AI in Smart

Energy
Systems
Lecture 6

Sohail Khan
Contents
1 Computational Methods for Energy Networks ........................................................................... 2
1.1 Energy networks ................................................................................................................ 2
1.2 Why model energy networks? ............................................................................................ 2
1.3 Conventional energy systems ............................................................................................. 3
1.4 Future energy systems ....................................................................................................... 3
1.5 Modelling cycle .................................................................................................................. 4
1.6 Model types ....................................................................................................................... 4
1.7 Transient behavior in gas and electricity networks ............................................................. 5
1.8 Summary............................................................................................................................ 5
2 Simulation of Transmission and Distribution Networks .............................................................. 6
2.1 Transmission vs distribution networks................................................................................ 6
2.2 Challenges.......................................................................................................................... 8
2.3 Summary............................................................................................................................ 9
3 Simulation of Gas Networks ..................................................................................................... 10
3.1 Unknowns in gas networks ............................................................................................... 10
3.2 Gas types ......................................................................................................................... 11
3.3 Gas Simulation ................................................................................................................. 11
3.4 Components to be modeled ............................................................................................. 12
3.5 Gas vs. electricity ............................................................................................................. 12
3.6 Gas-Electricity coupling .................................................................................................... 13
3.7 Energy Hubs ..................................................................................................................... 13
3.8 Summary.......................................................................................................................... 14
4 Easy and Complex Models ....................................................................................................... 15
4.1 Generator model.............................................................................................................. 15
4.2 Power flow models .......................................................................................................... 16
4.2.1 AC model:................................................................................................................. 16
4.2.2 DC model:................................................................................................................. 16
4.3 Combining models ........................................................................................................... 17
4.4 Challenges of models ....................................................................................................... 17
4.5 Summary.......................................................................................................................... 18
1 Computational Methods for Energy Networks
1.1 Energy networks

There are different kinds of energy networks. Let us consider the most important three:

1. Electricity network
2. Gas networks
3. Heat networks

All these networks are different. These different networks have distinct properties and
phenomena. One of the essential things to consider is how we can interconnect them? Or, how
can we integrate to make these networks which have distinct properties work together?

1.2 Why model energy networks?

There are various reasons why we might want to model energy networks. Some of them are:

1. We wanted to know what could happen in the present and the future. For example, what
if we add a new consumer or a new producer? Then it is essential to know what will
happen in the network. Hence, we can study what should be included and removed from
the network. We can also check the effects on the other consumers or producers.
2. To analyze the energy flow in the network. Energy flows from one network node to
another network node, leading to the solution of the simulation problem.
 3. To analyze the effect of defects and breakage in a line. This analysis helps to see
whether it is local. Or will there only be a problem in the neighborhood of these nodes?
Or will this effect globally? And how will that interact?
4. Another idea of network modelling is that you can adapt your models. This means that
you can model the network differently to solve the problem before it happens.

1.3 Conventional energy systems

If we consider a city to be energized, we can see that electricity, gas and heat systems will be
standalone; they do not communicate with each other. There is no interaction between the
networks.

Note that we have two lines in heat networks. A line that carries hot water to the houses. And a
line that carries cold water flowing back to the heat generator.

1.4 Future energy systems

In future, we could see a different model. The conventional energy sources are present, and
renewables like wind and solar power, alongside hydrogen gas, are introduced. This means that
the conventional generators play a smaller role. There will also be an interaction between two
networks: electricity and gas. This means that the energy systems are integrated, so part of the
energy can come from gas, and part of the energy can come from electricity.

Moreover, it is possible to transform electricity to hydrogen, store it, and later release it so that
it can be used in the network to provide energy.
1.5 Modelling cycle

Fig. 1. Modelling cycle

To explain, understand and do simulations on future grids, we need a model. To create the
model, we use the modelling cycle.

We start with a simple, straightforward model, and then after we have done this modelling, we
compute its solution, or in other words, simulate it. Then, we compare the results of this
simulation with reality. If it is good enough, we can stop and say, OK, this model is what we
want to execute and use.

