This journal paper, authored by6, presents a comprehensive conceptual framework for the successful integration of EVs into electric power systems. The framework encompasses two main domains: the technical operation of the grid and the electricity markets environment. The paper provides a detailed description of the various stakeholders involved in these processes and their respective activities. Additionally, several simulations are conducted to demonstrate the potential impacts and benefits that arise from the integration of EVs into the grid under this framework. These simulations include analyses of steady-state and dynamic behaviors7. The paper discussed three types of EVs that are currently relevant in the market: fully EVs, fuel cell EVs, and hybrid EVs. It acknowledges that the widespread adoption of these vehicles, particularly those relying solely on electric power, will have significant implications for the design and operation of electric power systems. In summary, this paper presents a comprehensive conceptual framework for integrating EVs into electric power systems, addressing technical and market aspects. It emphasizes the challenges and benefits of EV integration and proposes strategies for effective operation and management. The presented simulations provided insights into the potential impacts and highlight the importance of advanced charging control strategies and local-level control for optimal system performance.
This journal article, written by8, was centered on the creation of streamlined EV powertrain models for brand-new and current production cars, particularly the Tesla Roadster and Nissan Leaf. The models are based on published vehicle parameters and range information are compared with manufacturer specifications for range under various driving conditions and drive cycles. To validate the models, test results for the Tesla Roadster and Nissan Leaf are used, where a GPS-based smartphone app and Google Earth are used to simulate the geography of the test path. The paper demonstrated excellent correlations between the model projections, manufacturer data, and experimental outcomes. The study also considered the impacts of battery degradation over time and vehicle HVAC loads. These factors are considered to provide a more realistic assessment of the EV powertrain performance. Overall, the paper provides valuable insights into the performance of EV powertrains and their range capabilities.
Cai et al.9 discusses the energy issue and the state of ecosystem. According to them, the distribution network was facing new difficulties because of the growth of electric cars. The charging plan, the position and size planning for EV charging sites, and the cooperation coordination among EV and the distribution network all rely on the projection of charging demand and the scope of future EV growth. They assert that the growing size and charging capacity forecasting model for EV are established using the artificial neural network technique. The model’s accuracy was demonstrated through an example of Kunming, a significant city in West China, and the projection of the size and filling capacity of EVs. It has been determined that their study offers a fresh approach to forecasting China’s electric car market’s future growth size and charging load.
The author10 addresses the problem of range anxiety and the variable anticipated range left in electric cars, which prevents their widespread adoption, in this article. The goal of the research is to comprehend the root causes of potential range prediction errors and how intelligent transportation systems (ITS) might assist in finding solutions. Eleven participants made 141 documented trips, and the findings showed that the range projected by the EV and provided to the driver was overstated by around 50% in comparison to the actual trip distance. According to the research, driving aggressively results in increased mistakes and has the most influence on range forecast accuracy. In summary, the paper reveals that range predictions in EVs are often overestimated compared to actual journey distance. Driving style and journey characteristics play significant roles in range prediction accuracy. The study suggests incorporating these factors into range prediction algorithms and utilizing intelligent systems to improve accuracy. It also highlights the importance of the driver’s behavior in maximizing the available range of an EV.
The author11 sought to identify and measure links between the vehicle’s kinematic parameters and its energy usage. To build the energy consumption calculation models, they used actual statistics on EV energy consumption. With the vehicle dynamics equation as the base physical model, they built three models using multiple linear regression. The input variables (predictors) used by each model are aggregated at a separate level, allowing prediction based on the input data available. One approach aggregates kinematic parameters in trips; another extends this model with fine-grained acceleration parameters over the journey; and the third uses fine-grained kinematic parameter values to forecast micro-trips energy consumption. Multiple linear regression (MLR) having a much higher correlation coefficient was shown to be a successful strategy for estimating energy usage using aggregated tour data by accounting for a significant percentage of the variability included in the data. The findings showed that increasing the degree of detail in a model has the potential to produce one that is more accurate. Finally, concluded as non-linear effects of regenerative braking are mapped by connecting traffic situations and driving style with CMF to create accurate models.
