# Deep Learning LSTM Recurrent Neural Network Model for Prediction of Electric Vehicle Charging Demand

^{1}

^{2}

^{3}

^{*}

## Abstract

**:**

^{−5}are attained. Furthermore, the mean absolute error was evaluated to be 0.1083 and the mean square error pertaining to 4.25516 × 10

^{−10}. The results prove the efficacy of the prediction metrics computed with the novel deep learning LSTM neural predictor for the considered dataset in comparison with the previous techniques from existing works.

## 1. Introduction

## 2. Related Works and Motivations

- -
- -
- -
- -
- -
- -
- -
- -
- Requirement to analyze the fuel economy and drivability, else higher error variations [44]
- -
- -
- Lack in frequent data sharing between the charging stations and charging station providers [26]
- -
- Existence of non-linearity with time-dependent data [45]
- -
- -
- -
- High mobility and low reliability of electric vehicles [47]
- -
- Environmental factors and occupant behavior affects the performance of existing prediction models [48]
- -
- Difficulty in analyzing the long-term energy consumption prediction [49]
- -
- Dependency on the level of state-of-charge [50]
- -
- Implication of driving capacity on maintaining the charging capacity of electric vehicles [51]
- -
- -
- Poor descriptive ability of linear networks for complex environments [53]
- -
- -
- Certain predictive models are limited to short-term based prediction [54]
- -
- Difficulty of algorithms to handle temporal profiles [42]

#### Aim and Objectives of the Research Study

- -
- To model a novel deep learning based recurrent neural network model that employs auto-encoder and decoder for handling the non-linear data of the electric vehicle charging.
- -
- Employing the empirical mode decomposition (EMD) to decompose the data and attain the temporal features using the intrinsic frequency components.
- -
- Applying the arithmetic optimizer algorithm (AOA) to find the optimal weights and bias values of the designed deep learning neural model.
- -
- Designing the structure of the deep long-short term memory (DLSTM) neural network for performing the prediction with proposed training and training algorithms.
- -
- Testing and validating the EMD–AOA–DLSTM on the electric vehicle charging dataset of Georgia Tech, Atlanta, USA.
- -
- To ensure that for charging EVs, based on the prediction done, the waiting time gets reduced for charging in a 24-h time period.

## 3. Methods and Materials

#### 3.1. Empirical Mode Decomposition

_{j}(t) for j=1,2,…,m specifies the intrinsic mode functions (IMF) for various decompositions and D

_{m}(t) indicates the residue derived after the specified number of IMFs are carried out. For performing EMD, a suitable IMF should be defined satisfyingthe number of extrema and the zero crossing shall be equal or be different atleast by one and at any specific point, the average value of the envelop indicated by the local maxima and minima should be ‘0’ [55,56]. The steps to perform EMD for the electric vehicle charging time-series data are as follows:

_{upp}(t)] by connecting all the local maxima by a cubic spline and generate the lower envelope [S

_{low}(t)] by connecting all the local minima.

- -
- On [Y(t)] being an IMF then set X(t) = Y(t) and also replace [S(t)] with the residual [D(t) = S(t) − X(t)].
- -
- On [Y(t)] not an IMF, replace [S(t)] with [Y(t)].

#### 3.2. Arithmetic Optimization Algorithm—Revisited

_{max}and α

_{min}specifies the maximum and minimum values of accelerated functions, ‘Iter

_{current}’ indicates the current iteration and ‘Iter

_{max}’ specifies the maximum iteration.

_{1}and r

_{2}are small random numbers and the division operator performs when r

_{2}< 0.5 and the multiplication operator do not perform until ‘D’ operator completes the current operation. The best solution evaluated so far is ‘y

_{j_Best}’, ‘λ’ represents the control parameter for adjusting the search mechanism, ‘ε’ is the small integer number, LB

_{j}and UB

_{j}specifies the lower bound and upper bound of the present position, and ‘mop’ is math optimizer probability given by,

#### 3.3. LSTM Recurrent Neural Model

- -
- For overcoming the saturation of the training model and enabling to get convergence
- -
- For maintaining the balanced weights and bias at the time of training
- -
- Choosing suitable activation function to evaluate the network output
- -
- Unwarranted termination of training process is overcome
- -
- To possess a better slope value so that the gradient enables an efficient training mechanism
- -
- To increase the memory cells and classify the data; thereby formulating better training and testing process
- -
- The designed recurrent neural network model will be prevented from instability occurrences.

_{t}’ and the previous state ‘z

_{t−}

_{1}′ compute the function and is defined by,

_{gt}’ specifies the forget gate, ‘α’ is the learning rate metric, ‘W

_{gt}’ represents the weights of the model and ‘W

_{ogt}’ indicating the bias of the neural model.

