Fachbereich Elektrotechnik, Maschinenbau, Technikjournalismus
This paper provides a performance analysis of a wearable photovoltaic system mounted on the outer surface of a backpack. Three types of photovoltaic materials, commonly used for electricity generation, have been investigated under various conditions including sun irradiance, angle-of-incidence and sun inclination. The results of the investigation have shown that the system equipped with the rigid mono-Si panels performs 3.5 to 4.9 times better than the system equipped with a-Si flexible PV modules. The average power generated by the wearable photovoltaic system is about 30% of the maximum installed power for any photovoltaic type. This paper presents the test data resulting from the evaluation of the daily energy production of a wearable photovoltaic power supply.
Power train models are required to simulate hence predict energy consumption of vehicles. Efficiencies for different components in power train are required. Common procedures use digitalised shell models (or maps) to model the efficiency of Internal Combustion Engines (ICE) and manual gearboxes (MG). Errors are connected with these models and affect the accuracy of the calculation. The accuracy depends on the configuration of the simulation, the digitalisation of the data and the data used. This paper evaluates these sources of error. The understanding of the source of error can improve the results of the modelling by more than eight percent.
This paper proposes a new artificial neural network-based maximum power point tracker for photovoltaic application. This tracker significantly improves efficiency of the photovoltaic system with series-connection of photovoltaic modules in non-uniform irradiance on photovoltaic array surfaces. The artificial neural network uses irradiance and temperature sensors to generate the maximum power point reference voltage and employ a classical perturb and observe searching algorithm. The structure of the artificial neural network was obtained by numerical modelling using Matlab/Simulink. The artificial neural network was trained using Bayesian regularisation back-propagation algorithms and demonstrated a good prediction of the maximum power point. Relative number of Vmpp prediction errors in range of ±0.2V is 0.05% based on validation data.
