AI in Smart

Energy
Systems
Lecture 8

Sohail Khan
Contents
1 Trade-offs in Energy Systems................................................................................................ 2
1.1 The Energy Trilemma...................................................................................................... 2
1.2 Types of decisions in energy systems ........................................................................... 3
1.3 How to formulate a decision problem? .......................................................................... 4
1.4 Complexity in decision making ....................................................................................... 4
1.5 Summary: ......................................................................................................................... 5
2 Trade-Offs in Power Generation ............................................................................................ 5
2.1 Types of generation......................................................................................................... 6
2.2 Important limitations ........................................................................................................ 7
2.3 Meeting energy demand ................................................................................................. 7
2.4 The need for energy storage .......................................................................................... 8
2.5 Importance of forecasting ............................................................................................... 9
2.6 Summary .......................................................................................................................... 9
3 Capacity Planning ................................................................................................................. 10
3.1 Challenges of capacity planning .................................................................................. 10
3.2 Investing in generation .................................................................................................. 11
3.3 What is Dunkelflaute ..................................................................................................... 11
3.4 Capacity mechanisms ................................................................................................... 11
3.5 Infrastructure planning .................................................................................................. 12
3.6 Summary ........................................................................................................................ 12
4 Optimal Scheduling ............................................................................................................... 13
4.1 Formulating the optimal scheduling problem .............................................................. 13
4.2 Solving the optimal scheduling problem ...................................................................... 14
4.3 Negative prices .............................................................................................................. 15
4.4 Summary ........................................................................................................................ 15

Page 1 of 15
1 Trade-offs in Energy Systems

1.1 The Energy Trilemma

The goal of the Paris agreement: Reduce the carbon emissions from the energy
system radically. If we could build the required capacity in time, we would risk
frequent local blackouts when the sun doesn’t shine and the wind doesn’t blow. We
could use batteries when there is less generation from renewables, but that would be
prohibitively expensive.

With this, we would like our energy system to have three characteristics:

1. Sustainable: we want to phase out fossil fuels and their emissions.

2. Affordable

3. Reliable: This is essential for our modern society.

These three goals form a trilemma because they are often at odds with each other. It
is easy to design an affordable energy system if you disregard the other two goals.
And it is still manageable if you want that system to rely on renewable energy
sources, as long as you are willing to sacrifice reliability. However, doing all three at

Page 2 of 15
the same time is a great challenge. Decisions involve trade-offs - and they
involve making the best possible use of available information and
technologies.

1.2 Types of decisions in energy systems

1. Consideration: Thinking about the energy transition in the longer term means
we need to think about the system we would like to have – and how we will
get there.
Decision: The relevant decisions from this long-term perspective are
investment decisions. What do we build and where?

2. Consideration: In the shorter term, we need to consider the system we

actually have. How can we best use this system while balancing sustainability,
affordability and reliability?
Decision: That is primarily done using dispatch decisions: when do we use
energy – and from which source?

At the national or continental scale, where investment decisions could be about

creating a hydrogen backbone or the closure of nuclear power plants. Dispatch
decisions are primarily left to the markets in liberalized electrical power systems.
Still, the transmission system operators play an essential part in adjusting the market
dispatch if necessary, to ensure the grid’s reliability.

On the other hand, the same framework can be applied to local energy systems,
down to the scale of your own home. The long-term decision problem then includes
the option of investing in a home battery, whereas the operational dispatch problem
can be about charging your electric vehicle during off-peak hours.

The two decision problems are not independent. Clearly, past investment decisions
have influenced the operational decision problem, but that can be considered water
under the bridge. Those decisions have already been made. More importantly, one
cannot think about future investments without considering dispatch decisions. For
example, if reliability considerations mean we should curtail the amount of wind
power we use, it will affect the trade-offs for investment. As a result, investment
problems always encapsulate operational dispatch problems.

Page 3 of 15
1.3 How to formulate a decision problem?

There are two key elements to the formulation of a mathematical decision problem:

1. Objective: Here, we define a quantity that is to be maximized or minimized.

We might choose one of the three components of the energy trilemma, or we
could try to balance them by placing them on the same footing. For example,
if we assign a monetary value to reliability and carbon emissions, we can
minimize the total costs. Whichever choice we make, in this setup we
consider a decision optimal if it attains the best value of our objective, among
all other possible options.

