Let’s recap. In the middle of the day, California’s incredible solar resources generate more power than can be consumed given transmission constraints. Some other generation source is providing power to the homes and business that couldn’t be reached with cheap solar. But what is it?

One possibility is that imports meet that load. The grid is a net exporter of power during these periods, but it’s possible that some areas are importing power and others are exporting. But that doesn’t look like it’s the case. As solar generation picks up and prices fall, CAISO exports more to every one of its trading partners except one. (Only from IID in southeast California do imports increase.) (Rising numbers in the chart below mean more exports from CAISO to that neighboring grid; when the line goes down it means CAISO is importing energy.)

In fact, “exports” might give the wrong impression about what’s actually happening here. When you look at CAISO’s boundary map, a few of the regions are small and within California itself. In fact two of the three biggest balancing authorities that CAISO is exporting to are population centers within California — LADWP (Los Angeles) and BANC (Sacramento).

In summary, the largest force at play appears to be that high solar generation —> low prices —> CAISO exporting to “neighboring” grids, that may just be a population center in California that operates independently.

Above is a typical chart for CAISO so far this year. In the middle of the day, solar is curtailed (hence the bumpy renewables line). At the same time, gas is still being dispatched at a few gigawatts. And even though the grid had been importing power during the nighttime, it switched to exporting power when the sun came up.

I can think of three possible explanations for this chart:

• Resource adequacy rules - maybe the grid operator is requiring gas to be online in case the sun goes down at any moment, and that explains the ~1-3 gigawatts of gas dispatch.

• Transmission congestion - maybe the solar power being generated can’t reach the demand, and that demand needs to be met from other sources.

• Export demand high enough, and prices low enough, that the grid is still a net exporter, even though some imports are being used to meet the demand that would have been otherwise met by curtailed solar.

The second one — transmission congestion — I think is the largest contributor. Late last year, the Energy Information Agency did a great and concise look at this, noting that in the spring of 2023, solar curtailment peaked at nearly 700 GWh in a month. This curtailment is largely — almost exclusively — in the spring.

It’s also growing year over year. It 2019 they curtailed just under 1,000 GWh. By 2023, it was about 2,300 GWh. CAISO classifies curtailment into two categories: congestions and oversupply. Congestion is the biggest and more clear so let’s address that one first.

Here’s just the local congestion. It’s way worse this year.

According to CAISO’s definition, local “congestion occurs when available, least-cost energy cannot be delivered to some loads because transmission facilities do not have sufficient capacity to deliver the energy.”

CAISO doesn’t have a chart that show them together, but we can reconstruct it. The below chart shows the fuel mix for March 13, 2024. Renewables maxes out about 18 GW through the middle of the day. But in the middle of day, multiple GW of power was curtailed, peaking around 5 GWh curtailed around 3pm. The dotted green line represents what would have been dispatched if there had been no local transmission congestion.

The demand that solar was unable to meet was met some other way. But how? It doesn’t appear that CAISO switched on gas generators to meet the demand (at least not this year). Through the middle of the day, gas generation was only about 2 GW (and remember ~5 GW could have been served by solar). Since CAISO’s other forms of supply are not dispatchable — nuclear, and large hydro in particular — that suggests that the demand was met with imports from neighboring grids.

Before we dig more into imports and exports, let’s spend another moment on gas. While so far this year gas seems to peak about 2 GW, last year in May, gas supply did surge in the middle of the day, up to 6 GW at some periods.

It’s possible that we’ll see that again this year. As demand grows as warmer weather arrives, CAISO can get less power from neighboring grids and has to dispatch more gas. Or at least that’s a theory as to what could happen.

In future posts, I’d like to dig more into net exports (I think it’s possible some regions are importing even when the grid as a whole is exporting), and resource adequacy rules. It will also be fascinating to see if CAISO dispatches more gas later in the spring, or if that was a feature of the grid last year but not this year.

Prevention

The first thing to talk about is prevention. Everything would be easier if you could plan the upgrade and not deal with a flooded basement, ruined carpet, no hot water, etc. The water heater was 19 years old. I knew it was old. I didn’t know how old, but I should have. I thought it would give some kind of warning. Brown water or something. It didn’t.