If the model is not good enough, we improve it and create a more complex model. Then, we
repeat the simulation steps and check again whether it fits or not. If it fits, we stop and stick
with this model. Otherwise, we can go to the next modelling cycle until it satisfies.

This cycle is represented in figure 1.

The modelling cycle also depends on the application. A simple model might be good enough
for specific applications, whereas, for certain other applications, we would require a very
complex model to have a good description of reality.

1.6 Model types

There are two different model types:

 1. stationary model: which means we have a certain steady-state and the model is solved
for a steady-state condition.
2. Transient model: We go with the transient model when there is a transient event such
as an energy request in the network. Then, later, the system will most likely return to a
steady state.

1.7 Transient behavior in gas and electricity networks

 Transient behavior in an electricity network is very fast. This means that the transients
are transmitted from one end of the Netherlands to the other in the order of
microseconds.
 If a transient occurs at the gas producer side, it will take a particular time ΔT, depending
on the distance, to see the effect on the consumer side. This delay could range from
minutes to half an hour.

1.8 Summary

 Analyzing energy networks allows us to predict what would happen if an event

occurred.

 In the future, energy networks will be linked to each other

 Modelling cycles go from a very simple model to more advanced ones after each
iteration if it does not fit the actual scenario.
 Models can be stationary or transient. Stationary models are similar for electricity and
gas networks, while transient models of gas networks introduce a time delay.
2 Simulation of Transmission and Distribution Networks

2.1 Transmission vs distribution networks

In the transmission network, we could see the big producers and the high voltage network
that transports the electricity from the producer end to cities or factories. Then it goes into the
distribution network.

In general, an electrical distribution network is a low voltage network that is more localized.
Think about a city or part of it or a small group of small industries where only a limited
amount of power is consumed. This is the part where the distribution network is utilized.

A transmission network has more of a mesh structure which means that the cities through
which the electricity flows are interconnected, ensuring reliability. On the other hand, the
distribution network is more localized, for example, the street or neighborhood of a particular
city. Here you have one connection point, and energy is tapped from this point and
transported downstream. This is termed a radial or tree-like network structure.

Both transmission and distribution networks have three phases. But in the transmission
network, they are balanced, and in the distribution network, they are unbalanced.

The consequence is that for a transmission network, it is sufficient to simulate only one
phase. Because all phases are in balance, the other phases will simply be a phase shift of the
simulated phase. Due to its unbalanced nature, you have to compute all three phases
separately in a distribution network.

The responsibility of these two separate networks lies with two different utilities. In the
Netherlands, the transmission network is handled by only one Transmission System Operator,
a TSO utility called TenneT. In contrast, the distribution network is the responsibility of
different parties like Alliander or Stedin, also known as Distribution System Operators or
DSOs for short. It is to be noted that the TSO and DSOs operate separately. This can lead to
problems such as limited information sharing due to privacy issues. Hence, it is not easy to
simulate both networks in a coupled manner.
The main differences are summarized in table 1.

Goal: It is to be noted that the models of the transmission and distribution networks are
different. Moreover, these networks are standalone, so we have a separate simulation for the
transmission network and the distribution network. Due to the penetration of renewable
energy sources like wind and solar, the goal is to couple the transmission and distribution
networks and do the simulation on this coupled network.

How to simulate electrical networks?

We need some kind of solver that can simulate distribution networks, transmission networks,
and the coupled network.

The goal of simulation: Reliability

Input to the solver: Properties like the voltage, the resistance of power lines, the generated
power, and coupling of the different lines. Note: for all nodes at each instant, the energy
conservation property should be satisfied.
Result: To have an optimal operation of the different networks.

We already have solvers for the individual networks, but how can we simulate both networks
together? There are two approaches we can use:

1. Decoupled solvers: The electrical transmission and distribution networks go

separately into their own solvers and then iterate to know what happens. For example,
the information from the transmission network is sent to the distribution solver. Then,
the distribution part is solved, and the obtained solution is sent back to the
transmission solver. This is done a couple of times until we get a certain convergence
or in other terms until the solution is satisfactory.
2. Coupled solvers: This is where simulations are done together. Here both the
transmission network and the distribution network are described by their own models.
Those are combined in one big problem, and then we have one common solver that
tackles the problem. Hopefully, the solution is the same as what is done in the
decoupled solver, but the paths taken are different.