The author12 provides an overview of research on measuring and estimating energy consumption in EVs. The study emphasizes the significance of EVs in reducing oil dependence, improving efficiency, and mitigating carbon emissions. They developed a data collection system and analyzed 5 months of data to evaluate EV performance and driver behaviors. The findings revealed higher efficiency on in-city routes compared to freeways, with drivers adjusting behavior based on real-time energy usage information. The research explored the relationships between EV power and variables such as velocity, acceleration, and roadway grade, highlighting their impact on energy consumption. They proposed an empirical method and an analytical model for EV power estimation, successfully predicting instantaneous power and trip energy consumption. The study also demonstrated the feasibility of data collection and provides insights for enhancing EV energy efficiency. Limitations include a restricted dataset involving a single driver and vehicle, calling for broader studies with multiple drivers and commercially produced vehicles is required to validate the estimation model. Overall, this research contributes to understanding EV energy consumption and offers valuable insights for researchers and EV users.
In this article13, the difficulties in accurately estimating the energy consumption of EVs are discussed and potential solutions are provided. EVs are complex devices with multiple concurrent processes involving energy consumption and generation within different onboard systems. Achieving more precise simulations of energy consumption is crucial for gaining a better understanding of energy management in electric transport and ultimately contributing to a sustainable future with enhanced convenience in terms of distance range and recharging time. This paper addresses the problem of energy consumption simulation for EVs using various software packages and offers insights on making the simulation process more precise. By doing so, engineers can develop improved energy management strategies for EVs. Computer simulation plays a crucial role in the development and optimization of EV systems, and accurate modeling is a central challenge. Specific testing cycles such as NEDC, JC-08, and EPA cycle are utilized for powertrain simulation and optimization in different regions. In summary, this paper underscores the challenges of simulating energy consumption in EVs and proposes methods to enhance accuracy. It emphasizes the importance of precise modeling for improving energy management strategies, optimizing powertrain systems, and facilitating overall performance enhancement in EVs.
Several contrast methods14 using participatory sensing for predicting the individual energy (1) a comparison to the average using personalized customization; (2) two techniques to resemblance matching based on vehicle/driver/environment-dependent characteristics utilising pace profile comparison and driving habit match; and (3) an ensemble filtration strategy utilising matrix factorization. To find variables that depend on the driver, the car, and the surroundings, they also used a collaborative filtering technique based on matrix factorization and a black box framework. As there is a lack of systematic research on the effective use of the data for individualized prediction even though it can provide a variety of driving data.
Additionally, they conducted a case study using data from participatory sensing to forecast the distance to empty for electric vehicles, and they empirically assessed their results, that demonstrated that their approaches could greatly increase prediction accuracy. De Cauwer et al.15 to address the issue of range anxiety, they provided an energy consumption forecast method for EVs that was designed for energy-efficient routing. Using real-world measured driving data, geography data, and weather data, this data-driven technique anticipates consumption across every specific road in a road network.
They have approximated the energy consumption over road segments using a MLR models that links the energy consumption with minor driving characteristics and external factors. A neural network (NN) is utilized to forecast the unknowable microscopic driving parameters over a segment before departure based on the features of the road segment and the weather16. This method enables cost-optimization methods to choose energy-efficient paths and allows for the pre-departure forecast of every road in the system, there is energy usage.
The advantage of the data-driven method, according to their argument, is that the models can be easily modified in long run time to reflect changing circumstances. The development of a driving range estimation model for electric cars considering the influence of environmental temperature and driving circumstances is the main topic of this paper, written by17. Even though EVs are popular the limited energy density, high costs, and short cycle life of power batteries result in a restricted driving range compared to conventional vehicles. This research suggested a technique for estimating driving range that combines driving cycle recognition and prediction in order to overcome these difficulties. The driving cycle was first located using fuzzy C referring to clustering and Kernel principal component feature parameters. MATLAB/Simulink version R2023b18 is used to construct a fuzzy rule between the characteristic parameters and energy usage. The driving range estimation method is validated using a rotary drum test bench under the ECE 15 condition, and the results are compared with the estimation results of actual driving mileage. In summary, this paper presents a driving range estimation model for EVs that incorporates the effects of environmental temperature and driving conditions. The proposed method provided a new approach to estimate the driving range of EVs, considering various factors and optimizing energy consumption modeling.
