Premium
Monte Carlo‐based sensitivity analysis of an electrochemical capacitor
Author(s) -
Kannan Vishvak,
Somasundaram Karthik,
Fisher Adrian,
Birgersson Erik
Publication year - 2021
Publication title -
international journal of energy research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.808
H-Index - 95
eISSN - 1099-114X
pISSN - 0363-907X
DOI - 10.1002/er.6919
Subject(s) - monte carlo method , capacitor , sensitivity (control systems) , reliability (semiconductor) , materials science , degradation (telecommunications) , electrochemistry , durability , electrode , control theory (sociology) , computer science , biological system , reliability engineering , voltage , simulation , electronic engineering , engineering , mathematics , chemistry , physics , power (physics) , statistics , composite material , thermodynamics , electrical engineering , control (management) , artificial intelligence , biology
Summary The operation of electrochemical capacitors depends not only on extrinsic operating and design parameters, but also intrinsic physical, material, and electrochemical parameters. Fluctuations in these stochastic parameters can significantly influence the performance and may lead to quicker degradation of the electrochemical capacitors, thereby affecting their durability and reliability. Thus, it is important to quantify the sensitivities of these extrinsic and intrinsic parameters and correlate them with the performance, to delay the inevitable performance degradation. To achieve this, we perform Monte Carlo simulations (MCS) followed by sensitivity analysis under high and low charge/discharge current (load) conditions. The MCS is statistically performed by varying all the stochastic parameters simultaneously. We then identify the critical parameters that affect the performance under different load conditions, which provides insights into optimal operation of electrochemical capacitors. The thickness of positive electrode and the radius of active material were identified as the most significant parameters under low and high load conditions, respectively. Furthermore, we derive reduced order surrogate models with at least 95% accuracy using supervised machine learning techniques to predict the performance without solving the full physics‐based model.