Figures
Explore figures and images from publications
Figure 1 - available via license: CC BY
Content may be subject to copyright.
Basic RNN structure.

Basic RNN structure.

Source publication
Article
Full-text available
Load forecasting is one of the major challenges of power system operation and is crucial to the effective scheduling for economic dispatch at multiple time scales. Numerous load forecasting methods have been proposed for household and commercial demand, as well as for loads at various nodes in a power grid. However, compared with conventional loads...

Contexts in source publication

Context 1
... load forecasting problem, when the load value at time t-1 as the input of RNN, the model can output the load value of time t, RNN models can better capture the characteristics of input data by using recurrent structure shown in Figure 1. ...
Context 2
... R2 of LSTM is very close to one in three time steps, which indicates that the LSTM has been significantly competitive for the super-short-term EV charging load forecasting. The metrics and error comparison histograms in three scenarios are shown in Figure 10. It can be seen more intuitively from the histograms that LSTM has the lowest error and the best goodness of fit compared to five other methods. ...
Context 3
... with the best results obtained from all the counterparts, the LSTM method has an average of 30% lower errors on all the index criteria. Figure 11 shows the actual data curve and forecasting curve for each model in the three scenarios of a single day and a whole week. For the single day in the left column, 1440 points is described in the curve. ...
Context 4
... LSTM model again perfectly predicts the load of each point and captures all the slight step changes. For a whole week in the right column of Figure 11, the data from 15 February 2018 to 22 February 2018 are adopted and 10,080 points are considered in the curve. It could be easily seen that, due to the inherent working mode of public transports, the curve shows a certain periodicity. ...
Context 5
... this scenario, the forecasting accuracy of the RNN model is reduced, where the forecasting value of peak and valley load is higher than the actual value in the one time step case and it is lower than the true data in the 15 time steps case. As shown in Figure 12, the forecast results of the holiday (dry season) and the working day (rainy season) are compared. From the forecast of the working day (rainy season), the charging load peaks at 11:00 p.m., begins to decrease at 3:00 a.m., and increases again at 8:00 a.m. ...
Context 6
... load curves of the PEV charging station persist highly periodical characteristics, which might not be convincing enough to demonstrate the superior performance of deep learning methods. In order to further verify the validity of the proposed model, another minute-level load dataset of a PEV aggregator for commercial building chargers in Shenzhen was used to validate the proposed super short-term model, and the prediction results' curves are shown in Figure 13. It could be observed that the charging behavior of a commercial charges aggregator is more random and fluctuating, which is completely different from the charging station profile. ...

Similar publications

Article
Full-text available
p>The main focus of this paper is a study that empowers us to understand how the temperature variation affects the transmission line resistance and as a result the power flow analysis with a specific end goal to assess losses in the electrical network. The paper is composed of two sections; the first part is a power flow study under normal conditio...

