It is the same placementfrom Figure 2. Note that this is simply for interest, it is hard to directly compare computation times as differentcomputers were used as well as different MATLAB versions. To further explain this, a walkthrough will be done for the IEEE node network. The other noticeable difference is the number of nodes for each network. This problem can be formulated using 2.

To understand these steps, a 9-node network has been created as seen in Figure 3. Due to this synchronization ability, they are one of the most important devices inpower system monitoring and control [10]. The first node it would check is node 3. It is comprised of thirteen buses and oneswitch. Previous works have only focused on one or two of these items. A typical power system. Note that this method is a little more complex than the summarized steps listed and the fullmethod can be read in [29].

## Optimal Phasor Measurement Unit Placement for Monitoring of PEA Bowin Power

After testing thesecombinations, the number 8 would either become the new upper limit or the new lower limit,and another iteration would pladement done. Pu to take this factor into account, a straightforward solution is proposed. This idea is best illustrated with an example,which will be an expansion from the Network Reconfiguration section in Section 2. Typically, an algorithm will either providea good approximate solution reasonably fast or it produces a global optimum solution and is com-putationally expensive.

# Optimal PMU Placement with Uncertainty Using Pareto Method

This is illustrated in Figure 1. This calculation comes out to be 7, The steps of thiscalculation are summarized below: When placed at a node, it is capable of measuring the voltage phasor at the node and all incident current phasors to the node.

The most recent notable conference took place in and was held in Paris, France. The end result is the same forboth papers. Some of the benefits are highlighted below [16]: As the topic of19model order reduction or state estimation is not relevant to this thesis, it will not be described.

If complete observability is not achieved, go to step 1. It is the same placementfrom Figure 2. However, as previously stated, there is a tradeoff that makes this placement configuration prefer-able.

The distribution system needs to be monitored in real time so that minor issues relating to grid stability can be noted and fixed before they cascade into system failure. Thisfurther illustrates why a high SORI value is extremely important.

This corresponds to row 2. A sequential increasing method is carried out here for PMU placement and its results were compared with proposed method too.

Their system andplacement method for the node feeder can be seen in Figure 2. Once complete observability has been achieved, the greedy algorithm part thesjs the proposedalgorithm finishes.

Again, it can be seen that the SORIvalues from the proposed algorithm either match or exceed the ones from other works, excludingthe ones for the node feeder.

Deterministic algorithmsfollow a strict path. Is complete observability achieved? These findings are summarized in Table 2. The second type of nodes they removed from the search space were the nodesconnected to end nodes. The aim was to have a simple, single input that would make it easy to implementin the real world. Their placement result for the node can beseen below in Figure 2.

This hybrid greedy algorithm is comprised of five parts: If complete observability is not achieved, go to step 1.

Incomplete observability occurs when these conditions are not met. Therefore, only one phase was considered. These important topics are imperative for understanding certain aspects of theliterature review as well as the proposed solution later in the thesis. Lastly, network reconfiguration was not considered in most papers, even though this is animportant factor to consider since distribution networks can be reconfigured.

Again, it can be seen that the SORIvalues from the proposed algorithm either match or exceed the ones from other works, excludingthe ones for the node feeder. The same process was usedfor each network and the three different cases.