding works. Intermittency of Renewable Generation and the Need for Storage Given the intermittency of power supply from wind and solar generators, the obvious response is to add energy storage to intermittent sources, allowing for stored reserves to be accessed when environmental conditions are unfavor- able. The fundamental challenge with such an approach lies in selecting a suit- able amount of storage and in optimally using that storage. For an isolated microgrid, the storage must be sufficient to satisfy gaps between generation and load across both small and large time-scales. Thus, in order to correctly size the storage devices in advance, accurate and reliable behavioral models for intermittent sources are required. The use of such models inevitably raises concerns about quality, and the fidelity of intermittent source models is seldom high. Moreover, even if the models are accurate, selecting storage that minimizes cost while satisfying both short- and long-term commit- ments is a complex optimization task. Dispatch of stored energy is complicated by a similar need to serve both short- and long-term goals. In order to assess which stores should be accessed and at what rate they should be discharged, it will be necessary to implement intelligent control systems that are capable of handling noisy and dynamic data. Such systems must also correctly assess when energy storage devices should charge and how. Beyond the complexity of building such intelligent controls, it is also likely that the controllers will require high-speed communi- cations between microgrid devices in order to capture system state information.