I had two water sensors in the basement for exactly this purpose. One next to the water heater, and one behind the washer about 20 feet away. They are — err, are supposed to — tie into my home automation system which is Home Assistant. The one next to the water heater alerted at 7:30 am. Unfortunately, I didn’t have the alerts set up properly in Home Assistant, so I’m sure it beeped but no one heard it. At 1:30 pm water must have reached the washing machine. The other sensor alerted. My wife got a push notification about a water leak. She texted me asking what it meant. I was in meetings, she was in meetings. She didn’t go down there until 5pm, but the water had reached half way across the room by 1:30. Lesson: get water sensors and make sure they can alert you properly.

Remediation and Insurance

By the time I got home around 10pm, very kind neighbors had helped her turn off water to the water heater, turn off the gas, drain the tank bucket-by-bucket, and clean up a lot of the water. We’re in their debt.

The next morning I called the insurance company and started a claim. (I barely knew who it was but found it on some mortgage docs). There’s a $1,000 deductible, which they said is standard. They have a list of mitigation companies. Apparently this happens a lot. After some very quick googling we picked one off their list and they came that day. Their job is to get rid of the moisture so it doesn’t turn into mold. They thought they could pull back the carpet on the edge and put the fans around the edges. That wasn’t enough. They wound up cutting out half the carpet. Because they had access to both sides of the affected walls, they didn’t feel the need to cut out the wet drywall. They left about 8 fans and 2 dehumidifiers that ran from Thursday until Monday.

I would later find out that insurance won’t cover a new water heater in this situation, which is only fair. We’re still working through it, but it seems like they value the things were ruined (like the carpet), apply a depreciation percentage (for example saying the carpet was 5 years old out of a 10 year lifespan. Also seems fair), and pay out what they call a “recoverable depreciation”.

Plumbers

By this point I knew I wanted a heat pump hot water heater. In fact, I knew it standing in the airport. I haven’t been able to justify ripping out things that were working to make them more efficient. But I certainly had been planning to do efficiency upgrades every time something needed to be replaced. You probably know this by now, but a heat pump water is very efficient because it pulls hot air out of the room and turns it into hot water, running a compressor like an air conditioner or a fridge to do it. It has a resistance electric heater as a backup.

Thursday morning I called every plumber I have worked with previously. None of them could come until the middle or end of the following week. Finally, I called one of the big companies that a friend had used. They would come out the same day.

I also started working my way down the list of plumbers on Rheem’s site. I figured any plumber that went to the trouble of listing there and checking the box for “hybrid water heating” was someone I would like to talk to. I probably made 3 or 4 calls — googling reviews beforehand — before I found an independent company that would come out the next day, Friday.

The plumber from the Big Company showed up. They carry 50 gallon gas hot water heaters on their truck. If I wanted to replace like-for-like, we could have had the hot water back on Thursday. I think my wife had done a little googling on heat pumps and understood my interest and was on board, even if we had to wait longer. “It’s a 15 year decision” I said over and over. Big Company guy quoted 50, 65, and 80 gallon options for gas and hybrid electric. (I’ll get to the $ numbers a little later).

When Independent Company guys showed up the next day, it was a similar conversation. In both cases, they were lightly pushing the gas version. “There’s a real flame down there”. “The hybrid electric is way more expensive. What’s the payback period versus gas? I bet it takes a long time to pay back.” I would mention decarbonizing the economy in my response, but it didn’t seem to resonate. The customer is always right, and they were both willing to do hybrid electric, so it went fine.

Choosing a water heater

I had started googling what model I wanted on Wednesday night. Rheem and A.O. Smith seem to be the two big options for hybrid hot water heaters. Home Depot sells Rheem. Lowe’s sells A.O. Smith.

I have an unusual additional requirement on many home things which is I want them to integrate with Home Assistant natively so I can monitor and control them with everything else in the house. Rheem devices have WiFi and an “EcoNet” app which has an easy existing integration with Home Assistant. For A.O. Smith there are some vague threads where people express interest in reverse engineering the protocol A.O. Smith uses to issue commands to water heaters. Not promising. I decided on Rheem.