The amount of energy needed for a future journey19 rely on a variety of elements, including driving style, knowledge of the topography of the road, the weather, and the flow of traffic. To anticipate the energy consumption for a future journey, Jiquan Wang, Igo Besselink, and Henk Nijmeijer discussed an algorithm that comprises of an offline algorithm and an online algorithm. They contend that the offline algorithm is intended to offer information for the driver to create future driving plans, whereas the online algorithm is intended to adjust the energy consumption prediction result based on current driving. They can confirm their energy consumption prediction algorithm with the aid of 30 driving tests.
The observed energy consumption for all trips was within the offline algorithm’s allowable range, and the bulk of the differences between the measurement and nominal prediction was less than 10%, according to a comparison of the data20. Over the years, there has been a steady rise in the desire for electricity. A decent predictive model is, therefore, necessary to comprehend future consumption. Additionally, it aids in creating new plants and networks in order to prepare for anticipated increases in demand for electricity. Hence, the extensive use of ARIMA models for time series prediction by Praphula Jain, Waris, and Rajendra has produced encouraging findings. Using the ARIMA model, they have tried to predict how much energy will be used. After analyzing the electricity usage in IIT, (ISM) for the years 2004 to 2008, the seasonal ARIMA model was determined to be the best model. It was also able to predict consumption for the years 2008 and 2009.
Driving ranges21 of many EVs running on highways are projected by the authors, and this knowledge was utilized to develop a recommendation system. By using actual statistics on EV trips to build data-driven models, they can compute recommendations based on energy consumption. For well-known EV models, extremely precise prediction models are created using the authors’ method. Prediction accuracy for new EV models, however, is lower than for popular EV models since fewer journeys by new EV models are made on the motorway than by well-known EV models. In order to solve this problem, the authors proposed a novel transfer learning strategy. A type of machine learning called transfer learning creates prediction models utilizing more than enough data on well-known EV models. They have put forth a novel transfer learning technique that incorporates more data to develop forecast models, more full information on popular EV models. Additionally, they tested their approach using real EV trip data. They claim that consequently, their approach had a prediction error rate that was 30\% lower than the traditional approach. Finally, it was found that the suggested method might be able to predict new EV models’ energy consumption more precisely.
A Model for Predicting Electric Energy Consumption (EECP-CBL) that predicts electric energy consumption by combining a convolutional neural network (CNN) and a bi-directional long short-term memory (Bi-LSTM)22. Two CNNs are used in this study’s initial module of the individual house electric power consumption dataset, that gathers the key information from a variety of sources. To create predictions, a Bi-LSTM module with two Bi-LSTM layers uses the data indicated above as well as time series patterns in both the forward and backward routes. They basically carried out these experiments to compare the forecasted results of the proposed model and the state-of-the-art models for the IHEPC dataset with various changes. Finally, findings showed that the EECP-CBL framework outperformed the most recent methods in terms of several performance measures for predicting electric energy usage. The efficacy of the electric energy usage prediction model is said to be enhanced in the future using various methods, including evolutionary algorithms.
This literature23 summarizes the research conducted by on the stability and power consumption of EVs using different modern control strategies. The study introduces a novel control approach based on Artificial Neural Networks (ANNs) to predict the yaw moment for ensuring the lateral stability of EVs with four in-wheel motors. Computer simulations are conducted to evaluate the robustness and power consumption of the proposed Neural Network Controller (NNC) compared to Sliding Mode Control (SMC). The results show that the NNC achieves stable motion near the driving limits with slightly lower power consumption compared to SMC. Additionally, the soft computing controllers exhibit robustness against system uncertainty and consume satisfactory energy from the electric motors. This research contributed to understand stability control in EVs and highlighted the potential benefits of employing modern control strategies to enhance stability and reduce power consumption.
Time series ML analysis is used for forecasting in many different industries. ML-driven data series analysis can help predict the following: Demand and sales. ML can help analyze historical data to predict customer demand or sales. In some cases, there may be a scarcity of historical time series data, making it challenging for ML models to learn meaningful patterns. In certain cases, time series data may lack clear patterns or trends, making it difficult for ML algorithms to identify meaningful relationships and make accurate predictions. Poor quality or inconsistent time series data can negatively impact the performance of ML models, leading to inaccurate predictions and unreliable insights. Certain industries or domains may have unique challenges that are not easily addressed by generic ML models.