2. Constrains: Our freedom to select an optimal decision is constrained by the

limits of our system. For example, we cannot produce more power than a
generator is physically capable of producing. We cannot spend money we do
not have without resorting to a more expensive loan. This category can also
include limits we choose to impose, such as limits on CO2 emissions or limits
that ensure the physical safety of maintenance workers.

Together, the objective and the constraints make up the decision problem. Once it
has been formulated precisely, we can try to solve it. That is, we try to identify the
optimal decision.

1.4 Complexity in decision making

Finding that optimal decision may be easy, or difficult – that depends on the problem.
However, in general, we can define a few factors that significantly influence problem
complexity. These factors are

1. Timespan
Decisions that only concern the short term are generally the simplest. At the
other end of the scale, complexities stack up if we consider long-term plans or
the impact that short-term decisions have on future decisions.

2. Scale
This is the most obvious factor of the problem you are dealing with. In a
national electricity network, there are many more possible actions and
constraints to consider than for the case of a single home.

Page 4 of 15
1.5 Uncertainty

These could be uncertainties about the future – for example, related to the weather
forecast – or uncertainties about the past and present – for example, if you are
concerned about measurement errors. To deal with uncertainty:
 In the simplest case, we simply ignore them!

 Doing one step better, we can define a range of scenarios and consider our
decisions against each scenario.

 At the very complex end, we may even consider the knowledge that we can
re-assess our decisions later and we have the so-called “multi-stage
stochastic programming” framework – which tends to be very complex to
solve.

1.6 Summary:

 The fundamental balancing act between sustainability, affordability and

reliability when designing and operating energy systems is known as the
Energy Trilemma.

 In practice, these decision problems are often classified into investment – also
known as planning – problems and operational problems, such as dispatch
decisions.

 These decision-making problems are mathematically stated in terms of their

objective and constraints

 The complexity of such problems is linked to the time span considered, the
level of detail of the energy system and how we deal with uncertainty.

2 Trade-Offs in Power Generation

Dispatch decisions in energy systems are about deciding who supplies energy,
when, and where. We will need models to reason about such dispatch decisions in a
quantitative way—models for the costs and emissions of various generators and
models for the constraints on their abilities.

Page 5 of 15
2.1 Types of generation

In real energy systems, there are many types of units generating energy. The
different properties of these units determine how they are best used. Some of them
based on CAPEX, OPEX, emissions and reliability are:

1. Renewable energy sources: such as wind and solar power. The cost of
construction – the CAPEX – is substantial, but once we have built them, the
cost of generating energy is effectively zero. And there are no CO2 emissions
to worry about. But the main concern with these energy sources is that their
output is highly variable and, to some extent, unpredictable. So you cannot
depend on these generators alone for a continuous supply of power.

2. Traditional generators: They generally have fairly low costs – both CAPEX
and OPEX – and high reliability. But of course, as we all know, emissions are
the problem for this category.

3. Carbon Capture: As a result of huge emissions from traditional generators,

there are many efforts to reduce emissions of such fossil power plants, for
example, by equipping generators with carbon capture and storage
technology. This category is characterized by reduced emissions, obviously,
but this results in higher costs, both for the initial investment and for
operation.

4. Nuclear: This nuclear fission generator type has the largest capital
requirements but produces relatively cheap and reliable energy. It also does
not emit harmful substances into the atmosphere – but there is the long-term
problem of nuclear waste products to deal with.

Note: The generators in a system contribute to the various system-level objectives

for decision-making, which are cost, emissions and reliability. Costs are divided into:

 Capital expenditures (CAPEX): These are the long-term costs associated with
building and maintaining a generator.

 Operating expenses (OPEX): The day-to-day running costs of the generator.

Page 6 of 15
2.2 Important limitations

In order to make informed, quantitative decisions about which type of generator to

build and which to use, we require mathematical models for each kind of generator,
and each of these aspects: costs, emissions and reliability. But that is not enough:
some of the biggest challenges in decision-making for energy systems are related to
the dynamic behavior of these units. In particular, two types of constraints have an
essential role in driving dispatch decisions:

1. Ramping limits: There are limits to how quickly generators can adjust to
changing demand. Generators can only increase their power output so much
or decrease it by this amount in a given time frame. This is especially true for
nuclear generators, which are typically designed to run at a constant power
level, so it’s difficult to adjust them to follow daily demand cycles.

