Electricity is an increasingly complex industry in the midst of transition to renewables and decarbonization. Using my 25 years’ experience as an engineer, policy analyst, and academic, I help my consulting clients think through their toughest technical challenges and formulate their best business strategies.

Posted in consulting | Tagged | Leave a comment

Is demand side the way to go??

Because wind and solar production depends on weather conditions, it is subject to the variability and intermittency of weather. The challenge of renewable integration is to cope with the resulting variability of the “net load,” or total load minus intermittent renewable production. (Click here to read my blog post on the duck curve.)

Chemical batteries are often proposed as a solution for variability, and they do seem to be roughly competitive at wholesale in the supply of ancillary services (services additional to energy that are required for reliable operation of the electricity system). Chemical batteries are cost-competitive for the ancillary service of frequency regulation and possibly for managing contingencies. Frequency regulation and contingency services maintain supply-demand balance in the very-short-term to the short-term (in the seconds-to-tens-of-minutes time scale). 

However, broadly speaking, chemical batteries are not currently cost-competitive for storage of renewable electricity to match supply and demand at longer time scales of hours to weeks, given current wholesale electricity prices in most systems worldwide. Although chemical battery storage may eventually be cost-effective for medium-term storage, current costs imply that stored electricity would be much more expensive (including the wholesale cost of the electricity and the cost of storage) than the wholesale cost of electricity alone. Until (and if) costs come down by considerable factors, cost-effective progress toward very high levels of renewables must face the medium-term variability of renewable production with only a relatively small amount of chemical battery storage.  

There is also a challenge at even longer time scales as well. In particular, seasonal imbalances between renewable supply and electrical demand over weeks, months, and years are unlikely to be solved by chemical storage or any demand-side adaptations—we need to think about other ways to compensate for seasonal mismatches. Countries and regions with pumped-storage hydro and reservoirs with multi-year capacity will clearly be better able to utilize seasonal excess energy from renewables than regions without large, long-term storage.

That said, I now want to focus on the medium-term and to emphasize demand-side options. There are multiple, partially overlapping ways to help match renewable production and demand in the medium-term by harnessing or increasing flexibility to respond to renewable variability. While some of these actions are on the supply side, I want to focus here on demand-side actions.

Previously I have discussed one such demand-side option: AC pre-cooling to help align electrical consumption to the general pattern of solar production by taking advantage of storing “cool” at the end-use. (Click here to read the blog post.) This helps to adapt consumption to what could be called the “temporal endowment” of renewables, which, for solar, is during the hours in the middle of the solar day. Also, I have explored water pumping aligned to solar production (click here to read the blog post) and electric vehicle charging aligned to wind production in Texas, which is mostly overnight (click here to read the blog post).

RossBaldick.com
Air separation plants can potentially adjust their consumption to help compensate for renewable fluctuations. Credit: Image provided by courtesy of Linde GmbH, Germany, 2020.

Another promising approach to demand-side management comes from UT chemical engineering professor Michael Baldea and his graduate students, including Morgan Kelley, Joannah Otashu, and Richard Pattison, with whom I’ve collaborated. The team has demonstrated flexibility of electrical consumption in industrial chemical processes through storing end-use products. In this research [see, for example, Morgan T. Kelley, Ross Baldick, and Michael Baldea, “Demand response operation of electricity-intensive chemical processes for reduced greenhouse gas emissions: Application to an air separation unit,” ACS Sustainable Chemistry & Engineering, 7(2): 1909-1922, February 2019], a case study considers the cryogenic separation of air into its constituent gases (in this case, nitrogen, but also oxygen and argon can be taken into consideration). This process is electrically intensive, but these products can be stored conveniently in liquefied form. Conceptually, this allows us to increase the production rate of the separation process (often referred to as an “air separation unit” or ASU, and illustrated above) in time intervals when, for example, renewable production is high, wholesale electricity prices are low, or marginal emissions from electricity production are low.

Sign up for Ross’s blog!

To implement this concept, however, how do we deal with the fact that price signals in wholesale markets are updated hourly or sub-hourly? And with the fact that renewable production and emissions can fluctuate at the sub-hourly level, while key variables of the chemical process have dynamics that can be longer than hourly? We need to consider the chemical process dynamics and the limits on the values of process variables, such as pressure and temperature and the chemical composition and purity of the products, when adjusting consumption to follow prices. What Professor Baldea and his students show us is how to represent the salient issues of the complex dynamics of the process variables into a demand-response model that can respond to fluctuating prices or renewable production while also maintaining process variables within allowed ranges. 