_{gt}indicates the input gate and Y

_{gt}assigns weights to the data through the sigmoidal function.

_{gt}’ represents the output gate and this gate presents the output from the memory cells and ‘z

_{t}’ specifies the current state from which the output is computed.

#### 3.4. Proposed EMD–AOA–Deep LSTM Recurrent Neural Predictor

_{gt}, g

_{gt}and R

_{gt}that combines with the cell output.

^{50×30}, as there are 30 convolutional filters placed. The output matrix of the convolutional layer contains the weight of one filter and at the end of the fourth convolutional layer, a single dimension max pooling layer (pool size −2) exist and attains the output U

^{25×30}. An LSTM operational layer follows the max pooling layer with 70 neurons, and here, there is a 30% recurrent dropout probability and a vector of U

^{1×70}is computed. Finally, a fully connected network with 70 neurons is formed with linear activations and then the final soft-max layer acts as the predictor. The EMD–AOA based DLSTM model evaluates the mean square error value along with the recurrent drop out, which facilitates in circumventing the over-fitting occurrences.For the modeled novel DLSTM, its encoder activation function ‘G

_{encode}’ and the labeled sample data points ‘X

_{data}’ formulates the encode matrix as,

_{f_encode}’ and ‘g

_{f_decode}’ specifies the encoder and decoder activation function of the deep learning predictor model, ‘W

_{0}′ represents the bias element and the weight matrices are ‘W

_{x}’ and ‘${W}_{x}^{T}$’. The error is evaluated during deep training process using,

_{encodeN+1}′ represents the trained values at the LSTM output layer and the new weights based on the gradients are evaluated to be,

#### 3.5. EV Charging Datasets

## 4. Results and Discussions

_{pre}) and they are evaluated with the following equations,

_{actual}’ represents the original EV charging farm data and ‘Y

_{predicted}’ is the predicted output obtained using the proposed predictor model. Figure 8 shows the EMD decomposed sub-series output of the considered EV data samples and these sub-series data are presented as input to the DLSTM model. Figure 9 presents the design of the proposed DLSTM predictor model in the deep network designer of the MATLAB environment and simulation results are attained henceforth by training the predictive model created.

_{pre}are evaluated and listed in Table 5. Figure 10 shows the plots of the predicted charging energy level with that of the actual charging energy level of the EV charging station. It is clear from Figure 10 that the predicted EV charging energy is on par with the original charging energy level for the EV charging station. Figure 11 depicts the convergence curve attained during the deep learning process of the proposed predictor model. At the time of training, the convergence was obtained at 251st epoch with an MSE of 4.25516 × 10

^{−10}and for testing and validation the evaluated MSE value is 5.96333 × 10

^{−10}and 5.5317 × 10

^{−10},respectively. The prediction accuracy attained during the convergence of the DLSTM predictor model is 97.14% with a minimal MSE and an MAE of 0.1083.

^{−10}. This indicates that the proposed predictor model will be able to improve its learning phase only when there is a small change in its weights and bias. This is well supported by the plot shown in Figure 11.

^{−10}elapsed at 251st epoch for training process and during testing process, the MSE was 5.96333 × 10

^{−10}at 251st for testing process. Table 7 provides the sample of predicted EV charging demand value with that of the actual EV charging demand value at the Georgia Tech charging outlet. The predicted values prove their values are on par and near equal to that of the actual EV charging energy (kWh) for the considered charging station dataset. At the 251st Epoch, it has reached the convergence and attained the minimal MSE value.

## 5. Comparative Analysis

^{−10}and prediction accuracy of 97.14% has proved its superiority than other previous techniques from earlier works. The proposed model incurred a minimized computational time of 7.4 s than the Bayesian ELM model, which incurred 16.14 s [1]. The average training and testing efficiency of proposed predictor was 98.62% and 98.03%, better than other compared models proving the efficacy of deep learning mechanism.TheEMD based sub-series decomposition and the AOA presence to attain optimal training parameters has enhanced the proposed DLSTM model to achieve predicted charging energy demand on par with that of the actual EV charging energy level.