Big Company only installed A.O. Smith, which they said about five times they thought were the best. Independent Company had a slight preference for A.O. Smith, but was willing to install whatever I wanted. The Rheem 65 gallon “Performance Platinum” is $2,267 at Home Depot. The “ProTerra”, the next model up which includes a leak sensor and automatic shut off, is $2,491. The A.O. Smith’s are about the same price. The plumbing supply stores, which is where the plumbers shop, charge a couple hundred more.

Unlike the gas hot water heaters, plumbers need to call around and find the right model from their supply stores. When Big Company called me on Friday afternoon to try to close the deal, he didn’t know if or when he could get one. He promised to call me back later Friday afternoon once he found one, but didn’t end up calling me back until Monday, saying he had finally found one and could install it on Wednesday. I had already decided to go with Independent Company.

It took some negotiating, and by this point I had multiple quotes. The price we agreed on (both the unit and installation) was $5,500 for the Rheem 50-gallon or $6,400 for the 65 gallon. I was tempted, and asked, if I could buy it at the plumbing supply store and they could install it for me. They didn’t want to do that. We were on for a Tuesday installation, but I had to have the electrical done.

Electrical

One of the “downsides” of the hybrid electric water heater is it needs electricity, obviously. Whereas the gas line is sitting right there next to the defunct gas heater, I needed a new dedicated circuit for the new hot water heater. We have an old house — built in 1936 — and a finished basement, so it would be a chore to run a new circuit.

Big Company has an electrical division they push on you. Their electrician came on Friday (T+2) to give an estimate. We really liked the guy. He was great. But his quote was $2,000, though he came down to $1,600 as it was clear we weren’t going to bite. It became clear — and I’ve had this experience before — that the big companies pay their guys on commission.

I’ve done a bit of DIY electrical. I love wiring things, it’s just part of who I am. I’ve probably replaced a dozen light switches in the house with smart switches, including one or two three-way switches. And I’ve run a lot of low-voltage stuff, such as ethernet, HDMI, USBs, and coax through an old house. I try to draw the line at messing with the electrical panel itself. Also, my wife wanted a professional.

Friday night I went to Home Depot and got the wire - $300 for 100 feet of 10/3 wire and a 30amp breaker ($25). I got to talking to the guy in the electrical department. His family came from Serbia when he was a kid, and his dad is an electrician. He had confidence I could do it myself, but I got his number and said I might want him and his dad to come and check my work.

Saturday morning I spent 3 or 4 hours fishing the the wire through the ceiling across the basement. It’s a basement so there are some existing access holes. I’ve used the same path for a lot of the low-voltage stuff, so I knew the route. That part went great.

But it’s an old house. The top of the electrical panel was totally full of wires. There was no way to get the wire into the top of the box. I considered cutting holes in the drywall to run the wire down the side and in the bottom of the box. But there is wood everywhere. I was tired so I left the wire dangling in front of the panel and texted my friend from Home Depot. He would come with his dad first thing Monday morning.

When they showed up Monday morning they made quick work of it. His dad, as if by magic, moved the wire to the left side of the box, and in about 1 minute pulled the wire through a knockout hole I hadn’t considered. They were at my house about an hour asked for $180. I rounded up and it’s still the best money I have ever spent.

Install

The water heater install went well. It’s a big machine. Getting it down the stairs was harrowing but ultimately uneventful. They capped off the gas line, and put a cap in the flue.

One other thing to consider: just like an air conditioner, a heat pump generates condensate (water) which needs to find a way out. I have an air conditioner right next to the water heater, and it already has a pump attached. So it wasn’t an issue in my case but many people would need a pump and a way to reach the outside.

It was a full day job. They had three guys who were excellent. Certainly the highlight was connecting it to WiFi. One of the guys asked “what’s ‘M A C address’ mean?” That part I knew.

Takeaways

Here’s are the takeaways on cost:

• $6,400 for a 65 gallon heat pump water heater, installed

• ~$350 in wire and breaker parts

• ~$200 to the electrician

• Probably about $200 in tips along the way

Minus, and this is huge:

• $650 in a D.C. rebate because the plumber is a DC Master Plumber

• $2,000 (30% of the cost up to $2k) in federal tax credit, thanks IRA!

