Prediction and optimization of the design decisions of liquid cooling systems of battery modules using artificial neural networks

Abstract

Summary Liquid cooling systems are effective for keeping the battery modules in the safe temperature range. This study focuses on decreasing the power consumption of the pump without compromising the cooling performance. Artificial neural network (ANN) models are created to predict the effects of the height and width of the cooling channel and the mass flow rate on the maximum temperature, convective heat transfer coefficient, and pressure drop. The ANN models are used as surrogate models for the design and optimization of the liquid cooling battery system. Particle swarm optimization (PSO) and genetic algorithm (GA), which are commonly utilized optimization methods in many areas, and chaos game optimization (CGO) and coot optimization algorithm (COOT) methods, which are recently presented methods, are adopted to minimize the power consumption of the pump. The results are compared in terms of computational performance and best, worst, average, and SD values. Despite all of the optimization methods used giving similar results, the CGO method comes forward due to fast converging, SD, and finding the minimum power consumption of the pump among other optimization methods. A 22.4% decrease in the power consumption of the pump is achieved with the use of the ANN‐based CGO method while conserving the cooling performance. When comparing the ANN predicted and CFD results, the relative errors are less than 1%.

Experimental study on the thermal properties of Al 2 O 3 ‐CuO /water hybrid nanofluids: Development of an artificial intelligence model

“…Seven measurements were taken at each concentration for each temperature. In the current work, the experimental data of NF is compared with the data predicted by models presented in Equations ( 7), ( 8), ( 9), ( 10), (11), respectively.…”

“…However, these variations are more for Al 2 O 3 -CuO HNF. Since Equations ( 7), ( 8), ( 9), (10), (11) are used to forecast the TC of Al 2 O 3 NF, not for HNF. There is no acceptable model to estimate the TC of HNFs due to the complex interactions between NPs.…”

“…Artificial intelligence‐based models have become popular in recent years for analyzing the behavior of NFs' thermophysical characteristics. These models deliver higher precision results with shorter analysis times and lower computational needs 11 . Authors like Said et al, 12 Kanti et al, 13 Milewski et al, 14 and Sharma et al 15 used various prediction models.…”

“…These models deliver higher precision results with shorter analysis times and lower computational needs. 11 Authors like Said et al, 12 Kanti et al, 13 Milewski et al, 14 and Sharma et al 15 used various prediction models. Cascade-forward neural networks (CFNN) are similar to feed-forward networks (FFNN), but they include a connection from the input and every previous layer to the subsequent layers.…”

Metaheuristics‐optimized deep learning to predict generation of sustainable energy from rooftop plant microbial fuel cells

“…Models for predicting power generation can be divided into conventional approaches and artificial intelligence models 20 . Conventional approaches apply statistical methods and autoregressive integrated moving average (ARIMA) methods, 21 whereas artificial intelligence models include artificial neural networks (ANNs) 22 and support vector machines (SVMs) 23 . SVMs are primarily used to forecast short‐term time series data and nonlinear data 24 .…”