## 6. Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Conflicts of Interest

## References

- Wang, Q.; Sun, Y.; Huang, Y. SOC Prediction of HES in HEV based on bayesian extreme learning machine. ICIC Express Lett. Part B Appl. Int. J. Res. Surv.
**2014**, 5, 1735–1740. [Google Scholar] - Grubwinkler, S.; Lienkamp, M. Energy prediction for EVS using support vector regression methods. In Intelligent Systems; Springer: Cham, Switzerland, 2014; pp. 769–780. [Google Scholar]
- Majidpour, M.; Qiu, C.; Chu, P.; Gadh, R.; Pota, H.R. Fast prediction for sparse time series: Demand forecast of EV charging stations for cell phone applications. IEEE Trans. Ind. Inform.
**2014**, 11, 242–250. [Google Scholar] [CrossRef] - Chen, Z.; Li, L.; Yan, B.; Yang, C.; Martinez, C.M.; Cao, D. Multimode energy management for plug-in hybrid electric buses based on driving cycles prediction. IEEE Trans. Intell. Transp. Syst.
**2016**, 17, 2811–2821. [Google Scholar] [CrossRef] - Almaghrebi, A.; Aljuheshi, F.; Rafaie, M.; James, K.; Alahmad, M. Data-driven charging demand prediction at public charging stations using supervised machine learning regression methods. Energies
**2020**, 13, 4231. [Google Scholar] [CrossRef] - Zhang, Z.; Zou, Y.; Zhou, T.; Zhang, X.; Xu, Z. Energy Consumption Prediction of Electric Vehicles Based on Digital Twin Technology. World Electr. Veh. J.
**2021**, 12, 160. [Google Scholar] [CrossRef] - Fukushima, A.; Yano, T.; Imahara, S.; Aisu, H.; Shimokawa, Y.; Shibata, Y. Prediction of energy consumption for new electric vehicle models by machine learning. IET Intell. Transp. Syst.
**2018**, 12, 1174–1180. [Google Scholar] [CrossRef] - Liu, K.; Asher, Z.; Gong, X.; Huang, M.; Kolmanovsky, I. Vehicle Velocity Prediction and Energy Management Strategy Part 1: Deterministic and Stochastic Vehicle Velocity Prediction Using Machine Learning; SAE Technical Paper; SAE International: Warrendale, PA, USA, 2019; pp. 1–8. [Google Scholar]
- Mao, M.; Zhang, S.; Chang, L.; Hatziargyriou, N.D. Schedulable capacity forecasting for electric vehicles based on big data analysis. J. Mod. Power Syst. Clean Energy
**2019**, 7, 1651–1662. [Google Scholar] [CrossRef] - Hannah Jessie Rani, R.; Aruldoss Albert Victoire, T. A hybrid Elman recurrent neural network, group search optimization, and refined VMD-based framework for multi-step ahead electricity price forecasting. Soft Comput.
**2019**, 23, 8413–8434. [Google Scholar] [CrossRef] - McBee, K.D.; Bukofzer, D.; Chong, J.; Bhullar, S. Forecasting long-term electric vehicle energy demand in a specific geographic region. In Proceedings of the IEEE Power & Energy Society General Meeting (PESGM), Montreal, QC, Canada, 3–6 August 2020; pp. 1–5. [Google Scholar]
- Zhang, J.; Xu, F.; Zhang, Y.; Shen, T. ELM-based driver torque demand prediction and real-time optimal energy management strategy for HEVs. Neural Comput. Appl.
**2020**, 32, 14411–14429. [Google Scholar] [CrossRef] - Huang, X.; Wu, D.; Boulet, B. Ensemble learning for charging load forecasting of electric vehicle charging stations. In Proceedings of the IEEE Electric Power and Energy Conference (EPEC), Edmonton, AB, Canada, 18 January 2020; pp. 1–5. [Google Scholar]
- Sun, D.; Ou, Q.; Yao, X.; Gao, S.; Wang, Z.; Ma, W.; Li, W. Integrated human-machine intelligence for EV charging prediction in 5G smart grid. EURASIP J. Wirel. Commun. Netw.