2.2 Challenges

There are certain problems associated with solving coupled transmission and distribution
networks. The important one are:

1. Non-linear solution: If we solve both transmission and distribution separately, we get

a unique solution for both the network and the coupled network. Remember that it is
not a linear problem which means that we cannot add the results together. Instead, we
have to do that in an integrated way.
2. Numerous equations: Questions such as: Can we use them?, Do we need to use them
all? or Can we only partly use these equations? arise. We have to see the right amount
and the accurate equations to solve.
3. Simulation techniques: one of the key differences between the two networks is in the
balancing of the networks and needing to simulate only one or all three phases. This
leads to having two different simulation techniques. We must determine if it’s best to
do decoupled or coupled simulation, both for the transmission and distribution grid.
And also, how can we combine them accurately and efficiently?
 4. Confidentiality: By law, it is not always allowed to transmit data from the TSO to the
DSO or the other way back, which makes it hard to solve a coupled system. Thus,
sometimes we cannot use the best or the most optimal solver because we do not have
sufficient data available. This also means that it is not only a problem from the
mathematics or the physics point of view, but it also becomes a juridical and ethical
issue. In the future, the interaction between these different network owners should
become possible.
5. Data sufficiency: if insufficient data is available and a coupled simulation is required,
we can still do it, but it becomes more challenging to couple the two systems.

2.3 Summary

 The transmission grid is meshed, balanced, and requires only a single-phase

simulation as opposed to the distribution grid. Also, both parts are handled by
different entities.
 It is needed to couple both the networks in the future. For this, two methods can be
used: a decoupled and a coupled approach.
 There are challenges when coupling the two networks. These include the non-linearity
of the solutions, insufficient data exchange, and a different type of simulation needed
for each network.

Note:

 The Transmission system is used to deliver power from power station to consumption
centers. It is used for transporting power using a 3 phase (when considering AC
transmission). Unless there is an unbalanced fault event, the transmission system is
balanced.
 The Distribution system is typically used to power consumers. Different types of
consumers can be found:
o Three-phase power consumers (eg. motors), typically balanced.
o Single-phase or two phase consumers (eg, lighting, electronic devices), which
introduced unbalance in the system.

Given the presence of these ones, the system is unbalanced.

Now, what is a balanced system:

1. Balanced three-phase voltage system: a system where the voltage magnitude of the
three phases is the same and the voltage angle is shifted 120º.
 2. Balance three-phase current system: a balanced 3Ph voltage system where power is
delivered to a balanced consumer, making the currents have the same magnitude in
each of the fases and a curreng angle shifted 120º.

3 Simulation of Gas Networks

3.1 Unknowns in gas networks

The significant unknowns in a gas network are:

1. Velocity: This variable indicates how fast gas goes from the beginning to the end of the
pipe. This is comparable to the current density in electrical circuits.
2. Pressure: To be more specific, this is the pressure difference of gas. For instance, if you
have a high-pressure difference, the resulting velocity will be high; the velocity will be
low if you have a low-pressure difference. This is similar to the voltage in an electrical
circuit.
3. The temperature of the gas

It is essential to keep in mind a coupling between the parameters. The pressure can be coupled
with temperature, and velocity can also be connected with temperature. Hence, it is required to
use different quantities in gas simulation.
3.2 Gas types

There are different types of gases available. Table 1 analyzes a few gases based on their
calorific value and density. The calorific value indicates how much heat is produced when you
burn it.

Table 1. Different gases and their properties

As we can see from table 1, each gas is distinct. Hydrogen is much lighter than all the above-
compared gases but stores the least energy. To use these gases in a house, a city, or an industry,
we must remember that most burners or furnaces are based on specific calorific values. If the
calorific value of the supplied gas differs, we have to mix it with nitrogen to get it back to the
original value. Therefore, the combining should be accounted for in the simulation as well.