I think it would be easy to spend a lot more. The newest technology, lots of subsidies available, customers who can afford it — it is easy to see how prices could go higher.

And I certainly found the least expensive electric option that wasn’t 100% DIY.

The biggest obstacle to adoption of heat pump water heaters will be the inertia behind replacing a gas water heater with another gas water heater. Six days without a hot shower [at home] for a full family is a long time. The showers wouldn’t come on at all. The dishwasher didn’t work. I was dedicated to the heat pump, and my wife was on board. It’s easy to image choosing the easy road.

emergency situation and the 6 days this took.

1. The electricity used in making hydrogen needs to be generated in the same hour as it’s used/credited for producing hydrogen.

2. The electricity used in making hydrogen needs to be generated locally to the hydrogen plant.

3. And, the electricity used in making the hydrogen needs to be from new renewable resources, capacity have come online since the IRA passed.

Even though smart people feel strongly that the three pillars are necessary, fulfilling all three of these feels onerous. In particular, I wanted to believe that clean energy generated at the same time, in the same place as the hydrogen is green hydrogen. But I’m now convinced the other way: that renewable generation could make the grid look green at any particular hour, while the energy that was dispatched to meet the big hydrogen load is actually from a far away gas plant.

The embedded emissions from a hydrogen electrolyzer (which makes hydrogen from electricity) look something like this:

In this formula, 50 kWh will generate 1 kg of hydrogen using an average electrolyzer. At an emissions rate of .4 kg of CO2 for every kWh, that means 20 kg of CO2 were emitted to generate that 1 kg of hydrogen. As we’ll see, .4 kg CO2 per kWh equates to a fairly efficient gas plant emissions level.

The IRA tax credits for hydrogen production start at $0.60 per kg of hydrogen produced if you can achieve less than 4 kg of CO2 emissions for every kg of hydrogen produced. But the real prize is a subsidy of $3 per kg if emissions fall below 0.45 kg of CO2 for every kg of hydrogen.

In this post I’ll step through an example from recent days in the California CAISO grid and use that to imagine how the tax credit might work. California has been generating buzz for setting solar generation records. According to Gridstatus, renewables’ generation hit records on April 17th at 22.8 GW. Here’s what the CAISO grid looked like on April 17th:

There’s a huge midday peak in solar, and a lot of wind steady around 5 GW. But the second biggest generation technology in the middle of the day is natural gas, which is generating over 5 GW throughout the day. There’s been an active thread on energy Twitter about why this happening. It could be Resource Adequacy rules in CAISO, transmission congestion, how gas plants ramp up, incentives for exports, or maybe something else. It feels important to figure out, but isn’t the point of this post.

At each point in time, the emissions from the grid are the sum of the emissions from each technology. And each technology emits CO2 at a certain average rate per kWh of energy produced. Renewables obviously emit zero CO2 per kWh of energy produced. In California in 2021, gas power plants emitted 0.44 kg of C02 per kWh of energy produced, according to EIA. All of the other CO2-emitting technologies are rounding errors compared to the gas generation.

From this, we can calculate the emissions of the grid at any interval. In the peak renewable generation periods of the day, the grid is emitting about 0.1 kg of CO2 for every kWh of energy produced. With an average electrolyzer using 50 kWh to produce 1 kg of hydrogen, that means during these few hours in the morning, hydrogen producers could make hydrogen at about 5 kg of CO2 emissions for every kg of hydrogen produced. This isn’t enough for even the lowest level production subsidy.

It’s important to note here that even though a hydrogen producer wouldn’t get any subsidies during this day using the grid-wide emissions, they come close, and it’s entirely due to renewables driving down the average. Remember, gas is emitting at 0.44 kg CO2 per kWh but is weighted downwards by the renewables.

But this is the grid-wide average. One of the requirements of this subsidy would be the energy is produced “locally”, though I haven’t heard specific proposals about how how that would be defined. If you divide up a grid like CAISO you are almost certain to find some regions that are doing most of the renewable production compared to others. So it’s easy to imagine that two of the three “pillars” to achieve this subsidy — clean energy at a certain hour, in a certain place — are easily achieved.