**2020**, 2020, 139. [Google Scholar] [CrossRef] - Khan, N.; Haq, I.U.; Khan, S.U.; Rho, S.; Lee, M.Y.; Baik, S.W. DB-Net: A novel dilated CNN based multi-step forecasting model for power consumption in integrated local energy systems. Int. J. Electr. Power Energy Syst.
**2021**, 133, 107023. [Google Scholar] [CrossRef] - Deb, S.; Goswami, A.K.; Chetri, R.L.; Roy, R. Bayesian optimization based machine learning approaches for prediction of plug-in electric vehicle state-of-charge. Int. J. Emerg. Electr. Power Syst.
**2021**, 22, 753–764. [Google Scholar] [CrossRef] - Pan, C.; Tao, Y.; Liu, Q.; He, Z.; Liang, J.; Zhou, W.; Wang, L. Grey wolf fuzzy optimal energy management for electric vehicles based on driving condition prediction. J. Energy Storage
**2021**, 44, 103398. [Google Scholar] [CrossRef] - Quan, S.; Wang, Y.X.; Xiao, X.; He, H.; Sun, F. Real-time energy management for fuel cell electric vehicle using speed prediction-based model predictive control considering performance degradation. Appl. Energy
**2021**, 304, 117845. [Google Scholar] [CrossRef] - Schmid, R.; Buerger, J.; Bajcinca, N. Energy management strategy for plug-in-hybrid electric vehicles based on predictive PMP. IEEE Trans. Control Syst. Technol.
**2021**, 29, 2548–2560. [Google Scholar] [CrossRef] - Lin, X.; Wu, J.; Wei, Y. An ensemble learning velocity prediction-based energy management strategy for a plug-in hybrid electric vehicle considering driving pattern adaptive reference SOC. Energy
**2021**, 234, 121308. [Google Scholar] [CrossRef] - Xin, W.; Zheng, W.; Qin, J.; Wei, S.; Ji, C. Energy management of fuel cell vehicles based on model prediction control using radial basis functions. J. Sens.
**2021**, 24, 25–37. [Google Scholar] [CrossRef] - Thorgeirsson, A.T.; Scheubner, S.; Fünfgeld, S.; Gauterin, F. Probabilistic prediction of energy demand and driving range for electric vehicles with federated learning. IEEE Open J. Veh. Technol.
**2021**, 2, 151–161. [Google Scholar] [CrossRef] - Shahriar, S.; Al-Ali, A.R.; Osman, A.H.; Dhou, S.; Nijim, M. Prediction of EV charging behavior using machine learning. IEEE Access
**2021**, 9, 111576–111586. [Google Scholar] [CrossRef] - Cadete, E.; Ding, C.; Xie, M.; Ahmed, S.; Jin, Y.F. Prediction of electric vehicles charging load using long short-term memory model. In Tran-SET 2021; American Society of Civil Engineers: Reston, VA, USA, 2021; pp. 52–58. [Google Scholar]
- Liu, Q.; Dong, S.; Yang, Z.; Xu, F.; Chen, H. Energy management strategy of hybrid electric vehicles based on driving condition prediction. IFAC-Pap. Online
**2021**, 54, 265–270. [Google Scholar] [CrossRef] - Zhao, Y.; Wang, Z.; Shen, Z.J.; Sun, F. Data-driven framework for large-scale prediction of charging energy in electric vehicles. Appl. Energy
**2021**, 282, 116175. [Google Scholar] [CrossRef] - Lin, X.; Wang, Z.; Wu, J. Energy management strategy based on velocity prediction using back propagation neural network for a plug-in fuel cell electric vehicle. Int. J. Energy Res.
**2021**, 45, 2629–2643. [Google Scholar] [CrossRef] - Malek, Y.N.; Najib, M.; Bakhouya, M.; Essaaidi, M. Multivariate deep learning approach for electric vehicle speed forecasting. Big Data Min. Anal.
**2021**, 4, 56–64. [Google Scholar] [CrossRef] - Basso, R.; Kulcsár, B.; Sanchez-Diaz, I. Electric vehicle routing problem with machine learning for energy prediction. Transp. Res. Part B Methodol.
**2021**, 145, 24–55. [Google Scholar] [CrossRef] - Aguilar-Dominguez, D.; Ejeh, J.; Dunbar, A.D.; Brown, S.F. Machine learning approach for electric vehicle availability forecast to provide vehicle-to-home services. Energy Rep.
**2021**, 7, 71–80. [Google Scholar] [CrossRef] - Lin, X.; Zeng, S.; Li, X. Online correction predictive energy management strategy using the Q-learning based swarm optimization with fuzzy neural network. Energy
**2021**, 223, 120071. [Google Scholar] [CrossRef] - Zeng, T.; Zhang, C.; Zhang, Y.; Deng, C.; Hao, D.; Zhu, Z.; Ran, H.; Cao, D. Optimization-oriented adaptive equivalent consumption minimization strategy based on short-term demand power prediction for fuel cell hybrid vehicle. Energy
**2021**, 227, 120305. [Google Scholar] [CrossRef] - Al-Gabalawy, M. Reinforcement learning for the optimization of electric vehicle virtual power plants. Int. Trans. Electr. Energy Syst.
