Modelling of Hydro-Power System

System description
Overview
A hydro power system is an human-made structure or arrangement of structures usually built
with the main purpose of generating electric power, though it may also be used for recreation,
flood mitigation or other reasons.
This system can be composed by lakes, reaches, dams, pumps, valves, turbines, ducts which
connect the elements. Some of these elements must be controlled, either manually or
automatically, in order to generate electricity while avoiding and minimizing the chances of flood
or other accidents, such as dam breaking. For that, some constraints are usually imposed at the
levels of the lakes and reaches, as well as on the flow rater of pipes and rivers.
In this series of post, we will see how to develop a model for an academic case of a hydro-
power system, how simulations can be used to predict possible scenarios in such system, and
how automatic control techniques can be used to keep track of certain properties of the system
and avoid problems, or optimize energy production, for example.
The hydro-power system
For the sake of simplicity, we will consider a system formed by 3 lakes and 3 reaches as shown
in the figure below

A lake is basically a reservoir of water, i.e a big tank with one or more inlets and outlets, where
water can accumulate (if there is more water entering the lake than leaving it). Such lakes may
be artificial (man-made) or natural. Artificial lakes are usually created by deviating a water
course or by excavation.
The lakes are interconnected using ducts, each one containing a controllable valve, turbine or
pump. These elements can be used to change the local flow rates.
A reach, as defined by USGS, is basically a river section with similar hydraulic attributes, such
as discharge, area and slope. There are other possible definitions but we will stick to this one.
Thus, a river with length of 100 km may be interpreted as a single reach if its properties are
more or less uniform, or it may be interpreted as multiple reaches, if its properties vary
significantly. In some cases, it may be worth to split a river into multiple reaches even when its
properties are uniform, since this approach can help in the modelling process.
The following table summarizes all the components of the system as shown in the figure above.
Element Description

L Lake
 Element Description

P Pump

T Turbine

R Reach

V Valve

D Dam

In Inflow

Out Outflow
Definition of variables and assumptions
To simplify the model, we can take the following assumptions:
Negligible dynamics of valves, pumps and turbines. This may be assumed considering that the
fast dynamics of such elements when compared with the flow rates from the interconnected
elements, thus the flow rates are equal to their own set points;
Since the objective is to control electricity generation, while meeting some constraints, it’s
important to define which variables of the system are controlled, which are manipulated and
which are disturbances (uncontrollable inputs).
Controlled variables
 flow rates in ducts and reaches,
 levels of lakes and reaches
 generated/consumed electric power

Manipulated variables
 opening rate of valves, in terms of flow rate reference (ui);
 opening rate of turbines, in terms of flow rate reference (ui);
 imposed flow rate of pumps, in terms of flow rate reference(ui);
Disturbances/ Uncontrolled variables
 water inflow due to upstream section
 water inflow due to tributaries

Theoretical model of the lakes

The main use of these lakes is to work as a battery, i.e energy storage. At times when energy is
at low demand, water can be pumped from the river to the lakes and stored there as potential
energy. When there is a greater demand and only the level of water from the river can not
provide enough energy, water can flow from the lakes to the river through turbines, using its
kinetic energy to generate extra energy. Notice that this process is, in a real-world case, not
100% efficient. This means that more energy is needed to pump the water from the river to the
lakes, then can be extracted by turbining from the lakes to the river. Yet, it can be a useful
technique to keep the balance in energy generation, and also to redirect volume of water when
its excess can cause floods downstream.
Mathematical representation of the lakes
The water stored in the lake changes according the inflow and outflow rates. As the volume of
water decreases, the level also decreases. In essence the mass conservation is the basic
equation to describe the lakes:

dmdt=win(t)−wout(t)

Where m is the mass of water (kg), t is the time (s), win is the mass inflow rate (kg/s) and win is
the mass outflow rate (kg/s). The above equation can be rewritten as:

d(ρhA)dt=ρqin(t)−ρqout(t)

