As the amount of uncertainty in online power system operations grows, new methodologies need to be devised in order to timely monitor and control the power grid. In this work, novel techniques for online voltage stability margin monitoring and control have been developed with a focus on reactive power reserves.

The maintenance of adequate reactive power reserves (RPRs) is a critical step in avoiding a voltage collapse. A thorough investigation of the relationship between different definitions of reactive power reserves and how they are related to voltage stability margin (VSM) is performed.

Multi-linear regression models are used to relate RPRs and VSM. Several operating conditions and a significantly large number of different network topologies, including NERC category B, C and D outages are considered as well. A classification tool is then developed in order to identify which regression model needs to be used based on system conditions and network topology. The approach is tested in the IEEE 30 bus test system and in a reduced case of the eastern power system interconnection of the United States. Results have shown that the approach can monitor voltage stability margin in real time based on the amount of system wide reactive power reserves.

In case degenerative system conditions are identified, control actions need to be put in place to increase the amount of RPRs and system VSM. A novel control method is proposed here in order to identify the location and amount of control necessary to recover RPRs, VSM and to remove existing voltage violations. The approach is based on the identification of a critical set of generators that, if exhausted, will directly contribute to a voltage collapse.

Potential control actions are investigated in order to recover those critical reactive power reserves, namely: active power re-dispatch, capacitor switching, active and reactive power load shedding. The effectiveness of each control variables on RPRs is calculated using reactive power reserve sensitivities, a concept introduced in this work. Once these sensitivities are calculated, the problem of recovering RPRs and VSM is formulated as convex quadratic optimization problem with a reduced dimension.

Results on the IEEE 30 bus test system and the IEEE 118 bus test system are used to illustrate the efficacy of the approach.