**2021**, 31, e12951. [Google Scholar] [CrossRef] - Ye, M.; Chen, J.; Li, X.; Ma, K.; Liu, Y. Energy Management Strategy of a Hybrid Power System Based on V2X Vehicle Speed Prediction. Sensors
**2021**, 21, 5370. [Google Scholar] [CrossRef] - Chinnadurrai, C.L.; Aruldoss Albert Victoire, T. Dynamic economic emission dispatch considering wind uncertainty using non-dominated sorting crisscross optimization. IEEE Access
**2020**, 8, 94678–94696. [Google Scholar] [CrossRef] - Liu, Y.; Li, J.; Gao, J.; Lei, Z.; Zhang, Y.; Chen, Z. Prediction of vehicle driving conditions with incorporation of stochastic forecasting and machine learning and a case study in energy management of plug-in hybrid electric vehicles. Mech. Syst. Signal Process.
**2021**, 158, 107765. [Google Scholar] [CrossRef] - Pokharel, S.; Sah, P.; Ganta, D. Improved Prediction of Total Energy Consumption and Feature Analysis in Electric Vehicles Using Machine Learning and Shapley Additive Explanations Method. World Electr. Veh. J.
**2021**, 12, 94. [Google Scholar] [CrossRef] - Rabinowitz, A.; Araghi, F.M.; Gaikwad, T.; Asher, Z.D.; Bradley, T.H. Development and evaluation of velocity predictive optimal energy management strategies in intelligent and connected hybrid electric vehicles. Energies
**2021**, 14, 5713. [Google Scholar] [CrossRef] - Zhou, H.; Zhou, Y.; Hu, J.; Yang, G.; Xie, D.; Xue, Y.; Nordström, L. LSTM-based energy management for electric vehicle charging in commercial-building prosumers. J. Mod. Power Syst. Clean Energy
**2021**, 9, 1205–1216. [Google Scholar] [CrossRef] - Wu, C.; Jiang, S.; Gao, S.; Liu, Y.; Han, H. Charging demand forecasting of electric vehicles considering uncertainties in a microgrid. Energy
**2022**, 247, 123475. [Google Scholar] [CrossRef] - Chen, Z.; Liu, Y.; Zhang, Y.; Lei, Z.; Chen, Z.; Li, G. A neural network-based ECMS for optimized energy management of plug-in hybrid electric vehicles. Energy
**2022**, 243, 122727. [Google Scholar] [CrossRef] - Powell, S.; Cezar, G.V.; Rajagopal, R. Scalable probabilistic estimates of electric vehicle charging given observed driver behavior. Appl. Energy
**2022**, 309, 118382. [Google Scholar] [CrossRef] - Titus, F.; Thanikanti, S.B.; Deb, S.; Kumar, N.M. Charge Scheduling Optimization of Plug-In Electric Vehicle in a PV Powered Grid-Connected Charging Station Based on Day-Ahead Solar Energy Forecasting in Australia. Sustainability
**2022**, 14, 3498. [Google Scholar] - Asensio, E.M.; Magallán, G.A.; Pérez, L.; De Angelo, C.H. Short-term power demand prediction for energy management of an electric vehicle based on batteries and ultracapacitors. Energy
**2022**, 247, 123430. [Google Scholar] [CrossRef] - Liu, Y.; Liu, W.; Gao, S.; Wang, Y.; Shi, Q. Fast charging demand forecasting based on the intelligent sensing system of dynamic vehicle under EVs-traffic-distribution coupling. Energy Rep.
**2022**, 8, 1218–1226. [Google Scholar] [CrossRef] - Shi, J.; Liu, N.; Huang, Y.; Ma, L. An Edge Computing-oriented Net Power Forecasting for PV-assisted Charging Station: Model Complexity and Forecasting Accuracy Trade-off. Appl. Energy
**2022**, 310, 118456. [Google Scholar] [CrossRef] - Akbar, K.; Zou, Y.; Awais, Q.; Baig, M.J.; Jamil, M. A Machine Learning-Based Robust State of Health (SOH) Prediction Model for Electric Vehicle Batteries. Electronics
**2022**, 11, 1216. [Google Scholar] [CrossRef] - Wang, W.; Guo, X.; Yang, C.; Zhang, Y.; Zhao, Y.; Huang, D.; Xiang, C. A multi-objective optimization energy management strategy for power split HEV based on velocity prediction. Energy
**2022**, 238, 121714. [Google Scholar] [CrossRef] - Malik, H.; Alotaibi, M.A.; Almutairi, A. A new hybrid model combining EMD and neural network for multi-step ahead load forecasting. J. Intell. Fuzzy Syst.
