This thesis proposes and implements a mini reactor to harvest energy from fluids in industrial environments. The results of the tests show that the developed system provides an efficiency of 20% and delivers a power between 15 and 333 mW to the load for air pressures in pipes of 20 to 163 mbar, respectively.

To analyze the feasibility of a dual power manager in collection systems, two types of reference collectors have been used in this work, one DC (solar panel) and another AC, the mini-reactor. The process to obtain the data to design and scale the power and efficiency of the AC / DC and DC / DC converters, requires modeling the collectors. Thus, this thesis introduces a structural and electrical model of the solar cell and a new electromechanical model of the mini-reactor. The results of the tests presented show that the DC system based on solar cells provides more energy for long periods, due to the periodic availability of light. On the other hand, the mini-reactor and its AC / DC converter provide more power to the load, but for shorter periods. The objective of future developments is to achieve a universal energy management system for n-number of energy sources.

The mathematical model and the artificial intelligence algorithm developed are dedicated to the applications of energy collectors for wireless networks of low power sensors. This task constitutes an improvement of the state of the art in the electro-mathematical models of super-capacitors for low power applications since it uses its most characteristic parameters. In addition, the simulation results and the tests show that it provides the same precision as the electrochemical models and greater accuracy than the more advanced mathematical models. Improving the electro-mathematical model with machine learning algorithms requires less experimental data than electrochemical models to achieve the same accuracy. In addition, the machine learning algorithm does not require any electrochemical parameter to implement the model.

This thesis assesses whether predictive methods and algorithms, taken from the field of artificial intelligence and advanced analytics, could become a solution to overcome the technical challenges of power management. The work carried out is based on the feasibility of predicting both the availability of energy and the requirements for energy consumption. Thanks to the aforementioned, the energy management system could make decisions in critical situations and address tasks related to the use and distribution of the available energy in a network of IIoT systems. However, the results obtained in this work show that not all the algorithms tested are suitable for IIoT systems, so further research is required in algorithms, collectors and energy collection applications. One of the relevant research fields should be the energy consumption and the energy efficiency of the algorithms.

Summarizing, the research presented in this thesis shows that energy collection systems are a feasible alternative to reduce or eliminate the dependence on batteries of IIoT devices, thus contributing to the paradigm of Industry 4.0 and its effective deployment.