ShareApplications of Artificial Intelligence to refrigeration systems
AI techniques, which are currently very topical, are the subject of intense research and development. Overview of their applications and benefits in the refrigeration sector, based on the latest articles published in the IIR’s FRIDOC database.
The Spanish authors of a paper (1) presented at the recent IIR-co-sponsored CYTEF conference, stress that over the last decade, the industrial sector has experimented a great revolution due to the adoption of new strategies based on the improvement of information and communication technologies (ITCs) and digitisation. New technologies such as system virtualisation, the Internet of Things (IoT), Big Data, cloud computing or Artificial Intelligence (AI) have shaped what is known as Industry 4.0 or the fourth industrial revolution. All these technologies support the three fundamental bases of Industry 4.0, namely the interconnection of systems and devices, the massive acquisition of data and the processing of this information.
IoT consists of the interconnection of all kinds of devices and objects in a data network so that they can exchange information with others. Big data is a technology used for processing large volumes of information. Thanks to technologies such as the IoT, an increasing amount of data is available to industries. All this information generated by different nodes in the network must be collected and stored in a structured way for analysis.
AI algorithms have the ability to analyse large amounts of information to learn from the data and use this knowledge to perform certain tasks without the need to explicitly program each possible scenario. In this way, AI gives machines the ability to learn autonomously to detect complex patterns and trends in data, draw conclusions and act accordingly, just as an expert would.
Refrigeration systems modelling is mainly based on solving a mathematical problem in which one seeks to optimise one or more parameters of the system by reducing the cost of an objective function. In energy efficiency problems, these parameters are related to the energy performance of the system such as the coefficient of performance (COP) or the energy efficiency ratio (EER). Two main types of IA models can be distinguished for refrigeration system optimisation: experience-based and data-driven models.
Experience-based artificial intelligence models have the ability to establish relationships and understand a system based on prior knowledge of the system. In this field, different IA techniques have been successfully used to solve energy optimisation problems in industrial refrigeration systems. Some of the most commonly used methods are fuzzy logic, genetic algorithms, expert systems and stochastic optimisation methods. Energy savings of over 30% per day are reportedly achieved by applying these control techniques.
Data-driven artificial intelligence algorithms have the ability to model a system exclusively based on the correlation between input and output data, without any prior information or knowledge of the system. This set of techniques is known as machine learning (ML) and among themse techniques, Artificial Neural Networks (ANNs) stand out. ANNs are increasingly used in the modeling of industrial refrigeration systems for energy savings of up to 17%, according to the authors.
In a paper (2) presented at the latest IIR TPTPR conference, Italian researchers reviewed the practical results of 19 works relating to the use of AI algorithms in the refrigeration sector.
Fault detection and diagnosis (FDD) using machine learning is one of the most studied AI applications. Predicting the operation of the refrigeration system allows for better scheduling of maintenance operations to avoid hard failures, which occur abruptly and may cause the system stop functioning, and to limit the costs related to soft failures that cause a thermodynamic cycle variation and thus a degradation in system performance.
AI is also becoming widely used for energy performance prediction and control of refrigeration systems, with the potential to improve energy savings using black-box modelling approaches, calibrated on data from field case studies or laboratory experiments.
Defrosting optimisation is a significant example of the potential benefits of AI algorithms. The authors give the example of an open vertical display cabinet for which an ANN algorithm was developed for the defrosting process. The algorithm has been trained with data obtained from an experimental campaign and then used to optimise the controlled parameters of refrigeration time, defrosting time and air velocity and minimise the energy consumption of the process. A reduction of around 27% in total energy consumption relative to the total display area was obtained.
(1) Cerdán Cartagena F., Application of artificial intelligence to refrigeration systems. See in FRIDOC.
(2) Citarella B. et al, Use of artificial Intelligence in the refrigeration field. See in FRIDOC.