**2022**, 42, 1099–1114. [Google Scholar] [CrossRef] - Yan, Q.D.; Chen, X.Q.; Jian, H.C.; Wei, W.; Wang, H. Design of a deep inference framework for required power forecasting and predictive control on a hybrid electric mining truck. Energy
**2022**, 238, 121960. [Google Scholar] [CrossRef] - Shen, H.; Wang, Z.; Zhou, X.; Lamantia, M.; Yang, K.; Chen, P.; Wang, J. Electric Vehicle Velocity and Energy Consumption Predictions Using Transformer and Markov-Chain Monte Carlo. IEEE Trans. Transp. Electrif.
**2022**, 8, 3836–3847. [Google Scholar] [CrossRef] - Eddine, M.D.; Shen, Y. A deep learning-based approach for predicting the demand of electric vehicle charge. J. Supercomput.
**2022**, 78, 14072–14095. [Google Scholar] [CrossRef] - Eagon, M.J.; Kindem, D.K.; Panneer Selvam, H.; Northrop, W.F. Neural Network-Based Electric Vehicle Range Prediction for Smart Charging Optimization. J. Dyn. Syst. Meas. Control
**2022**, 144, 011110. [Google Scholar] [CrossRef] - Wang, Z.; Abdallah, A.B. A Robust Multi-Stage Power Consumption Prediction Method in a Semi-Decentralized Network of Electric Vehicles. IEEE Access
**2022**, 10, 37082–37096. [Google Scholar] [CrossRef] - Huang, N.E.; Shen, Z.; Long, S.R.; Wu, M.C.; Shih, H.H.; Zheng, Q.; Yen, N.C.; Tung, C.C.; Liu, H.H. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proc. R. Soc. Lond. Ser. A Math. Phys. Eng. Sci.
**1998**, 454, 903–995. [Google Scholar] [CrossRef] - Rilling, G.; Flandrin, P.; Goncalves, P. On empirical mode decomposition and its algorithms. In Proceedings of the IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing, Grado, Italy, 8–11 June 2003; Volume 3, pp. 8–11. [Google Scholar]
- Abualigah, L.; Diabat, A.; Mirjalili, S.; Abd Elaziz, M.; Gandomi, A.H. The arithmetic optimization algorithm. Comput. Methods Appl. Mech. Eng.
**2021**, 376, 113609. [Google Scholar] [CrossRef] - Agushaka, J.O.; Ezugwu, A.E. Advanced arithmetic optimization algorithm for solving mechanical engineering design problems. PLoS ONE
**2021**, 16, e0255703. [Google Scholar] [CrossRef] [PubMed] - Kaveh, A.; Hamedani, K.B. Improved arithmetic optimization algorithm and its application to discrete structural optimization. Structures
**2022**, 35, 748–764. [Google Scholar] [CrossRef] - Yu, Y.; Si, X.; Hu, C.; Zhang, J. A review of recurrent neural networks: LSTM cells and network architectures. Neural Comput.
**2019**, 31, 1235–1270. [Google Scholar] [CrossRef] - Udaiyakumar, S.; Aruldoss Albert Victoire, T. Week ahead electricity price forecasting using artificial bee colony optimized extreme learning machine with wavelet decomposition. Teh. Vjesn.
**2021**, 28, 556–567. [Google Scholar] - Sherstinsky, A. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Phys. D Nonlinear Phenom.
**2020**, 404, 132306. [Google Scholar] [CrossRef] - Campus Electric Vehicle Charging Stations Behavior. Available online: www.kaggle.com/datasets/claytonmiller/campus-electric-vehicle-charging-stations-behavior/metadata (accessed on 16 February 2022).
- Macioszek, E. Electric vehicles-problems and issues. In Scientific and Technical Conference Transport Systems Theory and Practice; Springer: Cham, Switzerland, 2019; pp. 169–183. [Google Scholar]
- Wang, R.; Xing, Q.; Chen, Z.; Zhang, Z.; Liu, B. Modeling and Analysis of Electric Vehicle User Behavior Based on Full Data Chain Driven. Sustainability
**2022**, 14, 8600. [Google Scholar] [CrossRef] - Macioszek, E. E-mobility infrastructure in the Górnośląsko-Zagłębiowska Metropolis, Poland, and potential for development. In Proceedings of the 5th World Congress on New Technologies (NewTech’19), Lisbon, Portugal, 18–20 August 2019; p. 108. [Google Scholar]