3.3 Gas Simulation

If we start with the best mathematical model for any gas transport, we will end up with the
Navier Stokes equation. This equation is given below:

This provides an accurate solution but is very complicated to solve, especially for massive
networks. For instance, the Navier-Stokes equation cannot solve the gas network of the
entire Netherlands.

What appears to be sufficient is the 1D equation. The Navier Stokes equation is in x, y, z

directions, the three spatial directions, and time. If you have only an equation in the length of
the pipe and time, it is already sufficient to have an excellent approximation. This means that
we do not have a complete description of the pipeline. But, this is sufficient for most
applications, except if we have to zoom in for details.

3.4 Components to be modeled

1. Pipeline resistance: This depends on the material that the pipe is made of. It is
measured by how much resistance the gasses experience when coming from one end
to the other.
2. Pipeline connection: This is a crucial component of the model. Things can become
difficult at the connection point, but we generally don’t use the Navier-Stokes
equation. Instead, we see it as a sort of hub and apply the conservation property,
which states that what’s coming in is equal to what’s going out. This is generally
sufficient for modeling.
3. Compressors: A pressure difference is needed to transport gas. To make the pressure
difference, we need to include compressors. They are more complicated to describe.
We don’t define them in much detail, but we treat them as black boxes. Another
complication is that they should be powered. This can be done using electricity or gas.
In this latter case, gas comes in, drives the machine, and then increases the gas’s
pressure to a higher value to obtain a high-pressure difference.

3.5 Gas vs. electricity

Although similar in many ways, there are also some fundamental differences between gas and
electricity. They are:

 Gas is much slower than electricity. Electricity can flow in microseconds from one
part of the Netherlands to the other. But for instance, if gas is transported from
Groningen to Rotterdam, it may take half an hour before it reaches Rotterdam.
 Gas is more dangerous than electricity. If something goes wrong in the electricity
grid, it is generally a local phenomenon. For example, a short circuit is obviously
hazardous, but its impact is limited because it can quickly be isolated. With gas, this is
more complicated. If there’s a lot of gas leaking and a spark occurs, you get an
explosion that can blackout a part of the city. So we could say that gas is hazardous
when compared to electricity.
  To store electricity is difficult, but storing gas is relatively easy. We can store gas in
two ways:

1. Preserve energy in the pipeline: If you increase the pressure, more gas will be
present in the pipeline, acting as a buffer.
2. Transformed to gas: The surplus of energy generation, for instance, by excess
wind, transformed into a gas, like hydrogen. That gas can then be stored and
reused in the future when there is excessive demand. This property enables gas
and electricity infrastructure coupling.

3.6 Gas-Electricity coupling

There are multiple ways the gas infrastructure is coupled to the electrical network:

1. Transport both electricity and gas in parallel. This allows you to choose between
electricity and gas.
2. Compressors: This is used in gas transportation lines. The electricity can be used to
drive the compressor, which leads to gas propagation.
3. Steam turbines: gas can be converted into electricity using steam turbines. This is
done in furnaces.
4. Power-to-gas method: Here, electricity can be used to produce hydrogen that is stored
in the container. Then, when needed, the gas can be used to produce electricity. So,
there’s coupling from power-to-gas and also from gas-to-power.
5. CHP plant: This stands for combined heat and power plant. Here, you can see that all
three sorts of energy are present: electricity, gas, and heat. We start with gas, feed it
to the CHP system, which then produces both electricity and hot water.

3.7 Energy Hubs

To combine different forms of energy, we use energy hubs which are black boxes. In this, on
the one hand, electricity can go in and out. On the right-hand side, gas can enter and leave. To
model this, we have to include specific conservation laws. What happens inside the box is not
essential when simulating; it is only necessary how much is lost in the transformation. So if
you go from electricity to gas, you’ll lose something, which also holds the other way around.
There is a specific efficiency involved. It can be 90% or 95%, but you always lose something,
which should be considered.

Importance:

1. We don’t have to make a detailed model of this energy hub. Instead, we only treat it
as a black box.
2. The interaction between the different energy networks can be simplified, making the
integration more straightforward.

