Diagnostics and Prognostics Using Temporal Causal Models for Cyber Physical Systems – A Case of Smart Electric Grid
Abstract:
Understanding failure in protection system using Temporal Causal Diagrams Situational awareness in large electrical systems and eventual fault source identification and possibly prognostics is a very challenging problem. State of the art relies on a network of protection devices that include relays to detect anomalies and engage circuit breakers to isolate the faulty component(s) in the system. These protection devices are designed to mask the fault effects locally, and sometime can be extended for wide area protection. Failure in protection system itself can lead to malfunction or wrong switching of breakers. It is important to analyze the protection failure events in a global context to improve the decision making for the protection engineers.
Our team has developed a modeling formalism called Temporal Causal Diagrams (TCD) that can be used to describe failure propagation in an electrical subsystem including the discrete timed behavior of the electrical components, especially protection relays. These TCD models can be connected together to capture the interactions between protection systems, and automatic controls using available information from the physical and the cyber components of this system. Models also capture the cascading effects of such behaviors, including their impact on the failure propagation through internal mode changes. This approach yields a natural, multi-level reasoning scheme, wherein a component or sub-system’s discrete timed behavior can be simulated that can be used to generate the alarms, mode-changes and event traces for single and multi-fault scenarios. More importantly, it can be used to generate data that can be used to test, validate and improve the quality and performance of the TCD based reasoner. Further the richness of the collected data set can be improved by integrating the discrete event simulation model with other physics-based dynamic models that can simulate the nominal and faulty operation of the physical plant. In this demonstration, we will show the integrated Simulink model that implements the discrete timed behavior of the protection devices for a two transmission line system shown in the figure above. It includes three substations (SS1, SS2, and SS3) and two transmission lines (TL1, TL2). Transmission line TL1 carries power between buses BU1 and BU2 while transmission line TL2 is between buses BU2 and BU3. The power sources exist on both sides of the transmission lines. Each transmission line has two breakers and two distance relay for protection.
Fault diagnostics in power distribution systems using Smart Meters data
Traditionally, power distribution systems consist of feeder switching and protective devices such as reclosers, sectionalizers, manually operated switches, fault current indicators and capacitor banks, with limited communications network for monitoring and remote operations. A shortcoming of such systems is the need for field crews to patrol feeders for identifying the fault location when a fault happens. The task is time-consuming. To overcome this hurdle, utilities are exploring the benefit of smart meter data for reducing the outage duration by fast detection of the fault location. The function of the last "gasp" of smart meters when smart meters are out of service can be fully utilized for outage management. The status of protective devices, such as reclosers, reclosers and fuses, can be estimated by using smart meters data. The location of fault, therefore, can be determined using the status information of protective devices. Following this idea, our team has proposed and implemented the logic reasoning rules to detect the status of protective devices in power distribution systems using smart meters data for identification of the fault location. This can be extended for detecting failures in protection devices or operation in an abnormal condition with hypotheses and reasoning algorithms. The general procedures of using smart meters data are: 1) implement the Tree technique to model the spatial connections of protective devices, which can be adopted to power distribution systems with different configurations; 2) logic reasoning rules are applied on the Tree to find the candidates of protective devices which may have acted to isolate the fault; 3) the corresponding logic rules to pinpoint the specific device which should have acted to isolate the fault. Under contingency (abnormal conditions), normally open switches may be closed for load transferring which results in a loop system and mis-coordination of protective devices along primary feeders. To enhance the capability of logic reasoning rules to deal with loop systems, a modified Tree method and corresponding logic reasoning rules areproposed and implemented. The aforementioned work serves as the foundation to build the Temporal Causal Diagram (TCD) for fault diagnostics and propagation in power distribution systems.