Citations

... On the other hand, evolving intelligent algorithms have been commonly used in recent years to forecast short-term load for EV charging, with their applicability verified in a large number of studies. Among them, intelligent algorithms such as neural networks are often used [34][35][36][37], support vector machine [38][39][40], and deep learning [41]. As shown in the preceding study, these studies do not account for changes in EV ownership, and thus are only applicable to short-term forecasting of EV charging load. ...
Article
Predicting the charging load of electric vehicles (EVs) is critical for the safe and reliable operation of the distribution network. Analyzing an EV's random charging characteristics and the uncertainty associated with its development scale are important to accurate prediction of its charging load. For this reason, we proposed a seminal method for predicting EV charging load based on stochastic uncertainty analysis. This included not only a probabilistic load model for describing the stochastic characteristics of the EV charging, but also an ownership forecasting model for estimating the EV development scale. EVs are classified into four categories based on their intended use: electric buses, electric taxis, private EVs, and official EVs. The corresponding load calculation model was developed by analyzing the charging behavior of various EVs. Simultaneously, the improved grey model method (IGMM) based on the Fourier residual correction is used to accurately forecast EV ownership. Finally, the scientific method of Monte Carlo simulation(MCS) was used to estimate the charging load demand of EVs. This method was used in Wuhan that has a lot of potential for EV production. As compared to the basic grey model method (BGMM), the IGMM outlined in this work can triple the prediction effect. Due to the large-scale charging of EVs, Wuhan's maximum daily total load would rise to 15,532.9 MW on working days and 15,475.5 MW on rest days in 2025. Additionally, the total load curves on working days and rest days will show a new peak load with the value of 14751.3 MW and 14787.2 MW at 14:01, resulting in an increase of 13.56% and 13.83% respectively in the basic daily load stage. As a result, it is necessary for grid operators to build adequate capacity to meet EV charging demands, while developing rational and orderly charging strategies to avoid the emergence of new load peaks.
... They found that the regression method, XG-Boost, outperforms the other two methods in predicting the charging demand. Also, ( Zhu et al. 2019 ) conducted a comparative analysis of deep learning approaches to forecasting the short-term charging load of electric vehicles. Their results indicated that deep learning models could accurately forecast super-short-term PEV charging load. ...
Article
Full-text available
The transport sector is a prominent source of increasing fuel consumption and greenhouse gas (GHG) emissions. Electric vehicle (EV) is deemed an appealing solution for those problems. However, due to the variation in charging demands, the high penetration of electric vehicles may cause different problems to the distribution network and its reliability. Therefore, several approaches are employed to predict the EVs charging demand and avoid the corresponding challenges. This paper compares the performance of four well-known deep learning models, namely artificial neural networks (ANN), recurrent neural networks (RNNs), long short-term memory (LSTM), and gated recurrent units (GRUs), in predicting charging demand for EV users after a charging session begins. We use a dataset consisting of 2000 observations of charging events collected from two public charging stations in Morocco. According to numerical data results, the first layer of the GRU regression approach marginally beats the other three methods in estimating power charging needs. Specifically, the GRU regression model has an RMSE and MAPE of 1.39% and 0.50% in the training stage and 2.90% and 0.76% in the testing stage, respectively. These findings can assist the National Office of Electricity and Water in Morocco in ensuring the reliability of grid utility in the short run and guiding them to construct additional charging stations in the long run.
... The deep neural network-based EVs charging profile forecasting has also been studied in the previous literature. For instance, in [33], the LSTM model is used to predict the charge demand of EVs based on raw data. Furthermore, in [34], the CNN model is used to predict the charge required by EVs charging stations. ...
... As noted before, gated recurrent units (GRUs) are incorporated into the designed network to enhance getting to know capacity via shooting temporal capabilities. GRU is a time-efficient version of LSTM, which performs extra correctly than LSTM in 1D-time series forecasting [33]. GRU includes important gates such as the update and the reset gates. ...
Article
Full-text available
Residential electric vehicle (REV) is an advanced technology with a rapid growth rate in transportation and electric grids. One key challenge in the operation of REVs is the necessity of the accurate, reliable, and practical forecasting method to provide accurate information of the charging profile in the look-ahead hours. In power system, in order to optimize the production and consumption as much as possible, in addition to accurately predicting the amount of electricity consumption, it is necessary for the stability of the grid to take into account the imminent probabilities. This paper presents the main principle of the probability density function forecasting approach in residential electric vehicle (REV) charging profile. To this end, an end-to-end deep learning structure is designed and integrated with kernel density estimation (KDE). The designed network is composed of four major blocks, i.e., convolutional layers to extract full spatial features, gated recurrent unit (GRU) to fully understand the temporal features as a time-efficient version of the gated deep network, an autoregressive (AR) to model the long patterns including battery type, REV type, and number of REVs and kernel density estimator block. Furthermore, to improve the learning ability of the designed network, an attention mechanism is integrated into the design network. The numerical results on the actual REVs (about 348 REVs) demonstrate the effectiveness and superiority of the proposed network through several cases and comparison with several well-known deep and shallow-based methods.