This work summarizes the research on a novel ensemble method for forecasting EV power consumption in Spain24. The study addresses the challenges associated with increasing EV usage and the need for power companies to adapt their generation accordingly. The proposed approach combines ARIMA, GARCH, and PSF algorithms using ensemble learning to forecast EV power consumption. The non-stationary nature of the time series adds complexity, leading to the dynamic weighting of algorithms over time. The study demonstrated the effectiveness of the approach using the Weighted Absolute Percentage Error (WAPE) metric. The research contributes by developing an ensemble algorithm, analyzing non-stationary time series, applying the method to real data, and implementing it at the EV Control Center. Coefficients are periodically updated to handle evolving data, and future research could focus on optimizing update frequency and exploring specialized treatments for holidays and different charging station aggregations. Overall, this research advances the field of EV power consumption forecasting and showcases its practical applicability in the Spanish system.
To forecast the demand for charging new EV models with larger battery capacities, A DCFCS dynamic planning approach that considers user behaviour and probabilistic driving patterns was developed by Marjan Gjelaj and colleagues25. EVs (electric cars) seem to be a viable option for promoting green transportation and reducing CO2 emissions in urban areas. It is recommended to have saved charging need and synchronized charging need to lower the peak load from EVs and the expense of the charging infrastructure.
They considered a few configurations, such as synchronized storage feeding demand and charging demand, inside the stochastic planning technique. To reduce the running costs of the DCFCS and the peak demand for EVs, they have proposed an ideal BES as a replacement strategy. They also used setup strategies as a optimum multipurpose design issue with the primary objective of lowering grid-reinforcement costs. The last phase was to carry out an economic analysis to evaluate the technical and financial elements of DCFCSs, the life-cycle costs of BESs, and the financial performance of BES costs in relation to grid-reinforcement expenses. Predicting the charging demand of PEVs (the energy consumed during the charging session) could aid in the efficient management of the electric grid26. One of the key components of green buildings and microgrids is now energy usage tracking. A dataset with information on charging via public charging facilities in Nebraska, USA was used, over the course of seven years. They have also boldly used data from many stations to apply the predicted framework.
In that case, it is probable that those input variables have an even stronger link with consumption of energy, improving estimates for a smaller percentage of consumers. The same framework may be used with data from a smaller area or even just one station. The challenge they ran into in this study was trying to build a prediction model to explain charging and parking habits after analyzing a substantial amount of semi-random data. They assert that by examining the charging trends at both public and private charging stations, this analysis can be strengthened.
In this article, Kim and Kim2 examine the most recent EV energy usage models with the goal of offering recommendations for the development of EV apps in the future. They discuss EV energy consumption models in terms of modeling scale (microscopic vs. macroscopic), methodology, and they essentially divide the influencing aspects of EV energy consumption to four categories: vehicle dynamics, traffic, environment-related factors, and automobile component (data-driven as opposed to rule-based). Their investigation shows trends of rising macroscopic models that can be used to predict trip-level EV energy consumption, as well as growing data-driven algorithms that estimated EV consumption of energy via vast amounts of real-world information and machine learning techniques. And it is discovered that rule-based models predominate earlier literature, whereas data-driven EV energy usage estimation and its applications have been drawing growing study interest in the last few years. Finally, they concluded that multi-scale energy estimation models should be developed as a comprehensive modeling strategy and that energy estimation models appropriate for applications linked to vehicle-to-grid integration should also be developed.
Using many scaled geographic datasets over a 2-year period, Kim and Kim27 compare the forecasting approaches used to account for increases in power consumption brought on by the growing acceptance of EVs. They compared various modeling techniques based on historical data and exogenous variables, including long short-term memory (LSTM) modeling and Trigonometric, Box-Cox, auto-regressive moving average (ARMA), trend, and seasonality (TBATS) are other names for trigonometry exponentially smoothing state space. It was found that the historical data’s importance was confirmed, and the effects of exogenous variables are evaluated on both macro and micro-scale geographic regions. When they compared time-series techniques with machine learning techniques, they found that machine learning had the advantage of requiring comparatively less model-fitting assumptions and straightforward hyperparameter tuning, but that it was not always highly predictive. They finally reached a decision. For macro data with generally straightforward patterns, the ARIMA model with regressors produced the best results.