2. Switching on/off-limits: This is related to starting and stopping generators.

Often, especially for generators with complex thermal systems, there is a
minimum on time, a minimum off time, and a minimum notice period for
starting up. You cannot simply instruct a nuclear generator, or large coal-fired
power plant, to run for an hour to cover the demand for electric ovens at
dinner time.
Moreover, in addition to the hard limitations, one should consider the cost
involved with starting up and stopping a generator. Even if it’s technically
feasible, it may be economically unwise to do so.

2.3 Meeting energy demand

What is the challenge we face when determining an optimal dispatch for the
system? We have a variety of resources at our disposal to supply the energy
demand. We would like to find an optimal way to supply energy over a certain
horizon – let’s say one day. To do so, we start with the total load we need to supply,
which varies considerably over time.

How should we use our generators to supply that load? We have solar and wind
generation, and because these sources are zero-cost and zero-emissions, we want
to use them whenever we can. But as you can see, this is not always enough to fully

Page 7 of 15
cover the consumption. This remainder needs to be supplied using a combination of
our other generators: clean and dirty fossil, and nuclear.

Keeping in mind the operational costs and emissions, we’d like to use the nuclear
units to supply as much energy as possible, but because of the limitation on their
ramp rates and minimum up- and downtime, we’d only use them to supply the
constant baseload. For the remainder, we’d ideally use the clean fossil category, but
that’s relatively expensive, so it may not be worth it for the evening peak. So we
might still call upon the dirty fossil plants to cover the peak.

2.4 The need for energy storage

There is a fundamental challenge in renewable energy systems: The presence of

solar power generation can result in extremely low remaining demand during the
mid-day hours, followed by a steep ramp to a high evening peak. If you recall the
ramp limits on generators, this might really impact one’s ability to use low-cost or
low-emission sources of energy.

Shift the energy production of solar panels to the evening, instead of during the day?
This is what batteries and other forms of energy storage can effectively do. They
store the energy produced by renewable generators, allowing us to use it when it is
most useful to us. And this changes the remaining demand that other generators
must supply, meaning it is cheaper, more environmentally friendly, and more reliable
to operate our system.

Because of this ability, storage technologies are a critical component of renewable

energy systems. They allow you to achieve a better trade-off between affordability,
sustainability and reliability.

If we want to quantify this impact for generators, we’ll need to specify a model for
energy storage units. In general, battery models can get really complex, but a simple
model can be constructed by considering the following technical properties:

1. Capacity: is the maximum amount of energy you can store inside.

2. Charging and discharging: There is a maximum power that the battery can
consume when charging, and a maximum power that the battery can deliver
when discharging.

Page 8 of 15
 3. Efficiency: when we are charging or discharging, there are losses. You
notice this by the battery getting warm to the touch. This is represented by the
charging and discharging efficiencies.

Disadvantage: Purchasing cost. Although the cost of large-scale energy storage,

especially in the form of batteries, is still very high, it is coming down rapidly. The
exploding use of batteries in electric vehicles is driving down costs to the point where
battery storage is finding ever more use cases.

2.5 Importance of forecasting

When we want to determine the best possible dispatch for this system, we need to
deal with sharp changes in power requirements, for example, in the late afternoon
when there is a huge demand. How should we optimally use the resources at our
disposal to deal with this rapid variation?

One crucial complicating factor is the ramp constraints of generators. They link what
a generator can do in the future to what it is doing now. And that means we must
consider what will happen in the future when making decisions about the present
time. This is done using forecasts – forecasts of future energy demand, and
forecasts of future renewable energy production.

In the example where there is huge demand later in the afternoon, based on our
forecasts, we might choose to start ramping up our fossil generators early in the
afternoon, because otherwise, they may not be able to reach their desired output in
time for the evening peak.

2.6 Summary

 There are different ways in which the type of generators impact the overall
objectives of investment and dispatch decisions. This is done based on
several objectives: costs, emissions and reliability.

