An Optimal Sensor Configuration Method using Genetic Algorithms for a Fault-Tolerant Energy Management for Smart Buildings

Abstract

Smart buildings increasingly rely on interconnected devices, particularly sensors, to optimize energy usage. The Internet of Things (IoT) paradigm enables advanced Building Energy Management Systems (EMSs). In particular, energy management in smart buildings should monitor energy usage, identify high-energy-consuming devices, schedule device usage, and prevent energy waste, among other aspects. A key challenge in deploying such systems lies in determining the optimal configuration and placement of sensors to ensure reliable and efficient operation. The problem is to define the configuration of sensors to perceive the information of the context useful for the energy management system. The main question is: What are the sensors required to guarantee the correct operation of the energy management system? The goal of this work is to propose a method based on multi-genetic algorithms and structural analysis theory to obtain an optimal sensor configuration. Thus, this paper addresses the problem of sensor configuration in EMSs from a diagnosability perspective, i.e., ensuring that the system can detect and isolate faults. This combination of approaches allows studying the diagnosability property in an EMS for smart buildings, which identifies the sensors required to ensure that a system can function without failures. In particular, this property is studied based on the generation of Analytical Redundancy Relations using minimal test equation supports over a bipartite graph. Using this paradigm, the multiple genetic algorithms analyze the different properties to achieve the diagnosability property in a system, specifically, the detectability and isolability properties, and determine the placement of the new sensors required to reach them. In this way, our method studies the sensor configuration problem for EMSs to fulfill their diagnosability property. Our method systematically identifies the minimal set of additional sensors required to guarantee EMS diagnosability while considering trade-offs in cost and energy consumption. Experimental results are very promising, since they demonstrate the effectiveness of the method in real-world scenarios by analyzing the diagnosability property to achieve fault-tolerant EMS configurations and highlight the flexibility of the method to offer alternative sensor designs based on environmental constraints