Deep learning approach for battery aging model

1. A computer-implemented method executed on a processor for predicting a life span of a battery storage unit by employing a deep neural network, the method comprising:

collecting energy consumption data from one or more electricity meters installed in a structure;

analyzing, via a data processing component, the energy consumption data;

removing one or more features extracted from the energy consumption data via a feature engineering component;

partitioning the energy consumption data via a data partitioning component after removing the one or more features; and

predicting battery capacity of the battery storage unit via a neural network component sequentially executing three machine learning techniques,

wherein the first machine learning technique is multiple linear regression (MLR), the second machine learning technique involves a neural network model, and the third machine learning technique is a Long Short-Term Memory (LSTM) neural network model.

2. The method of claim 1 , wherein test set data in the MLR technique are normalized by using a mean and standard deviation of training data.

3. The method of claim 1 , wherein the neural network model is trained to predict a battery faded capacity percentage.

4. The method of claim 1 , wherein the LSTM neural network model includes a selection gate, a forgetting gate, and an ignoring gate, where all weights of a graph are calculated so that a loss value is minimized with every input.

5. The method of claim 1 , wherein the one or more features extracted from the energy consumption data include date, time, charging power, discharging power, maximum temperature, and battery state of charge (SoC).

6. The method of claim 5 , wherein energy throughput is calculated based on the charging power and the discharging power features.

7. A system for predicting a life span of a battery storage unit by employing a deep neural network, the system comprising:

a memory; and

a processor in communication with the memory, wherein the processor runs program code to:

collect energy consumption data from one or more electricity meters installed in a structure;

analyze, via a data processing component, the energy consumption data;

remove one or more features extracted from the energy consumption data via a feature engineering component;

partition the energy consumption data via a data partitioning component after removing the one or more features; and

predict battery capacity of the battery storage unit via a neural network component sequentially executing three machine learning techniques,

wherein the first machine learning technique is multiple linear regression (MLR), the second machine learning technique involves a neural network model, and the third machine learning technique is a Long Short-Term Memory (LSTM) neural network model.

8. The system of claim 7 , wherein test set data in the MLR technique are normalized by using a mean and standard deviation of training data.

9. The system of claim 7 , wherein the neural network model is trained to predict a battery faded capacity percentage.

10. The system of claim 7 , wherein the LSTM neural network model includes a selection gate, a forgetting gate, and an ignoring gate, where all weights of a graph are calculated so that a loss value is minimized with every input.

11. The system of claim 7 , wherein the one or more features extracted from the energy consumption data include date, time, charging power, discharging power, maximum temperature, and battery state of charge (SoC).

12. The system of claim 11 , wherein energy throughput is calculated based on the charging power and the discharging power features.

13. A non-transitory computer-readable storage medium comprising a computer-readable program for predicting a life span of a battery storage unit by employing a deep neural network, wherein the computer-readable program when executed on a computer causes the computer to perform the steps of:

collecting energy consumption data from one or more electricity meters installed in a structure;

analyzing, via a data processing component, the energy consumption data;

removing one or more features extracted from the energy consumption data via a feature engineering component;

partitioning the energy consumption data via a data partitioning component after removing the one or more features; and

predicting battery capacity of the battery storage unit via a neural network component sequentially executing three machine learning techniques,

wherein the one or more features extracted from the energy consumption data include date, time, charging power, discharging power, maximum temperature, and battery state of charge (SoC).

14. The non-transitory computer-readable storage medium of claim 13 , wherein the first machine learning technique is multiple linear regression (MLR), the second machine learning technique involves a neural network model, and the third machine learning technique is a Long Short-Term Memory (LSTM) neural network model.

15. The non-transitory computer-readable storage medium of claim 14 , wherein test set data in the MLR technique are normalized by using a mean and standard deviation of training data.

16. The non-transitory computer-readable storage medium of claim 14 , wherein the neural network model is trained to predict a battery faded capacity percentage.

17. The non-transitory computer-readable storage medium of claim 14 , wherein the LSTM neural network model includes a selection gate, a forgetting gate, and an ignoring gate, where all weights of a graph are calculated so that a loss value is minimized with every input.