Where ρ is the density of water, h is the water level and A is the cross-section area of the lake.
Since any liquid can be reasonably considered incompressible (no change of volume with
pressure), the density ρ can be considered constant, and thus cancelled out. The cross-section
area A may be constant (like a cube) or it may be a function of the water level. For better
generalization.

let’s say A=A(h), thus the final equation is:

d(hA(h))dt=qin(t)−qout(t)

The above equation is an ordinary differential equation, relating the rate of change of volume
(hA(h)) with the inlet and outlet flow rates. It can be solved using numerical integration, if all the
other variables (qi) are known

Power generated/consumed by pumps and turbines

The power p generated/consumed by pumps and turbines is directly proportional to the flow
rate q and the difference in water height H upstream and downstream. The following equations
describe this relation in a simple form:

H=hup−hdown
p=K⋅q⋅H

Where K is constant of proportionality, which can be referred as the pump/turbine coefficient

(positive for turbines and negative for pumps).

Pipes and valves

The connection between lake 1 and 2 is made through a valve. The discharge through this
element can in general be modelled by a non-linear relationship with the difference in height
upstream and downstream H as:

q=−sign(H)⋅A⋅2g|H|

Where A is the duct section.

Analytical solution of lake storage with sinusoidal inflow and constant
outflow (pump)
It is always useful to have some analytical solution of ODE problems to compare how good are
numerical solutions obtained in a later step. Coming back to the mass conservation:

d(hA(h))dt=qin(t)−qout(t)

Let’s consider a very simple lake in the form of a cube. Thus, the cross section area is constant.
A≠A(t)
The equation simplifies to:

dhdt=qin(t)−qout(t)A

Say the outlet is regulated by a pump, with a constant flow rate of qout, and the inflow is a
sinusoidal flow with the shape, provided by a pump.

qin(t)=Aq+Bqsin⁡πtCq
dhdt=1A[Aq+Bqsin⁡πtCq−qout]
Call A∗=Aq−qout
dhdt=1A[A∗+Bqsin⁡πtCq]
Integrate it.
∫dh=1A∫(A∗+Bqsin⁡(πtCq))dt
h=A∗tA−BqCAπcos⁡(πtCq)+Const

Which gives us the general solution to this problem. Now let’s fix some numerical values for
simulation.
 A=1
 qout=5
 Aq=5
 Bq=2
 Cq=1

h=−2πcos⁡(πt)+Const
Apply initial condition t=0, h0=0
Const=2π
The final analytical solution is,
h=−2πcos⁡(πt)+2π

The equation above is used to calculate the water level profile of the lake for any time t≥0, as
shown in the Figure below.
Figure 1 - Water level profile for first case.

Another analytical solution, with variable cross-section

Let’s perform a similar analysis, as the one shown above, but now using a lake which has a
variable cross-section area. Say that the cross-section area follows the pattern below:

A(h)=Eh2

Where E is a constant value. Let’s perform the same analytic integration process that was done
above.
(Eh2)dhdt=1A[A∗+Bqsin⁡πtCq]
Integrate it.
∫(Eh2)dh=1A∫(A∗+Bqsin⁡(πtCq))dt
h=[3AEA∗t−3AEBqCqπcos⁡(πtCq)+Const]13
Which gives us the general solution to this problem. Now let’s fix some numerical values for
simulation.
 A=1
 qout=5
 Aq=5
 Bq=2
 Cq=1
  E=1
h=[3EA∗t−3EBqCqπcos⁡(πtCq)+Const]13
Apply initial condition t=0, h0=0
0=[−3EBqCqπ+Const]13
(0)3=([−3EBqCqπ+Const]13)3
3EBqCqπ=Const

Therefore the specific solution is,

h=[3EA∗t−3EBqCqπcos⁡(πtCq)+3EBqCqπ]13

Substituting the values here we find that,3EBqCqπ≈1.91

The final analytical solution is,

h=[−32πcos⁡(πt)+1.91]13

The equation above is used to calculate the water level profile of the lake for any time t≥0, as
shown in the Figure below.

