Multi-Scenario Voltage Optimization using Load Profiling

The load profiling method and the coordinated dispatching of active-reactive power are combined for voltage optimization in this paper. Load profiling aims to find out the typical pattern of the customs and has been widely used in the demand-side response, where only the active power is considered and it is usually a one-dimensional time series. With both active and reactive power considered, the load pattern then includes multiple angles of information in its scopes. To reveal such knowledge into the system operation, a machine learning based classification method combining the supervised and unsupervised learning can identify the typical load pattern on each node from big data by classifying them into a small number of groups. These typical patterns can therefore form the boundary condition for the voltage optimization, which can be sampled using the Monte Carlo simulation to form multiple scenarios. Then, a more accurate operation model of coordinated dispatching of active-reactive power is introduced in the voltage optimization considering the renewable energy source, the energy storage system and the controllable load. One step further, such a multi-scenario operation method can be used to facilitate power system planning, such as the voltage evaluation and the reactive power compensation, as this study briefly outlooks.