With the help of other researchers, Straka et al.28 created a data-centric methodology to examine how the functions, behaviors, and attributes of the environment around slow-charging facilities affect how energy is distributed there. They examined the likelihood distribution of energy use and its relationship to indicators of charging occurrences to gain some fundamental knowledge. They gathered geospatial information and produced numerous potential features that modeled the physical environment in which the charging infrastructure functions. It was found and analyzed a relatively small subset of the most significant characteristics that are correlated with energy usage using statistical techniques. They distinguished the chosen characteristics by applying the approach to a particular class of charging infrastructure, as decided, for example, by the rollout strategy employed. The proximity of public venues, working residents, and business categories are associated with greater energy usage at carefully placed charging infrastructure. Age-related characteristics of the populace have an impact on how much energy is used at recharge stations that are installed according to demand. It was discovered that their work offered insightful information on the types of data to gather and incorporate into prediction models to help with the implementation of charging infrastructure and the planning of power networks.
A thorough breakdown of the many aspects of the EV industry and its charging system is provided by Nagaraju Dharavat, Naresh Kumar, and others29. Additionally, It provides a step-by-step procedure for implementing the Vehicle to Grid (V2G) concept, explains how artificial intelligence can be used to capture data from the EV battery, and analyses the costs and benefits of using the V2G method effectively. This article also includes a list of various EVs, storage locations, methods for charging EVs using DGs combined with EVCS, as well as a variety of other socio-technical concerns with EVs. The acceptance rate and current stage of EVs globally have received attention.
EVs are a desirable way to reduce rising fuel use and GHG emissions, but their widespread use may compromise the distribution system’s reliability30. As a result, many methods are used to forecast the charging of EVs. Using a dataset made up of 2000 observations of charging events gathered from two public charging stations in Morocco, Mouaad Boulakhbar and colleagues compare the performance of four well-known deep learning models, namely ANN, recurrent neural networks (RNNs), LSTM, and gated recurrent units (GRUs), in predicting charging demand for EVs. The findings demonstrate that all four regression methods are capable of accurately predicting Morocco’s demand for EV charging. They claim that these findings can both short-term ensure the grid’s dependability and long-term direct the Moroccan National Office of Electricity and Water to create more charging stations. Mediouni et al.31 suggested a hybrid strategy to achieve this goal by accounting for factors like extra weight, road conditions, and driving habits. Physical and equation-based models are used to simulate the major EV parts. They primarily used a large synthetic dataset to illustrate various driving situations. A city car was also used to gather data from the real world. To connect mechanical and electric power, they created a machine-learning algorithm. In terms of R2 and root mean square error (RMSE), it turned out that the proposed models performed well. They also added that it might be useful for route planning by EV users to lessen range-anxiety, as well as for people making judgments regarding autos for the appropriate sizing of parts like the battery and powerplant.
This paper by Wang and Abdallah32 introduces the concept of a semi-decentralized robust network of electric vehicles (NoEV) integration system for efficient power management in a smart grid platform. The primary objective of this system is to tackle the challenge of balancing energy supply and demand, particularly in the face of fluctuating energy generated by renewable resources. To achieve this, the proposed approach integrates an aggregator with EV fleets using a blockchain framework. The system employs a multi-stage algorithm executed by the EVs, that incorporated a novel federated learning algorithm called Federated Learning for Qualified Local Model Selection (FL-QLMS) to predict power consumption accurately. The paper emphasized the significance of efficient distribution and utilization of renewable energy, highlighting the virtual power plant (VPP) as a vital intermediary in the smart grid ecosystem. NoEV addresses critical factors such as robustness, cost-efficiency of data storage, rapid response to demand, and scalability. In conclusion, this paper presented a sophisticated semi-decentralized system, leveraging blockchain technology and innovative federated learning algorithms, to integrate EVs into power management effectively. The NoEV system demonstrated improved efficiency, accuracy, and robustness in addressing the challenge of balancing energy supply and demand in the smart grid domain.
In order to meet the FCS’s demand for electricity while reducing the cost of production and energy loss, energy management and optimization are consequently crucial for the mixed energy system33. A mixed-integer linear programming-based energy management technique was created as part of a model reference adaptive control to address this issue and enhance the charging station’s performance. The results of modeling and testing the proposed system with the MATLAB/Simulink software are discussed. The evaluation finds that the recommended energy management system offers a performance that is optimum for the rapid charging station’s integration with nuclear and renewable energy.