Figure 2 - Water level profile for second case.

 Theoretical model of the reaches
a reach is a section of a river, with more or less uniform geometry, such that it can be
interpreted as one component in the complete model.
The state of the art way of modelling this dynamics is the De Saint Venant Equations. It consists
in a pair of nonlinear partial differential equations (PDE) that represent one dimensional
hydraulics. One equation describes the mass conservation while the other describes momentum
conservation.
Mass conservation of the reaches
Let’s start by the mass conservation equation for the reach. As shown for the lake, the
accumulation of mass m increases with the mass entering the reach win and decreases with the
mass wout leaving it , such that:
dmdt=win−wout
Which can be rewritten in terms of volume V and density of water
ρ,d(ρV)dt=ρqin−ρqout
Where q is volumetric flow rate (e.g m3/s). The liquid water can be considered an
incompressible fluid, so the density is constant and can be taken out of the differential term.
Additionally, the volume V can be written in terms of the the reach geometry, which are the
wetted area A and the length of the reach
xd(Ax)dt=qin−qout
The reach length x does not depend on time and it can be taken out of the differential term.
dAdt=qin−qoutx
For an infinitesimally small length of the reach, the above ordinary differential equation (ODE) is
written as a PDE:
∂A∂t=−∂q∂x
Or the most known form of this equation, incorporating the possibility of a lateral inflow in the
reach is,
∂A∂t+∂q∂x=q
latIt’s important to note here that the wetted area A is dependent on the geometry of the reach
and on the amount of water. By knowing the relation between this variable and the water level h,
it’s possible to calculate one from the other.
With the mass conservation well defined, let’s move on to the momentum conservation.
Momentum conservation of the reaches
According to Newton’s second law, the time rate of change of the linear momentum of a system
must be balanced by the sum of all external forces acting on/by the system. This property
directly affects the velocity, since the momentum is by definition the product of mass and
velocity of a system. It can be written as:
∑F→=∂∂t∫CVρv→dV+∫CVρv→(v→⋅n→)dA
The first term on the right-hand side of the above equation represents the time rate of change of
momentum “stored” in the control volume CV, where (→v) is the velocity of the system. The
second term is the net outflow of momentum across the control surface.
The net outflow of momentum is given by the difference in moment outgoing and moment
incoming in the control volume. The inflow is defined as:−ρ(Q+qdx)The inflow is translated into
momentum by multiplying it with the velocity. A correction factor β is normally also incorporated
into the equation, due to non-uniform velocity distribution across the reach cross-
section.∫inletρv→(v→⋅n→)dA=−ρ(βvQ+βvxqdx)Where vx is the (averaged) lateral inflow. The
factor β is calculated as:β=1v2A∫∫v2dAAs mentioned by Gündüz, O., in most
cases 1.01≤beta≤1.33, where lower values are used for reaches with geometry that resembles
a prism, and higher values for those like river valleys with floodplains.
The moment outflow is calculated from the mass outflow of the reach. Since it is known that this
is a function of mass inflow, Taylor series expansion may be used, and the moment outflow
is:∫outletρv→(v→⋅n→)dA=ρ(βvQ+∂(βvQ)∂xdx)Putting moment inflow together with outflow we
have:∫CVρv→(v→⋅n→)dA=−ρ(βvxQ−∂(βvQ)∂x)dxRegarding the first term on the right-hand side
of the momentum equation, it represents the time rate of change of momentum stored in the
control volume . It is associated with the elemental lenth x, and can be written
as:∂∂t∫CVρv→dV=ρ∂Q∂tdxFor the left-hand side of the momentum equation, the 5 forces that
may act on it are:
 gravity force Fg because of the weight of water.
 pressure force Fp due to atmospheric and hydraulic pressure