How would this flexibility be utilized in an electricity market such as ERCOT, which has both a day-ahead and a real-time market? To participate in the day-ahead market, the air separation facility could first roughly forecast when wholesale prices are expected to be lowest in the coming day. Currently in most electricity markets, this is typically overnight and on weekends, but as solar penetration increases, hours in the middle of the day will also likely have low prices, as is already happening in California. Given a daily target for air separation products and the required electrical energy over the day, the plant could bid to consume electricity during the lowest price hours in the coming day, bearing in mind limitations on how fast it can change production from hour to hour. This would lock in day-ahead prices for consumption targeted to the lowest priced hours: if the plant follows in real-time its position from the day-ahead market, then it will pay exactly the amount determined in the day-ahead market. Because the consumption is focused on low-priced hours, the total energy bill would be lower than, for example, producing the nitrogen, oxygen, and argon at a constant rate throughout the day.

In real-time, electricity prices can vary due to unexpected outages and other events, including renewable fluctuations, providing an opportunity to further benefit from swings in prices. In a market such as ERCOT, with a price cap of $9,000/MWh, the swings can be considerable. If a market participant deviates in real-time from its day-ahead position, then it effectively purchases or sells that deviation at the real-time price. Naturally, it would be advantageous to sell when real-time prices are unexpectedly higher (effectively requiring that the plant have some additional product stored to satisfy its customers for such an event when it is actually producing less than planned).  

Historically, few loads have participated in real-time markets in ERCOT, because the requirements for dispatchability by ERCOT and other market operators have not been attractive. Depending on the circumstances, an air separation plant might have the ability to satisfy these requirements and bid into the real-time market. However, even if it is not attractive to fully participate by bidding into the real-time market, the plant can still take advantage of wholesale price fluctuations by simply adjusting its consumption compared to the day-ahead position. Again, such deviations require some additional storage capacity for the end product. Effectively, the air separation plant can provide flexibility to the electricity market by having additional storage and additional stored product that enables it to cut its production when the price is right.

The air separation plant is a particularly good candidate for providing such flexibility, but many other industrial processes may also be able to adapt to varying conditions and thereby help with accommodating high levels of renewables. Historically, many industrial processes have agreed to be interrupted in emergencies in return for lower electricity rates. Today, the market allows them a much more direct way to profit from their flexibility.

In the future, we may also see new uses for electricity that are inherently more flexible than traditional consumption, or that lead to other ways to store the electricity in an end product or energy carrier. Such futuristic applications include the electrolytic production of hydrogen, the production of chlorine (for example, see: Joannah I. Otashu and Michael Baldea, “Demand response-oriented dynamic modeling and operational optimization of membrane-based chlor-alkali plants,” Computers & Chemical Engineering, 121: 396-408, February 2019), or the production of ammonia. These approaches add to the growing list of ways to use demand-side flexibility to accommodate high levels of renewables.

Posted in research | Leave a comment

Is demand-side the way to go?

Because wind and solar production depends on weather conditions, it is subject to the variability and intermittency of weather. The challenge of renewable integration is to cope with the resulting variability of the “net load,” or total load minus intermittent renewable production. (Click here to read my blog post on the duck curve.)

Chemical batteries are often proposed as a solution for variability, and they do seem to be roughly competitive at wholesale in the supply of ancillary services (services additional to energy that are required for reliable operation of the electricity system). Chemical batteries are cost-competitive for the ancillary service of frequency regulation and possibly for managing contingencies. Frequency regulation and contingency services maintain supply-demand balance in the very-short-term to the short-term (in the seconds-to-tens-of-minutes time scale). 

However, broadly speaking, chemical batteries are not currently cost-competitive for storage of renewable electricity to match supply and demand at longer time scales of hours to weeks, given current wholesale electricity prices in most systems worldwide. Although chemical battery storage may eventually be cost-effective for medium-term storage, current costs imply that stored electricity would be much more expensive (including the wholesale cost of the electricity and the cost of storage) than the wholesale cost of electricity alone. Until (and if) costs come down by considerable factors, cost-effective progress toward very high levels of renewables must face the medium-term variability of renewable production with only a relatively small amount of chemical battery storage.  

There is also a challenge at even longer time scales as well. In particular, seasonal imbalances between renewable supply and electrical demand over weeks, months, and years are unlikely to be solved by chemical storage or any demand-side adaptations—we need to think about other ways to compensate for seasonal mismatches. Countries and regions with pumped-storage hydro and reservoirs with multi-year capacity will clearly be better able to utilize seasonal excess energy from renewables than regions without large, long-term storage.

