The End of the Steampunk Grid

I just finished reading Alexandra von Meier’s book, Electric Power Systems: A Conceptual Introduction. It’s an overview of how the generation, transmission, and distribution system works, and how it’s worked for pretty much the whole history of the grid, stretching back to the end of the 19th century. More than anything I came away with an appreciation for the gloriously analog nature of the machine. We have a steampunk grid, a massive artifact of the Victorian era, hiding behind and powering our increasingly digital world. This isn’t an engineering textbook, but it’s not exactly meant for a popular audience either. There’s an ongoing stream of complex numbers, calculus references, vectors, matrices, and electromagnetic fields… and without some understanding of them, a lot of the core ideas in the book will probably not come across very well.

At the upstream edge of the grid, we have thousands of gigantic machines, spinning in almost perfect synchronization. Massive amounts of iron and copper, literally turned by steam. They’ve gotten bigger and hotter and more precise and efficient over time, but they’re fundamentally the same type of generation the grid grew up with a century ago.

At the downstream edge of the grid, in large part we have the same kind of machines… but running in reverse, taking the undulating waves of electricity, and turning them back into rotation, through an invisible, smoothly spinning force-field. It’s like magic, but it’s something we’ve all lived with our entire lives. It’s so normal we don’t think about it.

Between these spinning machines we have masses of iron and tightly wound copper stepping voltage up and down, mechanical switches that look like something out of Frankenstein, and very little in the way of instrumentation and automation — at least by present day standards. And with a few exceptions, the electricity really does flow from one edge of the grid to the other in a dendritic network.

A Digital Grid is Coming

A lot of this is about to change. The end of the Steampunk Grid is within sight. We have sinusoidal three-phase power because we make electricity with giant spinning dynamos, and we use a lot of that power to run motors, which are just dynamos running in reverse. The rotational inertia of the generators is the energy storage mechanism of the grid, regulating voltage and frequency at the microsecond scale. But once it’s on the transmission system, all you can see is the phase and magnitude of voltage and current, in three channels for three phase power — there’s no information left about where those waveforms came from, and they obey Kerchoff’s laws whether we like it or not.

Spinning dynamos are designed to naturally generate AC power. Time varying AC power is nice because it can be stepped up to high voltage with transformers and transmitted long distances efficiently — line losses mostly come from current, and the same power can be transmitted with high voltage and low current as with low voltage and high current.

Inverters take low voltage DC power and turn it into time varying AC They can be built to synthesize any waveform you want in real time. Solar PV and battery storage put out DC, so they have to use inverters to connect to the AC grid. Wind turbines make AC power, but they spin at variable speeds, depending on the wind, so they can’t typically be connected directly to the AC grid. Today they mostly use inverters to take that variable frequency AC power and turn it into DC, and then back into grid-synchronized AC power, allowing them to operate at a wide variety of speeds. Large solid state power electronics can also be inserted into the transmission system explicitly for the purpose of re-shaping and phase shifting the power waveforms, removing long-standing stability constraints in transmission systems, and avoiding the delicate synchronicity that the AC grid maintains today through analog means.

This means that a large wind farm or battery bank or PV installation can make power that looks like almost any kind of power station, as the demands of the transmission and distribution systems change over time. If the devices become cheap enough, maybe they could also exist at the interfaces between individual consumers or microgrids or distribution circuits, and the rest of the grid, and subtly condition power to increase grid efficiency and stability. Normal electricity consumers don’t often think about these “ancillary services” that go beyond raw kWh of energy, but grid operators do. Constantly. This is a subtle but large change in the nature of the grid.

Of course, there are other big changes happening now too, that seem to be getting more attention, maybe because they’re easier to communicate without using complex numbers and matrix calculus. At least, until reading this book I certainly didn’t appreciate just how much of the grid’s physical infrastructure is fundamentally dedicated to and designed around managing the AC waveform with purely analog systems on a massive scale.

Now my sense is that the digitization of electricity is similar in importance to, say:

• The advent of cheap, utility scale battery storage.
• The dominance of electricity sources with close to zero marginal production costs (solar, wind, enhanced geothermal, maybe nuclear someday).
• The shift away from controlling electricity supply and toward managing electricity demand.
• The re-organization of the distribution system’s topology into a densely interconnected network, rather than a tree-like structure with clear upstream and downstream ends.
• The ability to collect and process detailed information about the grid’s operational state in close to real time.

In a lot of ways these epochal shifts complement each other. Detailed real-time information about the grid is necessary if you want to adjust voltage and current components of the system on the timescale of a single 60Hz cycle. The valuation of demand flexibility and power quality can substitute for the raw $/kWh energy valuation we currently rely on, creating a market that isn’t broken by massive amounts of zero marginal cost power. These value streams can also enable the demand management strategies that facilitate the integration of very large amounts of variable, forecastable renewable power. Safe and reliable operation of a multi-directional distribution and generation network, rather than a unidirectional centralized generation to transmission to consumption network will also require much more fine grained information about the state of the grid. Localized and semi-autonomous provisioning of voltage regulation by inverters and other digital power devices can make the grid more modular and flexible and able to accommodate more kinds of interconnected participants with less centralized oversight.

Rather than requiring generous gross supply margins — which was the only way we could be sure the system would work for most of the last century — we will be able to create other kinds of systemic robustness. We’ll have a detailed picture of the grid’s behavior and dynamics, especially in retrospect, and the system will learn from its own experiences. Von Meier’s current work focuses on collecting high volume data about power distribution systems, measuring current and voltage 120 times a second, at thousands of points in the system, and then distilling that massive data stream into something usable by a grid operator

Countries that are still undergoing their first wave of mass electrification today may never see a completely analog grid built out. In Sub-Saharan Africa, India, and other parts of Asia, maybe they’ll leapfrog directly to a more dynamic, digital electricity system, as they have with wireless telephone infrastructure. Without the trillions of dollars worth of sunk costs and legacy systems and legacy expertise that we have to worry about, maybe it will actually be easier for them to build that grid from scratch.

Human Factors

Interestingly, the last chapter of von Meier’s book is about the human part of the power system: the engineers that design it, and the operators that run it day to day. These groups have very different cultures, and ways of relating to the system. Engineers are much more quantitative, while operators are more intuitive. Engineers prize efficiency, and operators focus on robustness. Big engineering changes are coming, and von Meier highlights the need and difficulty of the required operational changes. In many cases it sounds like the conservative operational culture makes it hard to integrate new technology into the grid.

Ironically, the operational culture, valuing a robust, safe, reliable system above all else, and cultivating a do-no-harm mentality, sounds like exactly the kind of culture we need more of at the scale of our civilization, in relation to the Earth’s climate, and basic planetary operations. It reminds me of the optimality vs. robustness debate that plagues decision making under deep uncertainty.

As the systems we rely on get more complex and autonomous, our understanding of how they work is going to become fuzzier. We may know how they function in general, intuitive terms, or empirically, but they’ll become more subtle and less deterministic. More qualitative and less quantitative (from our point of view), despite the fact that they will generate and consume vast quantities of data. We won’t understand everything about them. They may do unexpected things. Already if we aren’t careful neural networks sometimes learn how to cheat.

The grid is on its way to becoming a seething, self-aware system with reflexes that can respond automatically in a sixtieth of a second to re-route power, shed loads, engage battery backups, shift consumption patterns in time and space, and deal with variable power sources spread across whole continents. We’re just at the beginning of understanding what this kind of power system will look like, and how it will be organized and operated, but it seems clear it will be very different than the grid built by Edison and Tesla that we all still live with today.