3.8 Summary

 Velocity, pressure, and temperature as unknowns, and the resistance, connection

points, and compressors as elements are required to simulate a gas network.
 Gas is slower, more dangerous, and has higher storage potential.
 There are different possible forms of combining gas and electricity: transporting them
parallel, using compressors, generating electricity with steam turbines, and power-to-
gas technology.
4 Easy and Complex Models

4.1 Generator model

Consider a generator. We can simulate it with a simple model as well as a complex model.

For instance, in a simple model, we can look at the differential equations that represent the
synchronous generator. while complex models enables to simulate a generator with a
comprehensive model to describe its behavior accurately. Based on the application, we can
define the level of complexity.

Examples

1. Expansion planning model: Here, we want to find out which types of wind turbines or
which types of pipelines we need in, say, ten years.

Usually, it is not needed to have a detailed description because we are only interested in the
general trends. For instance, we want to know how the consumption is increasing, linearly or
quadratically, so that we can invest appropriately in offshore wind farms. Then, it is sufficient
to have only a global overview, and no detailed model is needed. Therefore, grid planning
generally requires simpler models.
2. System operation: Here, for example, we are interested in seeing the impact of an event on
the grid dynamics. So, we must know what happens exactly, not only now but also in five or
ten minutes. Thus, the model should be more accurate. Therefore, grid operations demand
more complex models.

4.2 Power flow models

There are two models for computing the power flow in the electricity network.

4.2.1 AC model:

 It leads to a non-linear system, which means that it is also more difficult to solve.
 Inputs: generator power and voltages and the demand
 Output: Current and power flow in the lines, the voltage at the nodes, the power losses
in the system, the generator reactive power setpoints and the power output at the slack
bus.
 This model is complicated but comprehensive.
 accuracy of the AC model is higher
 more challenging to solve
 more computational power and CPU time
 requires complicated iterative solvers to solve this non-linear model efficiently

4.2.2 DC model:

 This model takes fewer inputs and has fewer outputs.

 we get a linear model that can be solved easily to give a rough estimate of the power
flow
 DC model is simpler.
 not accurate enough.
 relatively easy to solve.
 we can use more straightforward mathematical solvers.

Both the easy and complex models have their own advantages and disadvantages.
4.3 Combining models

It is possible to combine AC and DC models and thereby having the accuracy of the complex
model and the speed of the simple model together. It is done in the following steps:

1. We start with the inaccurate model.

2. we solve this model.

3. We use that solution as the starting solution of the accurate model.

4. Afterwards, we check if that solution is accurate enough.

5. If not, we can reiterate: we go back to the inaccurate model, solve it again, plug it in the
accurate model etc.

6. If the solution is good enough, we can stop.

We can not only use this combination of models in simulation but also in optimization. In
optimization, we try to find an optimal solution to a problem. It is not always needed to start
the optimization with an accurate model. We can start with the inaccurate model, try to
optimize that, plug that solution in the accurate model and reiterate. This way, we can get a
good enough solution in a fraction of the time.

4.4 Challenges of models

There are three challenges:

1. Performance: Some models are easier to solve than others. We have to make sure that
the models we are using are solvable within an acceptable time limit.
2. Interdisciplinary: such problems or models are interdisciplinary. It is needed to have a
clear view of the different components and closely cooperate between them. On the
one hand, we need to consider the physics and electric laws governing physical
behavior. On the other hand, computer science develops good solvers and efficient
algorithms that can be implemented on modern hardware. And mathematics is at the
core of all this.
 3. Stopping criterion: This is inherent to the iterative way of solving models. How do we
know when to stop? When do we consider the solution to be accurate enough? The
answer is not straightforward and should be tuned to the problem.

4.5 Summary

 Different applications require different types of models, some more complex than
others. Planning uses only general trends, while operations require an accurate system
model.
 There are differences between the AC and DC models, both with their advantages.
The AC model is more complex but more accurate, while the DC model is easier to
solve.
 There are methods that efficiently combines easy and complex models to get the best
of both worlds.
 There are three main challenges: ensuring good performance, close cooperation
between disciplines and stopping the simulation.