That said, I now want to focus on the medium-term and to emphasize demand-side options. There are multiple, partially overlapping ways to help match renewable production and demand in the medium-term by harnessing or increasing flexibility to respond to renewable variability. While some of these actions are on the supply side, I want to focus here on demand-side actions.

Previously I have discussed one such demand-side option: AC pre-cooling to help align electrical consumption to the general pattern of solar production by taking advantage of storing “cool” at the end-use. (Click here to read the blog post.) This helps to adapt consumption to what could be called the “temporal endowment” of renewables, which, for solar, is during the hours in the middle of the solar day. Also, I have explored water pumping aligned to solar production (click here to read the blog post) and electric vehicle charging aligned to wind production in Texas, which is mostly overnight (click here to read the blog post).

RossBaldick.com
Air separation plants can potentially adjust their consumption to help compensate for renewable fluctuations. Credit: Image provided by courtesy of Linde GmbH, Germany, 2020.

Another promising approach to demand-side management comes from UT chemical engineering professor Michael Baldea and his graduate students, including Morgan Kelley, Joannah Otashu, and Richard Pattison, with whom I’ve collaborated. The team has demonstrated flexibility of electrical consumption in industrial chemical processes through storing end-use products. In this research [see, for example, Morgan T. Kelley, Ross Baldick, and Michael Baldea, “Demand response operation of electricity-intensive chemical processes for reduced greenhouse gas emissions: Application to an air separation unit,” ACS Sustainable Chemistry & Engineering, 7(2): 1909-1922, February 2019], a case study considers the cryogenic separation of air into its constituent gases (in this case, nitrogen, but also oxygen and argon can be taken into consideration). This process is electrically intensive, but these products can be stored conveniently in liquefied form. Conceptually, this allows us to increase the production rate of the separation process (often referred to as an “air separation unit” or ASU, and illustrated above) in time intervals when, for example, renewable production is high, wholesale electricity prices are low, or marginal emissions from electricity production are low.

Sign up for Ross’s blog!

To implement this concept, however, how do we deal with the fact that price signals in wholesale markets are updated hourly or sub-hourly? And with the fact that renewable production and emissions can fluctuate at the sub-hourly level, while key variables of the chemical process have dynamics that can be longer than hourly? We need to consider the chemical process dynamics and the limits on the values of process variables, such as pressure and temperature and the chemical composition and purity of the products, when adjusting consumption to follow prices. What Professor Baldea and his students show us is how to represent the salient issues of the complex dynamics of the process variables into a demand-response model that can respond to fluctuating prices or renewable production while also maintaining process variables within allowed ranges. 

How would this flexibility be utilized in an electricity market such as ERCOT, which has both a day-ahead and a real-time market? To participate in the day-ahead market, the air separation facility could first roughly forecast when wholesale prices are expected to be lowest in the coming day. Currently in most electricity markets, this is typically overnight and on weekends, but as solar penetration increases, hours in the middle of the day will also likely have low prices, as is already happening in California. Given a daily target for air separation products and the required electrical energy over the day, the plant could bid to consume electricity during the lowest price hours in the coming day, bearing in mind limitations on how fast it can change production from hour to hour. This would lock in day-ahead prices for consumption targeted to the lowest priced hours: if the plant follows in real-time its position from the day-ahead market, then it will pay exactly the amount determined in the day-ahead market. Because the consumption is focused on low-priced hours, the total energy bill would be lower than, for example, producing the nitrogen, oxygen, and argon at a constant rate throughout the day.

In real-time, electricity prices can vary due to unexpected outages and other events, including renewable fluctuations, providing an opportunity to further benefit from swings in prices. In a market such as ERCOT, with a price cap of $9,000/MWh, the swings can be considerable. If a market participant deviates in real-time from its day-ahead position, then it effectively purchases or sells that deviation at the real-time price. Naturally, it would be advantageous to sell when real-time prices are unexpectedly higher (effectively requiring that the plant have some additional product stored to satisfy its customers for such an event when it is actually producing less than planned).  

Historically, few loads have participated in real-time markets in ERCOT, because the requirements for dispatchability by ERCOT and other market operators have not been attractive. Depending on the circumstances, an air separation plant might have the ability to satisfy these requirements and bid into the real-time market. However, even if it is not attractive to fully participate by bidding into the real-time market, the plant can still take advantage of wholesale price fluctuations by simply adjusting its consumption compared to the day-ahead position. Again, such deviations require some additional storage capacity for the end product. Effectively, the air separation plant can provide flexibility to the electricity market by having additional storage and additional stored product that enables it to cut its production when the price is right.

