Search

Deep neural networks (DNNs) are often coupled with physics-based models or data-driven surrogate models to perform fault detection and health monitoring of systems in the low data regime. These models serve as digital twins to generate large quantities of data to train DNNs which would otherwise be difficult to obtain from the real-life system. However, such models can exhibit parametric uncertainty that propagates to the generated data. In addition, DNNs exhibit uncertainty in the parameters learnt during training. In such a scenario, the performance of the DNN model will be influenced by the uncertainty in the physics-based model as well as the parameters of the DNN. In this article, we quantify the impact of both these sources of uncertainty on the performance of the DNN. We perform explicit propagation of uncertainty in input data through all layers of the DNN, as well as implicit prediction of output uncertainty to capture the former. Furthermore, we adopt Monte Carlo dropout to capture uncertainty in DNN parameters. We demonstrate the approach for fault detection of power lines with a physics-based model, two types of input data and three different neural network architectures. We compare the performance of such uncertainty-aware probabilistic models with their deterministic counterparts. The results show that the probabilistic models provide important information regarding the confidence of predictions, while also delivering an improvement in performance over deterministic models.

Condition-based and predictive maintenance enable early detection of critical system conditions and thereby enable decision makers to forestall faults and mitigate them. However, decision makers also need to take the operational and production needs into consideration for optimal decision-making when scheduling maintenance activities. Particularly in network systems, such as power grids, decisions on the maintenance of single assets can affect the entire network and are, therefore, more complex. This paper proposes a bi-level multi-agent decision support system for the generation maintenance decision (GMS) of power grids in an electricity market in the context of predictive maintenance. The GMS plays an important role in increasing the reliability at the network level. The aim of the GMS is to minimize the generation cost while maximizing the system reliability. The proposed framework integrates a central coordination system, i.e. the transmission system operator (TSO), and distributed agents representing power generation units that act to maximize their profit and decide about the optimal maintenance time slots while ensuring the energy balance. In the proposed bi-level approach, upper levels and one lower level (i.e. sub-problems) are solved by the independent agents and by the TSO, respectively. We derive the optimal strategy of the agents that are subject to predictive maintenance via a distributed algorithm, through which agents make optimal maintenance decisions and communicate them to the central coordinator, i.e. the TSO. The TSO decides whether to accept the agents’ decisions by considering market reliability aspects and power supply constraints. To solve this coordination problem, we propose a negotiation algorithm using an incentive signal to coordinate the agents’ and the central system’s decisions, such that all the agents’ decisions can be accepted by the central system. We demonstrate the effectiveness of the proposed algorithm with reference to the IEEE 39 bus system., Reliability Engineering & System Safety, 210, ISSN:0951-8320