# Projections#

When the control algorithm is trying to find the best control for given voltage constraints, it
could find a solution that is not “feasible” by the load. This means that the active and reactive
powers \(P\) and \(Q\) that constitute the solution lie outside the feasible domain defined by a part
of the disc of radius \(S^{\max}\) in the \((P, Q)\) space. In these cases, the solution has to be
projected into the feasible domain. We can choose how the projection is performed using three
available projection types:
the *Euclidean* projection, the projection at *Constant \(P\)* and the projection at *Constant \(Q\)*.

The projection accepts two approximation parameters: `alpha`

and `epsilon`

.

`alpha`

is used to compute soft sign function and soft projection function. The higher`alpha`

is, the better the approximations are.`epsilon`

is used to approximate a smooth square root function:\[\sqrt{S} \approx \sqrt{\varepsilon \times \exp\left(\frac{-{|S|}^2}{\varepsilon}\right) + {|S|}^2}\]The lower

`epsilon`

is, the better the approximations are.

Important

Please note that no projection is performed if the final \(\underline{S(U)}\) point lies in the disc of radius \(S^{\max}\).

## Euclidean projection#

A Euclidean projection on the feasible domain. This is the default value for projections when it is not specified.

```
from roseau.load_flow import Projection
projection = Projection(type="euclidean") # alpha and epsilon can be provided
```

Important

Please note that using the Euclidean projection may reduce the provided \(P^{\mathrm{th.}}\) and \(Q^{\mathrm{th.}}\) of the load. See the Feasible Domain page for more details.

## Constant \(P\)#

Keep the value of \(P\) computed by the control and project \(Q\) on the feasible domain.

```
from roseau.load_flow import Projection
projection = Projection(type="keep_p") # alpha and epsilon can be provided
```

Important

Please note that using the *Constant \(P\)* projection may reduce the provided \(Q^{\mathrm{th.}}\) of the load. See the
Feasible Domain page for more details.

## Constant \(Q\)#

Keep the value of \(Q\) computed by the control and project \(P\) on the feasible domain.

```
from roseau.load_flow import Projection
projection = Projection(type="keep_q") # alpha and epsilon can be provided
```

Important

Please note that using the *Constant \(Q\)* projection may reduce the provided \(P^{\mathrm{th.}}\) of
the load. See the Feasible Domain page for more details.