The air separation plant is a particularly good candidate for providing such flexibility, but many other industrial processes may also be able to adapt to varying conditions and thereby help with accommodating high levels of renewables. Historically, many industrial processes have agreed to be interrupted in emergencies in return for lower electricity rates. Today, the market allows them a much more direct way to profit from their flexibility.

In the future, we may also see new uses for electricity that are inherently more flexible than traditional consumption, or that lead to other ways to store the electricity in an end product or energy carrier. Such futuristic applications include the electrolytic production of hydrogen, the production of chlorine (for example, see: Joannah I. Otashu and Michael Baldea, “Demand response-oriented dynamic modeling and operational optimization of membrane-based chlor-alkali plants,” Computers & Chemical Engineering, 121: 396-408, February 2019), or the production of ammonia. These approaches add to the growing list of ways to use demand-side flexibility to accommodate high levels of renewables.

Posted in research | Tagged , , , , , , , | 2 Comments

There’s more than one way to deal with a duck (curve)

The good news is: solar. The bad news is: uncontrollable rooftop solar. How do we utilize the production of rooftop solar in the middle of the day?

This is an especially significant problem for California. The California ISO coined the term “duck curve” to describe that state’s net load – that is, load minus renewable production from solar photovoltaic (PV), wind, and run-of-river hydro production. On a mild sunny day, the problem that the duck curve illustrates is that net load falls so low that other generation cannot follow it.

Sign up for Ross’s blog!

The same challenge is emerging in other places, as well. In Australia, for example, Paul Simshauser of Griffith University, South East Queensland, describes the same situation there in his warm, northern state. (Click here to download his paper.)

RossBaldick.com

BTW, I would suggest we might replace “duck curve” with “magpie goose curve” in honor of an Australian native bird. (Photo credit: Fir0002/Flagstaffotos)

Interestingly, Simshauser’s data shows that air-conditioning (AC) is a significant contribution to the rising neck of the magpie goose during critical summer days (that is, a significant contribution to the increase in net load during the evening of such days after the sun goes down). Although I do not have the specific data on hour-to-hour consumption for inland California, I believe that a similar pattern would apply there and in many regions with high AC consumption. In Texas, for example, residential consumption in summer greatly exceeds that in spring and fall. It is well understood that summer peak consumption is driven by AC load during times of high temperatures that persist into evenings.

RossBaldick.com
Figure 1: Average net load (kW) of homes with AC and solar and net load with pre-cooling. (Ross Baldick)

To illustrate, Figure 1 shows net load for Central Texas during August 2018 (data courtesy of Grant Fisher and Esha Choudhary of Pecan Street Inc.). The blue line shows net load (consumption, including AC, minus rooftop PV production) for a sample of Austin homes. The data points are 15-minute average power consumption, averaged over the sample of homes with both AC and rooftop PV that Pecan Street monitors. Each day, net load falls (becoming negative, implying net export to the electric distribution system) during the day, but then rises again and experiences a peak in the late afternoon and evening, with solar production decreasing just as AC consumption increases: the upward sloping “neck” of the duck or magpie goose.

This combination of AC needs that persist after the sun goes down and PV production that falls precipitously at sundown suggests a way forward: to pre-cool houses in the hours before sundown. Researchers working under the U.S. Department of Energy’s Building America program have modeled two example homes for several regions using EnergyPlus. (Click here to download the report.) They found that pre-cooling is an effective way to reduce peak residential load.

However, pre-cooling will result in higher energy consumption by 2% to 8% overall. This increase in energy consumption can be thought of as analogous to “round-trip losses” in a battery storage system, implying that storing energy involves overall more energy consumption than using the energy when it is produced. Results depend on both weather patterns and the thermal insulation and thermal mass of the housing stock.

How does pre-cooling interact with the duck curve?  Pre-cooling can increase consumption when the sun is shining and decrease it after the sun goes down. This has two advantages.  First, with significant rooftop solar, there is significant export during the day to the electric distribution system. There are limits to the amount of such exports, and California is heading toward a situation where PV production may otherwise have to be curtailed during the day. Therefore, pre-cooling could offer significant benefits by increasing utilization of renewables during the day, while also reducing non-renewable production in the evening. Second, pre-cooling will reduce the “ramp rate” of net load; that is, the rate of increase in net load over time, which is represented by the upward sloping neck of the duck. Because net load must be matched by other generation, and because generation has limited ability to ramp, reducing the slope of the “neck” can ease the need for ramping capacity.