 Constraints on the ramping rates and on- or off-times influence the flexibility of
individual generators – and therefore of the system as a whole.

 Battery storage can help with alleviating these constraints, for instance, by
storing energy to use when the demand is highest.

Page 9 of 15
  Forecasting is important, especially when dealing with ramping constraints. It
is essential to anticipate significant changes in demand and renewable
generation so that you can dispatch other units on time. This way, we are well
prepared to face the unknown.

3 Capacity Planning

Despite the emerging trend for local energy production, we still rely heavily on
transmission infrastructure to transport energy across long distances. Building large-
scale infrastructure for transportation and generation is cost-efficient, but it’s capital
intensive and costs a lot of time, so it needs careful planning.

3.1 Challenges of capacity planning

There are three challenges in capacity planning:

1. Delayed effect: The overarching problem is that building a new generation or

transmission capacity takes time.
• It can easily take a decade to realize a new long-distance transmission line,
for example. This means that if you discover you have insufficient capacity,
it’s already too late to do anything about it.
• It’s essential to plan ahead, using forecasts.

2. Long-term uncertainty:
• Uncertainty about future energy demand.
• Uncertainty about the development and uptake of new technologies.
• Uncertainty about future energy policy.

3. Different actors: In deregulated energy systems, there are different actors

making decisions independently. On the one hand, we have the generating

Page 10 of 15
 companies investing in generation. On the other hand, there are transmission
and distribution system owners investing in transport capacity.

3.2 Investing in generation

Generating companies are the one that invests in a generation. These are typically
commercial investors, who decide on either investing in more generating capacity or
sometimes choose to close down existing generators.

As a society, we are dependent on such decisions by investors, collectively, for the

security of our energy supply. So it’s an important question to ask whether we risk a
shortfall of energy supply. Because, when we discover there is not enough, it’s
already too late. The studies we perform months or years ahead of time to estimate
the security of supply is called resource adequacy studies.

In recent years, these studies have highlighted concerns about the electricity supply
due to our increased reliance on renewable energy sources such as wind and solar
power. Once built, those generators can produce power at a negligible cost. This has
– by design – eroded the business case for investing in traditional fossil-fuel-
powered generators.

3.3 What is Dunkelflaute

The output of wind and solar generators is highly variable, and sometimes we face a
prolonged period where both are unavailable. This phenomenon is called a
Dunkelflaute – a period of days or even weeks where there is little sun and little
wind. It is sufficiently long that small batteries and flexible energy demand combined
may not be sufficient to ensure the security of supply.

Installing spare generating capacity – or long-term storage facilities – may be the

only solution for such cases, but investors might need additional incentives to build
them.

3.4 Capacity mechanisms

When a system operator, or a state regulator or government determines that the

security of supply is insufficient, they can set up a scheme to support the
construction of additional generators, batteries or network connections – or prevent

Page 11 of 15
investors from mothballing old generators. Of course, they will then need to compute
a price tag for these services, which is done using resource adequacy studies.

One particular capacity mechanism that is used in a few countries is a capacity

auction. The system operator or government determines how much capacity should
be purchased, and organizes an auction to purchase this capacity from capacity
providers: those can be generators, or sometimes owners of transmission cables
between different countries. And that last option nicely illustrates that we do need not
only generators but also transport capacity: in the case of electricity, these are
transmission lines.

3.5 Infrastructure planning

There are challenges faced by the infrastructure owner in planning ahead. The
requirements for the system operator in a liberalized energy system are as follows:

1. To facilitate markets, and not interfere with trades unless absolutely

necessary.

2. To respect the technical limits of the equipment.

3. To ensure a dependable supply.

Note: Building new transmission lines can take an exceptionally long time. This is a
complicated problem, for which forecasts are essential. Forecasts of demand, but
also of the sites and sizes of future generators. And finally, in all these decisions,
the system operator needs to trade off reliability and cost-efficiency. Avoiding all
risks would be excessively expensive, but investing too little may lead to costly
outages.

3.6 Summary

 Capacity planning is a complex problem in general due to long lead times,

significant uncertainty, and multiple players’ involvement.

 Looking specifically at companies investing in the generation, we are

increasingly facing challenges due to the variability of renewable generation.
There is a risk that the security of supply cannot be maintained during
extended low-wind, low-sun periods, called Dunkelflaute.

