Table 11 Comparitive analysis of proposed Rul prediction framework with models.
Reference | Algorithms | Model complexity | RMSE | Adaptability to noise and unseen conditions | Research gap |
|---|---|---|---|---|---|
DNN | 3 hidden layers × 128 neurons; Total Parameters: ~34,945 | 3.427 | Moderate | -The model parameter selection was not conducted comprehensively | |
Ant lion-SVR | Trainable parameter: 1000–2000 | 0.0307 | Low | -The LIB parameters such as voltage, current, impedance was not critically analyzed and used | |
BPNN | 1 hidden layer with 10 neurons, learning rate as 0.005, iteration as 20 and epoch as 1000 | 0.0819 | Moderate | -The BPNN model parameters selection was not accurate and thus the error was high | |
SAE | 3 hidden layers | 0.65 | High | -The model demonstrated computational complexity with high error | |
LSTM | Batch size 32, epoch 500, 3 hidden layers, hidden neuron as 128, trainable parameters as 331,393 | 5.6262 | High | -Use of suitable techniques to extract the data samples could be performed | |
GRU | Hidden layer as 64 neurons, dropout as 0.2, trainable parameters as 12,737 | 0.97% | Moderate | -High-dimensional data could be applied for training the model | |
Proposed | JFO-FNN | Iteration 100, epochs as 1000, trainable parameters vary between [7 241] | 0.0292 | High | -Discharge profile of the LIB can be analyzed in future research works |