Sometimes chemical battery storage is advocated as a solution to the mismatch between PV production and electrical demand. Interestingly, the higher energy consumption with pre-cooling found by the DOE Building America program is similar in magnitude to the round trip losses of a Tesla battery. In contrast to a chemical battery, pre-cooling does not require (much) capital investment, at least for a well-insulated home. While pre-cooling might not work for typical current Queensland housing stock, it might be effective in regions where there is already significant investment in insulation. Much housing stock in Texas, for example, has double-glazing as well as ceiling and wall insulation, and further investments in building efficiency would not only help with improving prospects for energy storage but also pay dividends in overall energy savings. I understand from Scott Jarman of Austin Energy that this Austin utility already practices pre-cooling in some of its controlled residential thermostats in preparation for critical peaks.

So, could we pre-cool all residences all the time? Pre-cooling homes could effectively be practiced more widely and not just on critical peak days. The idea would be to significantly pre-cool well-insulated homes while PVs were still producing significant power, and then to allow indoor temperatures to drift upward as the sun goes down. This would facilitate better utilization of PV production and reduce the slope of net load in the evening.

I have not performed the detailed modeling to evaluate the potential explicitly, but figure 1 suggests what might be possible for Austin. I considered shifting the AC consumption represented in the Pecan Street data to occur three hours earlier. I accounted for the round-trip losses by assuming that 10% more electricity for AC would be required when shifting consumption by three hours. The result is shown in the orange line, which has less variation than the blue line: peak consumption is significantly lowered, there is lower net export of solar to the grid, and the ramp rate of the net load is significantly reduced. To be clear: the blue line simply shows the effect on net load of bringing forward AC consumption by three hours and increasing it by 10%, whereas a more careful simulation is required to obtain actual results with a real home. 

RossBaldick.com
Figure 2: Average net load 9 (kW) of homes with AC and net load with pre-cooling, on August 1, 2018. (Ross Baldick)

What does data for a single day tell us? Figure 2 depicts net load for the specific day of August 1, 2018. We can see that the general duck-like shape of the net load as shown by the blue line has been flattened by bringing AC consumption forward in time: as shown by the orange line, peak of net load is lower, net electricity exports from homes have been eliminated for this day, and the “neck” of the duck rising to the peak has a lower slope. Simulation of a pre-cooling strategy would undoubtedly show a different detailed pattern of net load, but a similar general effect could be expected.

Won’t consumers balk at spending more money on higher electricity usage to pre-cool their homes? California is addressing this problem by introducing new lower time-of-use (TOU) rates for electricity during sunny hours.

Traditional TOU rates were designed to shift consumption to nighttime, say after 10pm or 11pm, when load is typically lowest. Some argue, however, that these traditional TOU rates are ineffective, and recent evidence from Bruce Mountain, Victoria University, Melbourne, (click here to download the presentation) supports that claim, by suggesting that such traditional rates, with low prices overnight, have not convinced homeowners to shift their consumption to nighttime in the state of Victoria. No one wants to do their laundry in the middle of the night to save a few pennies.

But with the new TOU rates they would be willing to do their laundry, dishes, and electric-vehicle charging – and pre-cool their homes — in the afternoon. The new, improved version of TOU with lower prices during middle hours of the day was mentioned in the DOE Building America study, and that’s exactly what California is doing.

Renewables challenge us to rethink our basic assumptions. To mix metaphors, there is more than one way to skin a duck — or a magpie goose. With high PV penetration, we cannot always control supply to meet demand. We need to change demand to follow supply. And that’s what pre-cooling will achieve.

Next time: more ways to change demand to follow supply.

Posted in commentary, research | 9 Comments

Applied Optimization book: new and improved!

Cambridge University Press bookstore

Last summer, when I was visiting the University of Cambridge to present a seminar, I happened upon the university press bookstore.

And what did I find inside? A hardback copy of my 2006 book, Applied Optimization.

Applied Optimization

Even if you are in Cambridge, I do not advise you to buy the hardback copy. Instead, there is now an updated paperback copy available that includes corrections and additions. Click here to purchase.

There are a number of other optimization books out there, but if you want a careful introduction to optimization, convexity, and optimization, with multiple power systems case studies, please consider this book and the slides for my associated graduate course, “Optimization of Engineering Systems.”

Posted in commentary, education, news, research | Leave a comment