... Elaborate ANN architectures such as CNNs [67] and Recurrent Neural Networks (RNNs) [46] have been explored for modelling EV load. [170] compares 12 different architectures including CNNs, RNNs. In this last article, the Long Short-Term Memory (LSTM) architecture showed the best performance on the dataset studied. ...
... Additionally, a data quality analysis was conducted on the six datasets. In [170], outliers were identified by using a set threshold from the variability between current and previous values. Instead, in this analysis, fixed boundaries were chosen and the following set thresholds were observed : -Charge and/or Park Duration has to be positive and less than 24 hours -Energy Consumption needs to be positive and less than 100 kWh The first criterion is important as some datasets have some obvious errors in the end times column which are set in 1970. ...
... Additionally, the large amount of records available is suited to test the scalability of queuing models [34,57] and spatiotemporal processes [27,36] which require travel information. It also provides an ideal setup for deep learning models which require large training sets [51,59,67,46,170]. ...
Thesis
The development of electric vehicles (EV) is a major lever towards low-carbon transport. It comes with a growing number of charging infrastructures that can be used as flexible assets for the grid. To enable this smart-charging, an effective daily forecast of the charging behavior is necessary. In this context, the objective of this thesis is threefold: (a) to identify current modeling techniques and open data available (b) to propose new EV charging methodologies to characterize their charging behaviours (c) to specify innovative techniques for daily peak load forecasting. The first chapter of the manuscript presents the industrial issues and introduces the modeling framework for EV charging. Chapter 2 is a review of state of the art EV load models as well as an exploration of 8 open charging session datasets. Chapter 3 offers a comparative study of 14 EV load and occupancy models on the 8 datasets presented in the previous chapter. Chapter 4 introduces a model for EV arrivals as a non-homogeneous Poisson process with additive spline and wavelet effects. Finally, Chapter 5 introduces a model for daily electrical peaks with a multi-resolution approach. We show that the approaches proposed in our work are competitive with the best existing alternatives by evaluating their performance on real-world data.
... There are three different gates in the LSTM cell; input, forget, and output ( Figure 1). The following equations are the LSTM cell states and parameters' updating scheme [12,20,32]: The following equations are the LSTM cell states and parameters' updating scheme [12,20,32]: ...
... There are three different gates in the LSTM cell; input, forget, and output ( Figure 1). The following equations are the LSTM cell states and parameters' updating scheme [12,20,32]: The following equations are the LSTM cell states and parameters' updating scheme [12,20,32]: ...
Article
Full-text available
Forecasting the aggregate charging load of a fleet of electric vehicles (EVs) plays an important role in the energy management of the future power system. Therefore, accurate charging load forecasting is necessary for reliable and efficient power system operation. A hybrid method that is a combination of the similar day (SD) selection, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), and deep neural networks is proposed and explored in this paper. For the SD selection, an extreme gradient boosting (XGB)-based weighted k-means method is chosen and applied to evaluate the similarity between the prediction and historical days. The CEEMDAN algorithm, which is an advanced method of empirical mode decomposition (EMD), is used to decompose original data, to acquire intrinsic mode functions (IMFs) and residuals, and to improve the noise reduction effect. Three popular deep neural networks that have been utilized for load predictions are gated recurrent units (GRUs), long short-term memory (LSTM), and bidirectional long short-term memory (BiLSTM). The developed models were assessed on a real-life charging load dataset that was collected from 1000 EVs in nine provinces in Canada from 2017 to 2019. The obtained numerical results of six predictive combination models show that the proposed hybrid SD-CEEMDAN-BiLSTM model outperformed the single and other hybrid models with the smallest forecasting mean absolute percentage error (MAPE) of 2.63% Canada-wide.
... Application of Reinforcement learning (RL) in the electro-mobility domain has attracted a lot of interest recently, leading to several published use-cases such as charging load forecasting (e.g., Zhang et al. 2021;Zhu et al. 2019), fleet assignment (e.g., Shi et al. 2020), charging station recommendation (e.g., Blum et al. 2021), and charging management (e.g., Chang et al. 2019;Wan et al. 2019;Ding et al. 2020;Dorokhova et al. 2021). Table 1 shows the summary of some exemplary studies using RL for AEV charging management. ...
Article
Full-text available
Electrification of the transportation sector provides several advantages in favor of climate protection and a shared economy. At the same time, the rapid growth of electric vehicles also demands innovative solutions to mitigate risks to the low-voltage network due to unpredictable charging patterns of electric vehicles. This article conceptualizes a stochastic reinforcement learning agent that learns the optimal policy for regulating the charging power. The optimization objective intends to reduce charging time, thus charging faster while minimizing the expected voltage violations in the distribution network. The problem is formulated as a two-stage optimization routine where the stochastic policy gradient agent predicts the boundary condition of the inner non-linear optimization problem. The results confirm the performance of the proposed architecture to control the charging power as intended. The article also provides extensive theoretical background and directions for future research in this discipline.