**Figure 1.**Statistics of EVs across countries (Courtesy—statista.com (accessed on 16 February 2022)).

Charging Mode | Power Rating (P) | Supply |
---|---|---|

Normal Power Charging | Less than or equal to 7 kW | DC and AC |

7 kW to 22 kW | DC and AC | |

High Power Charging | 22 kW to 50 kW | Only DC Supply |

50 kW to 200 kW | Only DC Supply |

Author | Method Adopted | Challenges |
---|---|---|

Mao et al. (2019) [9] | Parallel gradient boosting decision tree algorithm | Delayed scheduling capacity and energy demand |

Saputra et al. (2019) [10] | Variants of machinelearningtechniques | Only for short-term prediction |

Zhang et al. (2020) [12] | Extreme learning machine algorithm | Stuck with algorithmic stagnation issues |

Sun et al. (2020) [14] | Hybrid artificial intelligence techniques | Accuracy was not guaranteed due to rule base employed |

Thorgeirsson et al. (2021) [22] | Probabilistic prediction models | Invariant time frame of analysis |

Shahriar et al. (2021) [23] | Hybrid Machine learning algorithms | Longer session duration of prediction with higher error |

Cadete et al. (2021) [24] | Autoregressive and moving average models | Higher error variations with minimized accuracy |

Liu et al. (2021) [25] | Back propagation neural network | Over-fitting and under-fitting problems |

Zeng et al. (2021) [32] | Optimization-oriented adaptive training predictor | Difficulty in data sharing |

Ye et al. (2021) [34] | Deep learning algorithm | Increased layer complexity and delayed convergence |

Asensio et al. (2022) [44] | Kalman Filter scheme and auto regressive models | Difficulty in handling temporal files |

Liu et al. (2022) [45] | Intelligent sensing system | Difficulty in handling non-linear time dependent data |

Shi et al. (2022) [46] | Deep auto-encoded extreme learning machine | Difficulty in analysing the long-term energy consumption prediction |

Charging Time (hh:mm:ss) | Energy (kWh) | GHG Savings (kg) | Gasoline Savings (Gallons) | Cost Incurred (USD) |
---|---|---|---|---|

01:11:50 | 6.249 | 2.625 | 0.784 | 1.02 |

00:58:15 | 4.352 | 1.828 | 0.546 | 0.83 |

01:11:24 | 4.341 | 1.823 | 0.545 | 1.02 |

03:19:30 | 7.857 | 3.3 | 0.986 | 4.12 |

01:58:14 | 6.075 | 2.551 | 0.762 | 1.68 |

02:23:58 | 7.758 | 3.258 | 0.974 | 2.04 |

01:37:25 | 9.55 | 4.011 | 1.199 | 1.39 |

03:24:24 | 11.275 | 4.735 | 1.415 | 2.9 |

01:01:50 | 6.061 | 2.546 | 0.761 | 0.88 |

01:23:29 | 4.019 | 1.688 | 0.504 | 1.19 |

02:29:45 | 9.085 | 3.816 | 1.14 | 2.12 |

01:54:49 | 5.153 | 2.164 | 0.647 | 1.78 |

06:20:20 | 19.28 | 8.098 | 2.42 | 6.17 |

04:47:36 | 12.334 | 5.18 | 1.548 | 4.73 |

03:12:16 | 9.975 | 4.19 | 1.252 | 2.73 |

03:10:19 | 12.815 | 5.382 | 1.608 | 2.7 |

03:27:05 | 13.826 | 5.807 | 1.735 | 2.94 |

04:37:06 | 12.408 | 5.211 | 1.557 | 4.25 |

04:48:07 | 18.272 | 7.674 | 2.293 | 5.56 |

03:57:31 | 9.594 | 4.03 | 1.204 | 4.52 |

04:43:32 | 12.967 | 5.446 | 1.627 | 4.5 |

02:06:29 | 4.789 | 2.011 | 0.601 | 1.96 |

06:11:21 | 19.524 | 8.2 | 2.45 | 8.96 |

04:45:17 | 15.359 | 6.451 | 1.927 | 4.05 |

05:07:46 | 12.336 | 5.181 | 1.548 | 4.36 |

04:18:45 | 14.575 | 6.122 | 1.829 | 3.67 |

00:01:34 | 0.081 | 0.034 | 0.01 | 0 |

Parameters | Values of Predictor Model | Parameters | Values of Predictor Model |
---|---|---|---|

Number of Deep layers | 11 | Recurrent LSTM memory states | 8 |

Number of IMFs | 4 | Dropout probability | 30% |

Input neurons | 5 | Output neurons | 1 |

Convolutional layer neuron nodes | 30 | Activation | Sigmoidal function |

Pooling layer nodes | 30 | Max iterations | Till the convergence criterion |

LSTM layer neurons | 70 | No. of trial runs | 32 |

Learning rate metric | 0.01 | β | 5 |

Number of populations | 40 | λ | 0.5 |

Convergence criterion | 10^{−6} |

EV Datasets | Performance Metrics | |||
---|---|---|---|---|

MAE | MSE | RMSE | A_{pre} | |

Georgia Tech EV charging station, USA | 0.1083 | 4.25516 × 10^{−10} | 2.0628 × 10^{−5} | 0.9714 |

