Section: Application Domains
Probabilistic Algorithms for Renewable Integration in Smart grid
This reserach is developed by the Associate Team PARIS; http://www.di.ens.fr/~busic/PARIS/.
Challenges to Renewable Integration. With greater penetration of renewables, there is a need for tremendous shock absorbers to smooth the volatility of renewable power. An example is the balancing reserves obtained today from fossil-fuel generators, that ramp up and down their power output in response to a command signal from a grid balancing authority - an example of an ancillary service. In the absence of large, expensive batteries, we may have to increase our inventory of responsive fossil-fuel generators, negating the environmental benefits of renewable energy.
The goal of our research is to demonstrate that we do not need to rely entirely on expensive batteries or fast-responding fossil fuel generators to track regulation signals or balancing reserves. There is enormous flexibility in the power consumption of the majority of electric loads. This flexibility can be exploited to create “virtual batteries”. The best example of this is the heating, ventilation, and air conditioning (HVAC) system of a building: There is no perceptible change to the indoor climate if the airflow rate is increased by 10% for 20 minutes, and decreased by 10% for the next 20 minutes. Power consumption deviations follow the airflow deviations closely, but indoor temperature will be essentially constant.
A starting point in our research is the fact that many of the ancillary services needed today are defined by a power deviation reference signal that has zero mean. Examples are PJM’s RegD signal, or BPA’s balancing reserves (BPA balancing authority. Online, http://tinyurl.com/BPAgenload http://tinyurl.com/BPAbalancing .). We have demonstrated that loads can be classified based on the frequency bandwidth of ancillary service that they can offer. If demand response from loads respects these frequency limitations, it is possible to obtain highly reliable ancillary service to the grid, while maintaining strict bounds on the quality of service (QoS) delivered by each load [23] .
Control Design with Local Intelligence at the Loads. An emphasis of our research is the creation of Smart Communities to complement a Smart Grid: intelligence is created at each load in the community. For example, a water heater may be equipped with a simple device that measures the grid frequency – a measure of power mismatch that is regulated to stabilize the power grid. Larger loads may receive a signal from a balancing authority.
A challenge in residential communities is that many loads are either on or off. How can an on/off load track the continuously varying regulation signal broadcast by a grid operator? The answer proposed in our recent work is based on probabilistic algorithms: A single load cannot track a regulation signal such as the balancing reserves. A collection of loads can, provided they are equipped with local control. The value of probabilistic algorithms is that a) they can be designed with minimal communication, b) they avoid synchronization of load responses, and c) it is shown in our recent work that they can be designed to simplify control at the grid level (see the survey [23] and [20] , [30] ). Other researchers have introduced randomization (see in particular the thesis of J. Mathieu [59] ), but without the use of “local intelligence” (distributed control).