Predicting load in the utility business has never been easy. Witness the track record that the industry has accumulated by reviewing the NERC projections over the years. Even though NERC substantially improved its forecasting capability, projections have always been off. There are many reasons for this among which hard-to-forecast weather, shifting customer population trends, a shift from industrialization to services, and an unproven relationship between electricity demand and either GDP or the increased use of electronics in households.

Of course, the last 18 months have not made load forecasts any easier. For the most part, forecasts have always been up, with the exception of the 1991- 92 and 2000-2002 recessions. But the 2008 crisis has resulted in an unprecedented drop of 7.5% in energy sales down from its 2005 record. The latest statistics show that the total US utility load in 2009 (3575 TWh) was back to its 2004 level (3547 TWh) but several utilities have experienced load drops in the 6-10% range from their peak this past decade. Of course, industrial sales took the biggest hit (down by 13%). By contrast, residential sales have only dropped by 2-3% across the board but the small customer market may take another 2-3 years to recover in some franchise areas.


Jean-Louis Poirier

Load forecasting is not going to be any easier. Some still insist on asking when future loads will return to “normal” and, in their mind, it mostly depends on when there is “full” recovery. But that’s not all. Finally, after two decades of increasing talk about demand-side management and distributed resources, the smart grid is about to emerge. What that really means, nobody is really sure but it is likely that we will see between 70 million and 90 million of new so-called smart meters being installed by 2020 – or between 44% and 57% of the 159 million meters in operation by then. This trend will not only involve the 70+ large IOUs that we have in the US but also a couple of hundred municipalities and co-ops. Utilities are intent to use the new metering capabilities to deploy new “dynamic” rate structures (such as revised Time-of-Use rates, Critical Peak Pricing, or Peak-Time Rebates) designed to spur demand response (DR) in a way that has never been seen before. Briefly, estimates tend to range between 50 GW and 100 GW of additional DR capability by the end of the decade.

Of course, the devil is in the details. There are various types of smart meters being installed and different levels of control capability being deployed. In fact, there are very few smart grid deployments that look alike. Second, we still don’t quite know how much DR will be load control (much more 0-1) versus dynamic rate pricing (much more elastic). Third, the new dynamic rates must be designed and approved by the relevant public utility commissions (PUCs) – a process that can take some time and result in compromised program features. Then, we have to see whether customers respond the way the models say they will. Of course, the models do not all agree and estimates of peak reduction triggered by dynamic rates vary between 6% and 15% depending on how steep the rates are during critical times when peak reduction is most desired. Load reductions due to time-of-use rates tend to be smaller – between 2% and 6% for customers without high air conditioning loads and between 6-10% for customers with either high winter (electric heating) or high summer consumption. In large part, these models rely on the results of early pilots. But after reviewing nearly 20 of these, it is hard to conclude that they represent a strong statistical sample (mainly stemming from differences in weather conditions, rate structures, meter capability and customer segmentation). Pilots also do not give much information on the robustness of DR: is load shedding a sure thing from year to year; does it grow; and how will it play withnew retail prices in unregulated retail areas? Likewise, new rates and smart meters should trigger increased energy efficiency, but it is not always easy to estimate that component separately (and in some cases, it also means reduced gas usage).

So, we don’t really know… plus, the impact will be very utility specific as we can see from utility filings with expected paybacks on their smart grid investments often in the 7-10 year range. Some utilities may experience a 8-12% DR-induced load reduction while others next to them may be more in the 5-8% range as the result of not only dynamic rates but also the deployment of in-house displays and controls. The range is even wider when one looks at the Brattle Group’s estimates of DR-induced peak reductions; these range between 4% and 24% with an average of 11%. At the end of the day, though, there will be increased DR and, hopefully, this will be a one-time durable investment in customer empowerment. Now, for utilities, this may also mean lower revenues in some cases and increases in the fixed component of fixed rates – something called “decoupling” which is not really DR-friendly.

Having said all of this, the 50-100 GW estimate for added DR may not look that bad even though RTOs are likely to seriously evaluate the reliability of this emerging DR capability. Still, it can make a difference as evidenced by a recent PJM simulation that showed a recommended DR cap of 8.5% of system peak which would amount to a very respectable 16 GW of deemed-reliable DR capacity in that pool at this point in time.

Know your customer! it does help. Unfortunately, generators (and their regional councils, by extension) were used to rely on utility load forecasts. Now, that will be a bit harder not just because of smart grid roll-outs but also because of the impact of growing wholesale renewable resources, from both wind and solar. There too, there is a wide range – roughly anywhere between 100 GW and 150 GW of wind in place by 2020 subject to the amount of transmission capacity that can be invested in. These new resources deliver an intermittent load causing wide fluctuations but otherwise requiring more ramp up system flexibility. There may be more gas capacity back-up, and if possible investments in new power storage facilities. The most recent NERC scenario assessment (released in late 2009) assumes 155 GW of additional renewable capacity in 10 years, calling for up to 40,000 miles of additional transmission. The result is a 15% penetration for renewables – not the 20% that some have been asking for – but closer to the average of all RPS state mandate targets announced (or extrapolated) for 2019/2020.

Even then, the NERC analysis concludes, this would trigger many changes:

• Minimum generation limits during
light load conditions
• Increased ramp requirements and
out-of-phase ramping
• A much better coordination
between day-ahead ancillary
service markets and wind forecasts
and real-time monitoring of wind
• Additional operating reserves.

On that last point, the reserve margin increases could be substantial in some regions. Based on the NERC analysis, I estimate the following impact in terms of needed reserve margin increases:

• A jump from 13% to 28% in MRO
• From 20% to 27% in NEPOOL
• From 28% to 37% in SPP
• From 15% to 22% in SERC

So, it roughly is an increase of at least 40 GW in additional reserve margin that would be required in that 15% renewable penetration scenario.

