Energy and the environment are probably the most critical and massive problems of our time. The transformation of our energy system into a more sustainable form will take decades, determination, and sacrifices. In the case of power networks, several powerful trends are driving major changes. In this post, we will look at two of them.
The first trend is the accelerating penetration of distributed energy resources (DER) around the world. These DER include photovoltaic (PV) panels, wind turbines, electric vehicles, storage devices, smart appliances, smart buildings, smart inverters, and other power electronics. Their growth is driven by policies and incentive programs. California, for instance, has ambitious policy goals such as:
- Renewable Portfolio Standard (2002): 33% of retail electricity will be procured from renewable sources by 2020.
- Global Warming Solutions Act (2006): Reduce greenhouse gas emission to 1990 level by 2020.
- California Solar Initiative (2007): Offers solar rebates for customers of three CA investor-owned utilities, from 2007 – 2016.
- ZNE homes (2007): All new residential construction will be zero net energy by 2020.
- Energy storage target (2010): The three investor-owned utilities will deploy 1.325 GW of non-hydro storage by 2020.
Leading the world, in terms of percentage share of non-hydro renewable generations (at approximately 20% now), is Germany. Its relentless push for renewables, in the face of technical and financial challenges, will no doubt help find a way forward and benefit us all. See a recent New York Times article, where a proud German reader commented, “And that’s what I love about my country, it is a pain, it causes frustration and malice, but nobody questions the vision.” The question is not whether we should move to a sustainable future, but how we overcome the many challenges on the way (e.g., see Adam’s earlier post about Germany’s challenges), and the earlier we start, the less painful the process will be.
The second trend is the growth of sensors, computing devices, and actuators that are connected to the Internet. Cisco claims that the number of Internet-connected “things” exceeded the number of people on earth in 2008, and, by 2020, the planet will be enveloped in 50 billion such “Internet-of-things.” Just as Internet has grown into a global platform for innovations for cyber systems in the last 20 years, Internet-of-things will become a global platform for innovations in cyber-physical systems. Much data will be generated at network edges. An important implication on computing is that, instead of bringing data across the network to applications in the cloud, we will need to bring applications to data. Distributed analytics and control will be the dominant paradigm in such an environment. This is nicely explained by Michael Enescu (a Caltech alum!) in a recent keynote.
The confluence of these two trends points to a future where there are billions of DER, as well as sensing, computing, communication, and storage devices throughout our electricity infrastructure, from generation to transmission and distribution to end use. Unlike most endpoints today which are merely passive loads, these DER are active endpoints that not only consume, but can also generate, sense, compute, communicate, and actuate. They will create both a severe risk and a tremendous opportunity: a large network of DER introducing rapid, large, frequent, and random fluctuations in power supply and demand, voltage and frequency, and our increased capability to coordinate and optimize their operation in real time.
At Caltech, we are interested in developing an intellectual framework to understand and guide this historic transformation and to address engineering and economic challenges that will arise in the coming decades. There is no shortage of intellectually interesting and practically important problems – if you are starting out your graduate study, you should definitely consider this area.
For instance, a radically different control architecture for future grid is endpoint-based control where each DER self-manages through local sensing and computation, and communicates with its neighbors. They make local decisions based on their own states, and possibly information from their neighbors. The global behavior that emerges from the interaction of these local algorithms must be stable, robust, efficient, and above all, understandable.
An example of this approach in the context of load-side participation in frequency regulation is discussed in my previous blog post. Though frequency regulation has been mainly implemented on the generator side, there are three important advantages to ubiquitous continuous and distributed load-side participation:
- Unlike spinning reserve, load-side control does not waste fuel or produce extra emission.
- The ubiquity of load-side control can better localize disturbance and produce more reliable response through the law of large numbers. The paper “Achieving controllability of electric loads” by Callaway and Hiskens provides an example where 60,000 air conditioners can be controlled to track the output of a wind farm accurately without deviating from the desired temperature by more than 0.15 degree:
and an example where 20,000 electric vehicles can be controlled to track demand profile accurately without impacting their charging requirements:
- Finally, since loads have low or no inertia, they can respond fast. The simulation in our recent CDC2014 paper “Optimal decentralized primary frequency control in power networks” demonstrates that load-side participation can significantly improve both the steady-state and transient behavior over generator-only control: