**Generator Placement Methods In Network Reduction**

**Introduction**

In network reduction, after generating the reduced network, the next step is to place the external power injections: generators and loads. In this work, the goal is to find a generator placement method which can be integrated in the network-reduction process and yield good accuracy in terms of matching the full-model dc OPF results. In this work, the metrics used for accessing accuracy of matching the full-model dc OPF results are bus LMP, generation dispatch and the total cost.

Placing external load is less complex than placing generators since a load does not have an “identity,” unlike generators which have individual real and reactive power limits and production cost curves. One can split the load and distribute the fractional load across the reduced network in order to match the branch flows in reduced model to those in the full model.

For planning or market studies conducted by the (E4ST) application, formerly known as the SuperOPF, generators need to be moved whole. It is impractical to use fractions of the external generators, which are generated in the traditional Ward reduction for at least two reasons. First, Ward reduction distributes the external generators to all boundary buses.

This process will generate a huge number of fictitious generators in the reduced model. Consequently, in a large system where significant bus reduction takes place, there will be a large number of variables related to the splitting of the generators. Second, for each external generator, the power output of each fraction may not be scheduled independently in the planning study. Outputs of fractions of one external generator must always hold a fixed proportional relationship and their sum must be constrained to operate within the generator’s capability. Consequently, there will be a large number of constraints involved in the dc OPF study related to the generator fractions. Thus a reduced model with generator fractions may be more complex the than the original model, defeating the purpose of network reduction.