 friction force Ff, due to water contact with the bottom and the sides of the reach.
 contraction/expansion force Fe, related to drastic changes in the reach cross-section.
 wind shear force Fw
Gravity Force
Hydrostatic pressure is given by the column formed by water.dm=ρdV=ρAdxThe weight
force W is the mass multiplied by the gravity acceleration.W=gdm=ρgAdxWhere g is the
gravitational acceleration. For momentum , the relevant component of weight is the one in the
direction of the flow, which we interpret here as the gravity force Fg.Fg=ρgAdxsin⁡θWhere θ is
the inclination of the channel bed.
For small inclinations of the reach, it can be assumed that,sin⁡θ≈tan⁡θWith that, the gravity
force can be written as,Fg=ρgAdxS0Where S0 is slope of the channel bed.
Pressure Force
The pressure force Fp results from a balance between the hydrostatic pressure exerted on each
side of the control volume (Fpl at the left and Fpr at the right), as well as the pressure exerted
by the banks Fpb, such that:
Fp=Fpl−Fpr+Fpb
It can be shown (I will skip the derivations here) that this expression can be expanded to:
Fp=−(ρg∂y∂xA+∫0yρg(y−w)∂b∂wdw)dx+(∫0yρg(y−w)∂b∂xdw)dx
Where y−w is the depth that the element of water is immersed, which gives the hydrostatic
pressure ρg(y−w), and b is the width of the element across the reach.
Simplifying and rearranging, it can be seen that the resultant pressure force is expressed as:
Fp=−ρg∂y∂xAdx
Friction Force
In any real reach with rough bottom and sides, mechanical energy is transformed into heat
through shear stress τ. For a steady uniform flow, the shear stress can be written as:
τ0=ρgRSf
Where R is the hydraulic radius.
R=A/P
Where P is the wetted perimeter and Sf is the friction slope. This last one comes from Manning
Equation, and defined as,
Sf=n2v2μ2R4/3
Where n is Manning’s roughness coefficient, v is the flow velocity and μ is a constant to convert
between SI units (μ=1.0) and Imperial units (μ=1.49).
For an unsteady, non-uniform flow, the shear stress τ can be written in terms of τ0.
τ=−τ0Pdx
The final form of the friction force can then be defined as:
Ff=−ρgASfdx
Contraction/expansion Force
Drastic changes in the reach geometry causes energy losses due to turbulence. These forces
can be related to the friction force, with the difference that the friction slope Sf is replaced by the
energy grade slope Se, representing such loss due to contraction/expansion.
Se=Ke2g∂(Q/A)2∂x
Where Ke is a coefficient that is negative for expansion and positive for contraction. The final
equaion of contraction/expansion force Fe is very similar to the friction force.
Fe=−ρgASedx
Wind shear Force
Wind blowing against the free surface of the water generates a shear stress τw. The greater the
shear stress, the greater the wind shear force. The wind shear force is also a function of a wind
shear factor Wf and the water density, thus we can write:
τw=−ρWf
where:
Wf=Cf|vr|vr2
Here, Cf is a constant coefficient related to the shear stress and vr is the relative velocity of the
water with respect to the boundary. It can be defined as:
vr=QA−vwcos⁡ω
Being vw the velocity of wind, and ω is the angle that such velocity makes with the average
water velocity. The final form of the wind shear stress, considering that τw=τwBdx is:
Fw=−ρWfBdx
Final form of the momentum equation
Joining together all the pieces described above and simplifying, the momentum equation can be
written in the following form:
∂Q∂t+∂(βQ2/A)∂x+gA(∂h∂x+Sf+Se)−βqvx+WfB=0
Summary
The De Saint Venant equations consisted in a 1D unsteady hydrodynamic model which
describes the mass and momentum conservation of a system (in the present case, a reach or
river section). These two equations are always addressed together. In the differential form they
read:
∂A∂t+∂q∂x−qlat=0
∂Q∂t+∂(βQ2/A)∂x+gA(∂h∂x+Sf+Se)−βqvx+WfB=0
For a steady flow, the derivatives with respect to time becomes equal to zero, so for steady
flow:
∂q∂x−qlat=0
∂(βQ2/A)∂x+gA(∂h∂x+Sf+Se)−βqvx+WfB=0