Testing smart grid information and communication (ICT) infrastructures is imperative to ensure that they meet industry requirements and standards and do not compromise the grid reliability. Within the micro-grid, this requires identifying and testing ICT infrastructures for communication between distributed energy resources, building, substations, etc. To evaluate various ICT infrastructures for micro-grid deployment, this work introduces the Virtual Micro-Grid Laboratory (VMGL) and provides a preliminary analysis of Long-Term Evolution (LTE) as a micro-grid communication infrastructure.

This paper introduces a new method for performing actuator fault detection and diagnostics (FDD) in heating ventilation and air conditioning (HVAC) systems. The proposed actuator FDD strategy, for testing whether an actuator is stuck in a single position, uses a two-tier approach that includes a dynamic model-based detector and a fast-deciding steady-state detector. The model-based detector is formulated to provide detection performance that asymptotically bounds both the probability of miss and probability of false alarm. To provide a quick confirmation the actuator is working, the steady-state detector utilizes a goodness-of-fit detection strategy to decide if the measurements could be described by an actuator failure. An architecture is introduced that requires multiple steady-state detection experiments to decide that the measurements could be explained by an actuator failure before performing model-based detection. An experimental test bed using a the KTH Royal Institute of Technology campus HVAC system is described and used to evaluate the steady-state and model-based detectors. The experimental test bed is utilized to identify a building dynamics model, that is employed through monte carlo analysis, to characterize the detection performance of both the model-based detector and the steady-state detector.

This paper addresses the issue of detecting and isolating topology attacks in power networks. A topology attack, unlike a data attack and power injection attack, alters the physical dynamics of the power network by removing bus interconnections. These attacks can manifest as both cyber and physical attacks. A physical topology attack occurs when a bus interconnection is physically broken, while a cyber topology attack occurs when incorrect information about the network topology is transmitted to the system estimator and incorporated as the truth. To detect topology attacks, a stochastic hypothesis testing problem is considered assuming noisy measurements are obtained by periodically sampling a dynamic process described by the networked swing equation dynamics, modified to assume stochastic power injections. A centralized approach to network topology detection and isolation is introduced as a two-part scheme consisting of topology detection followed by topology isolation, assuming a topology attack exists. To address the complexity issues arising with performing centralized detection in large-scale power networks, a decentralized approach is presented that uses only local measurements to detect the presence of a topology attack. Simulation results illustrate that both the centralized and decentralized approaches accurately detect and isolate topology attacks.

This paper examines distributed event-triggered estimation over wireless sensor networks. In such systems an efficient utilization of the wireless communications must be performed since energy consumption and communication bandwidth are limit resources. We pose a global event-triggered communication policy for state estimation that minimizes a weighted function of the network energy consumption and the number of transmissions subject to constraints on the estimator performance. The global communication policy determines when sensors transmit measurements to the central estimator using a sensor-to-estimator communication channel as well as when sensors received other sensors' measurements (which have been transmitted to the central estimator) using a estimator-to-sensor communication channel. A distributed 1-step greedy heuristic is introduced for the proposed global minimization problem such that sensors determine their respective communication policies using only the local information available at each sensor. Simulation results demonstrate that the number of sensor transmissions can be reduced at a potential increase in network energy consumption (number of sensor transmission and receptions) with the added benefit of reducing network congestion.

This paper concerns the problem of large-scale source localization arising when many potential sources must be classified as either active or inactive such that the probability of missing an active source is bounded. A new iterative heuristic called the Iterative Source Localization Procedure (ISLoP) is introduced that reduces the complexity of a source localization problem with J potential sources from 2 J to J per iteration, while also providing a local bounds on the maximum probability of a missed source. The ISLoP separates the source localization problem into a likelihood maximization problem followed by an active source localization problem. A diffusion example is used to demonstrate the performance of the ISLoP when compared to an estimation-based approach, where the heuristic is shown to have increasingly better performance as the bound on the maximum probability of a missed source is decreased. An experimental evaluation of the heuristic with respect to common wireless sensor networking errors is provided using a test bed implementation for a CO 2 sequestration site monitoring problem.

Constraint-based control over wireless sensor networks(WSNs) require control strategies that achieve a desiredclosed-loop system performance while using minimal networkresources. In addition to constraints associated with distributedcontrol, WSNs have limitations on bandwidth, energy consumption,and transmission range. This paper introduces andexperimentally evaluates a new receding-horizon approach forperforming constraint-based control using a WSN. By leveragingthe system controllability, the receding-horizon controller isformulated as a mixed-integer programming problem which, ateach time step, simultaneously generates a control sequence andsensor selection schedule such that the desired performance isachieved while minimizing the energy required to perform dataacquisition and control. For systems containing many sensors,a multi-step state estimator is employed to implement thereceding-horizon controller using a conservative abstractionrelaxationapproach that simplifies the original mixed-integerprogramming problem into a convex quadratic programmingproblem. A wireless process control test bed consisting of8 coupled water tanks and 16 wireless sensors are used toexperimentally evaluate the receding-horizon controller.

This paper considers the minimum electricity costscheduling problem of smart home appliances. Operation characteristics,such as expected duration and peak power consumptionof the smart appliances, can be adjusted through a powerprofile signal. The optimal power profile signal minimizes cost,while satisfying technical operation constraints and consumerpreferences. Constraints such as enforcing uninterruptible andsequential operations are modeled in the proposed frameworkusing mixed integer linear programming (MILP). Several realisticscenarios based on actual spot price are considered,and the numerical results provide insight into tariff design.Computational issues and extensions of the proposed schedulingframework are also discussed.