Training | Testing | ||
---|---|---|---|

Epochs | Mean Square Error | Epochs | Mean Square Error |

10 | 4.2516 | 10 | 5.7164 |

50 | 1.0291 | 50 | 3.9917 |

100 | 0.2098 | 100 | 0.6310 |

150 | 3.0816 × 10^{−3} | 150 | 7.5518 × 10^{−3} |

200 | 7.1892 × 10^{−5} | 200 | 9.9421 × 10^{−5} |

250 | 2.1163 × 10^{−7} | 250 | 5.2219 × 10^{−7} |

251 | 4.25516 × 10^{−10} | 251 | 5.96333 × 10^{−10} |

Georgia Tech EV Charging Station Dataset | |||
---|---|---|---|

Actual Charging Energy Output (kWh) | Predicted Charging Energy Output (kWh) | Actual Charging Energy Output (kWh) | Predicted Charging Energy Output (kWh) |

13.575 | 13.500 | 9.903 | 9.900 |

5.952 | 6.014 | 0.976 | 0.971 |

19.58 | 19.726 | 7.053 | 7.048 |

5.617 | 5.229 | 5.552 | 5.488 |

18.795 | 18.903 | 1.179 | 1.184 |

4.302 | 4.007 | 13.137 | 13.130 |

10.818 | 10.800 | 18.278 | 18.275 |

6.32 | 6.296 | 2.126 | 2.130 |

19.485 | 19.501 | 6.939 | 6.900 |

9.132 | 9.127 | 10.812 | 10.810 |

12.507 | 12.500 | 9.039 | 9.010 |

10.314 | 10.310 | 8.025 | 8.022 |

7.8 | 7.779 | 9.208 | 9.201 |

12.269 | 12.261 | 0.891 | 0.890 |

19.737 | 19.730 | 6.952 | 6.950 |

13.902 | 13.900 | 11.996 | 11.990 |

10.033 | 10.000 | 16.789 | 16.790 |

Prediction Techniques Adopted | Georgia Tech EV Charging Datasets | ||||
---|---|---|---|---|---|

MSE Error | Training Efficiency % Mean | Testing Efficiency % Mean | Computational Time (s) | Prediction Accuracy % | |

Bayesian ELM neural model [1] | 5.1304 | 83.26 | 82.77 | 16.14 | 87.09 |

Federated learning approach [10] | 2.3478 | 85.19 | 84.76 | 15.83 | 89.66 |

Ensemble learning [13] | 0.3367 | 89.03 | 88.64 | 14.36 | 90.01 |

Bayesian optimization with ML [16] | 0.1649 | 89.48 | 88.91 | 14.87 | 90.73 |

Probabilistic prediction [22] | 0.0021496 | 91.45 | 90.67 | 15.04 | 93.26 |

LSTM model [24] | 0.00487632 | 93.51 | 92.48 | 11.60 | 95.03 |

Back propagation model [27] | 0.0002148 | 93.64 | 93.37 | 10.48 | 95.67 |

Deep inference framework [50] | 8.11476 × 10^{−4} | 94.18 | 94.02 | 9.84 | 96.48 |

Deep learning model [52] | 6.13857 × 10^{−6} | 96.44 | 94.27 | 9.15 | 96.91 |

Proposed EMD–AOA–DLSTM neural predictor | 4.25516 × 10^{−10} | 98.62 | 98.03 | 7.4 | 97.14 |

Proposed Neural Model | Correlation Coefficient | Coefficient of Determination |
---|---|---|

Deep LSTM neural model | 0.9967 | 0.9993 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Shanmuganathan, J.; Victoire, A.A.; Balraj, G.; Victoire, A.
Deep Learning LSTM Recurrent Neural Network Model for Prediction of Electric Vehicle Charging Demand. *Sustainability* **2022**, *14*, 10207.
https://doi.org/10.3390/su141610207

**AMA Style**

Shanmuganathan J, Victoire AA, Balraj G, Victoire A.
Deep Learning LSTM Recurrent Neural Network Model for Prediction of Electric Vehicle Charging Demand. *Sustainability*. 2022; 14(16):10207.
https://doi.org/10.3390/su141610207

**Chicago/Turabian Style**

Shanmuganathan, Jaikumar, Aruldoss Albert Victoire, Gobu Balraj, and Amalraj Victoire.
2022. "Deep Learning LSTM Recurrent Neural Network Model for Prediction of Electric Vehicle Charging Demand" *Sustainability* 14, no. 16: 10207.
https://doi.org/10.3390/su141610207