Page 12 of 15
  In response, some authorities are introducing capacity mechanisms to provide
extra incentives for generating companies.

 There are particular challenges faced by infrastructure owners, who invest in

infrastructure on behalf of all the users. They need to plan in advance,
requiring them to have accurate forecasts and balancing reliability against
cost-efficiency.

4 Optimal Scheduling
Optimal scheduling is the operation of generation and storage facilities to produce
energy at the lowest cost to reliably serve consumers, recognizing any operational
limits of generation and transmission facilities.

Although in this definition we mention the lower cost, we recognize that there may be
other important objectives as well, such as reducing emissions or improving
reliability.

 We might use the terms economic dispatch and optimal scheduling

interchangeably, but we consider this definition for both unless otherwise
indicated.

 We do not consider any transmission capacity limitation in the optimal

scheduling problem.

 Goal: We want to ensure continuous power balance in the system. This

means that there is as much power generation as there is consumption at a
time.

 There exist multiple solutions that lead to generation-consumption power

balance. We differentiate between them by assigning a cost to them based on
our objectives. These can include operational costs, reliability and emissions.

4.1 Formulating the optimal scheduling problem

To formulate the optimal scheduling problem, we need several elements:

1. Forecast of demand and Generation: We require a forecast of the expected

power consumption as well as the expected power generation from the
renewable-based generation facilities such as wind and solar parks.

Page 13 of 15
 2. Technical Specifications: We require the dispatchable generation units’
technical specifications. These can include:

o The unit’s energy efficiency

o The maximum and minimum generation capacities

o The ramping-up and -down limits

o The restrictions on minimum down- and up-time.

4.2 Solving the optimal scheduling problem

There are different steps to take to solve the optimal scheduling problem.

Step 1: Take all our system and units’ information

Step 2: Define a specific goal for the problem

Step 3: Translate the goals and information into mathematical models. In practice,
there exist many models for all components, some more complex and accurate than
others.

Step 4: With this mathematical model, Create an optimization problem

Step 5: Solve using any mathematical programming language, such as Pyomo,

developed for Python.

The optimal scheduling problem will determine what combination of generator power
outputs will lead to the lowest possible cost that covers the remaining consumption.
The outcome of this optimization is an optimal schedule for all the dispatchable units.
The cheapest and quickest generator is used the most.

Note 1: Although a specific hour might have a higher cost, these outputs are still
called optimal. This is because they are needed to decrease the overall expenses
over a planning horizon while meeting the power generation-consumption balance
constraint.

Note 2: The output of the nuclear reactor is flat. Due to its properties, it is slow in
changing its power output. This is why nuclear is not per se the magical solution to
bridge the gap between renewable generation and demand.

Page 14 of 15
4.3 Negative prices

Negative prices are an interesting phenomenon that can occur when we are unable
to control generation, for technical, regulatory, or market reasons.

One example of when this phenomenon might occur is when renewable energy is
given subsidies to produce energy and when there is an abundance of wind and
solar irradiance.

If there is an imbalance in power. There are two solutions: either decrease the
generation or increase the consumption. If the former does not happen because of
the incentives given to renewables, we have to do the latter. And this is achieved by
Consumers being paid to restore the power balance, resulting in negative prices.

The negative prices are paid by the owners of the renewable's units. However,
because of subsidies, it is usually more attractive for them to keep producing instead
of shutting down, resulting in negative prices.

It’s good to know that negative prices might also occur when such subsidies are not
available, but there simply isn’t a way to control generation. For example, if PV
panels on roofs – that are generally not controllable – threaten to supply more than
the total demand, or additional generators must run to keep the grid stable.

4.4 Summary

 Optimal scheduling problems can be used to define the output power of the
dispatchable generation units while maintaining the generation-consumption
power balance.

 Type of technical information required to state the optimal scheduling

problem: forecast of the loads and non-dispatchable units such as wind farms,
and the dispatchable generation units’ technical specifications.

 Negative prices: This happens when there is an abundance of renewable

generation and low consumption. This can have multiple causes but the
outcome is the same: consumers are paid to consume energy, resulting in
negative prices.

Page 15 of 15