Our own analysis shows that some states are likely to be much more impacted than others by the combination of high renewable requirements and extended smart grid deployment. We found that the 15 most impacted states would be, in decreasing order of severity, Connecticut, Maine, California, Maryland, Minnesota, DC, Texas, New York, Ohio, Pennsylvania, New Jersey, Delaware, Nevada, Utah and Vermont. Clearly, the threat is nationwide. These states are all over the US, not in one region.

Another complication is the impact of carbon legislation. If it is drastic, it will trigger some changes in dispatch intra- and inter-dispatch patterns and even reallocation of existing generation capacity. For example, a PJM study concluded on a range of impact between $7/MWh and $45/MWh by the mid-2010s, based on a carbon price between $10/ton and $60/ton. While that analysis was centered on the kind of bills debated in 2009, and while the more recent bills are different, it still is an indication of the kind of impact range that is at play. There is now more and more talk of plant closures for older mid-size coal plants. While these plants have survived much longer than any consultant ever thought, we may now approach the time where these plants finally shut down, not just on economics but also because their owners may be able to get in exchange accelerated permits for new super-efficient gas generation. To give an idea, that could be a swing of 20-50 GW if carbon prices exceed $30/ton by 2020. It also means that congestion costs could significantly change in amount and location for several East Coast RTOs.

So, generators have to deal with a triple moving target: utilities that experience more elastic loads in balancing areas exposed to more volatile renewable resources and with the possibility that the existing fossil-fired capacity does not quite behave like it used to. While any one of these factors is not uncommon, and even combining two of them is not unheard of, dealing with all three promises to be interesting.
To illustrate the point, we show on Exhibit 1, the range of uncertainties that are involved and estimate what we call the potential for load imbalance. We also know that in even the best organized RTO markets, small imbalances can trigger wide wholesale price volatility. That analysis shows that we would need another 20-50 GW of new (mostly gas-fired) fossil capacity – which is double the capacity of 20 GW of new combustion turbines and combined cycle capacity shown in the latest EIA run (issued in early 2010). We should also say that we don’t see much utility power storage capacity developed by 2020 – maybe 1 GW.

The estimates we show are large enough to suggest that there will be a need for mitigating measures.

First, utilities will hopefully invest and be successful in the use of analytics to track demand response as well as they can. So far, this has not happened but this will become a must by 2012-13.

Uncertainty Factor Unit 2020 Outcome
Smart metering development Million of units 70-90
DR peak reduction % 10-15
DR peak reduction GW 50-70
DR load reduction % 5-10
DR Load Reduction TWh 200-400
Additional Renewable Development GW 50-100
Dislocated generation
(due to carbon legislation)
GW 20-50
Potential Load Balance Impact GW 20-70
Fossil Capacity Needed to offset GW 20-50

Exhibit 1

Second, we can trust at least some utilities to want to become smarter in yield management and design new rates, especially in unregulated areas. Utilities will find out that not all DR is to be desired. Right now, though, we’ll take any DR. In addition, we should count on private DR providers to really mine the commercial market. We can already see signs of increased activity on this front with major corporations, such as ABB, AREVA, Cisco, GE, Honeywell, Johnson Controls, and Schneider Electric, acquiring smaller outfits that offer automated DR services and aggregated DR portfolios. This too should mushroom in the 2011-2013 period.

Third, by the mid 2010s, utilities will invest more in the other part of their operations that has been neglected: active distribution management (ADM) which will include increased investments for:

• automated feeders and RTU line monitoring
• faster restoration – so called Automatic Sectionalizing & Restoration (ASR)
• More adaptative volt/VAR control – new approach is called SCADA controlled Volt-VAR.
• More capacitors.

By 2015, ADM investments will grow and could average $4-5 billion annually by the end of the decade. Of course, some ADM investments could be done now but most utilities find it hard to ask for more monies from the PUCs and customers on the top of individual smart metering programs that can cost hundreds of millions of dollars when they involved a million or more new meters. For example, PG&E’s
smart grid plan calls for deploying some 4.6 million new meters at a cost of about $2.2 billion. Also, in all fairness, ADM investments are best done when utilities have a better sense of true demand response, thus the need for better analytics again.

With ADM and better grid sensors, distribution companies may be able to achieve true “Management by Wire”. That’s the “Holy Grail” when utilities can use their meter data management systems (MDMs) to continuously balance on one hand a fluctuating supply (at least in terms of price but also quantity available if there is a large component of renewables) and a malleable demand that is subject to a true demand-responsive rate structure well supported by regulators. At that point, utilities will also be able to better incorporate in their integrated resource planning strategies, new portfolio management tools to optimize smart grid commercial operations with demand side optionality and the incorporation of Virtual Power Plants, distributed generation and microgrids as an increasingly larger and more viable resource category. Of course, that highroad scenario does not cover vegetation management which will still have to be done in a more conventional way…..

But there will be other trends to help solve the equation:

1. New Intra-RTO “backbone” transmission
will become a major issue that will
prompt consideration for larger RTOs
2. Utilities and RTOs will have to figure
out ways to price new ancillary services
and new storage capacity
3. Repowering on brownfields may
see a second life – that’s flexible gas-fired
load balancing capacity near the loads
4. The business for turbine O&M
services by both OEMs and third-party
providers will continue to be a very good
one indeed as more CTs and CC machines
experience increases in start-up requirements
to balance the loads.

So, it won’t be easy. Plus I forgot to mention that by 2020, there may be the emergence of electric vehicles. The impact on the grid is not known either… but that’s for another decade to sort out.