The big problem is what type of generation comes online to meet demand when we’ve added hydrogen electrolyzers. Even if a producer can find a location with clean energy in that hour, at that location, the electrolyzer adds loads to the entire grid. So a new electrolyzer adds demand, and the only dispatchable technology to meet the demand — even in the middle of the day — is a gas plant.

It feels onerous at first glance. It feels like if we want to kickstart hydrogen production and maybe have it as long duration storage, we need to be making it easy, not difficult, to get these subsidies. But that’s wrong. The point of the subsidies is to incentivize green hydrogen, and anything other than the three pillars will subsidize dirty hydrogen.

Let’s look at the big picture. Texas generates and consumes a lot of power. While power consumption went up steadily from 2015-2021, 2022 set records. The records aren’t just at peak consumption in 2022 (though it did set those) but across every hour of the day. Here’s how to read this chart: in the top 1% of hours in the year — the highest load 87 hours of the year — the maximum load on the grid during those hours was 80 GW in 2022. In the hours that represent 50% of the year, max load was 47 GW, up 10% from 2021. Every hour of the day had load greater than 33 GW which was up 8% from 2021.

Thanks for reading Mike’s Substack! Subscribe for free to receive new posts and support my work.

Subscribed

Now let’s look back to what happened during Uri. The Texas grid generally sees its peak load about 6pm local time. That was no different on the 14th of February, 2021, though there did appear to be some load shedding or demand response through the peak demand hours. However, things came crashing down after 2am on the 15th of February.

At the time of the meltdown there was 63.9 GW of load on the grid. That’s a lot, especially for 2am, but isn’t an unprecedented load. In fact, load had been 69.7 GW just six hours before at 8pm which would have also been after dark in February. For context, there were 3,312 hours between 2015 and 2022 that had higher load than what brought down the grid during Uri.

As Doug Lewin has shown, the same thing happened again during Winter Storm Elliot right before Christmas 2022. He reports that more than 40% of coal plants were offline at some point during the storm, coal capacity hit a low of 57%, they asked for an emergency declaration from the Department of Energy, and numerous gas plants were offline. History repeated itself.

Getting back to the Texas proposal, the so-called Performance Credit Mechanism would award a subset of dispatchable generators (which would likely be dominated by gas) credits (money) based on their “availability across a set of high reliability risk hours” according to Utility Dive. According to the Chair of the PUCT, the goal is to “guarantee a pot of revenue for reliable resources to go compete for”. The proposed solution to unreliable generation is to give money to the reliable generation. But the problem doesn’t end there.

How Texas defined the “high reliability risk hours” makes a big difference. Let’s assume a high reliability risk hour is anything in the top 10% of 2022 load-hours. That means we’d consider anything above 66.6 GW is a threat to the grid. Using that metric, there have been 1,943 hours above that load since 2015. Not surprisingly, the vast majority of those high load hours are in the summer.

(The original study that proposed the PCM has a more complex high reliability risk definition: the ~30 hours per year with the “lowest additional operating reserves”. And the credit or penalty that would be assessed to generators would be based on generators’ availability to put power into the market during those high risk hours as compared to their accredited capacity. Though the study’s illustration (below) doesn’t have dates, it appears to represent months and support the assessment that these high stress hours would be during the summer.)

Without an even more complicated seasonality component, the PCM would be rewarding the dispatchable generation that was online during the hottest parts of the summer, not the load that stayed online through rare winter weather events.

As the PUC study plainly projects, this market design change would incentivize new gas power plants to come into the market. In that sense, it will likely add overall capacity, albeit dirty and expensive capacity. What it wouldn’t do it is target money for weatherization to protect against grid failure in rare winter events.

If the PUC and Texas legislators want to protect the grid in the face of rare events and surging overall demand, here’s a plan: (1) identify the cause of generators being offline when needed and get that fixed. Then (2), stay ahead of surging demand with incentives for short and long-duration storage, transmission, local generation, and demand response.