... Specifically, the reinforcement learning and deep learning based EV charging demand forecasting models were proposed in [38] and [39]. Furthermore, a comparative study of various deep learning approaches in the short-term forecasting of EV load was carried out in [40] and using real-world data to evaluate the performance of forecasting models. Meanwhile, using big data analysis based methods, [41] and [42] introduced online ridehailing trip data and internet-of-things based real-time data to further improve forecasting accuracy. ...
Article
Full-text available
The penetration rates of both electric vehicles (EVs) and distributed energy resources (DERs) have been increasing rapidly as appealing options to address the global problems of carbon emissions and fuel supply issues. However, uncoordinated EV charging activities and DER generation result in operational challenges for power distribution systems. Therefore, this article has developed a hierarchical transactive energy (TE) framework to locally induce and coordinate EV charging demand and DER generation in electric distribution networks. Based on a modified version of the alternating direction method of multipliers (ADMMs), two fully decentralized (DEC) peer-to-peer (P2P) trading models are presented, that is, an hour-ahead market and a 5-min-ahead real-time market. Compared to existing P2P electricity markets, this research represents the first attempt to comprehensively incorporate alternating current (ac) power network constraints into P2P electricity trading. The proposed TE framework not only contributes to mitigating operational challenges of distribution systems, but also benefits both EV owners and DER investors through secured local energy transactions. The privacy of market participants is well preserved since the bid data of each participant are not exposed to others. Comprehensive simulations based on the IEEE 33-node distribution system are conducted to demonstrate the feasibility and effectiveness of the proposed method.
... Electric vehicle charging load has strong randomness and volatility, and is affected by many factors, which increases the difficulty of charging demand forecasting [4] . At present, there are two main categories of electric vehicle load forecasting: physical model-driven [5][6][7] and data-driven [8][9][10] charging demand forecasting methods. ...
Article
Full-text available
Accurately predicting the electric vehicle charging station load is the basis for the optimal design of power supply planning. Aiming at the problem that the load data of electric vehicle charging station has strong randomness and volatility due to the interaction of various factors. The electric vehicle charging load prediction model is constructed, which considers the highest temperature, day type and weather type factors to select similar days. Firstly, an algorithm for optimizing variational modal decomposition parameters with relative entropy is proposed to decompose the extracted similar daily charging load data into multiple sub-sequences. Then, the decomposed data are respectively input into the combined prediction model of the least squares support vector machine optimized by the whale algorithm improved by the adaptive probability threshold and the random difference mutation strategy, and finally the predicted value of each subsequence is superimposed and summed to obtain the final predicted value. Based on the measured charging load data and weather forecast data of an electric vehicle charging station in Shanghai, simulation experiments are carried out to compare and verify the accuracy and effectiveness of the model in this paper.
... The findings indicated that the results of Multilayer Perceptron Training, as well as Jordan Education methods, are favorable, whereas the Radial Basis Function method has a higher error and computation burden for EV's load prediction. In [173], six different deep learning approaches have been compared from the viewpoint of EV load forecasting. The methods included ANN, RNN, canonical LSTM, gated recurrent units, stacked auto-encoders, and bi-directional LSTM forecasting methods. ...
Article
The role of electric vehicles (EVs) in energy systems will be crucial over the upcoming years due to their environmental-friendly nature and ability to mitigate/absorb excess power from renewable energy sources. Currently, a significant focus is given to EV smart charging (EVSC) solutions by researchers and industries around the globe to suitably meet the EVs' charging demand while overcoming their negative impacts on the power grid. Therefore, effective EVSC strategies and technologies are required to address such challenges. This review paper outlines the benefits and challenges of the EVSC procedure from different points of view. The role of EV aggregator in EVSC, charging methods and objectives, and required infrastructure for implementing EVSC are discussed. The study also deals with ancillary services provided by EVSC and EVs' load forecasting approaches. Moreover, the EVSC integrated energy systems, including homes, buildings, integrated energy systems, etc., are reviewed, followed by the smart green charging solutions to enhance the environmental benefit of EVs. The literature review shows the efficiency of EVSC in reducing charging costs by 30 %, grid operational costs by 10 %, and renewable curtailment by 40 %. The study gives key findings and recommendations which can be helpful for researchers and policymakers.
... Convolutional neural networks, long short-term memory (LSTM), and bidirectional LSTM (Bi-LSTM) models based on deep learning were used to forecast EV charging loads [10], [11]. Similarly, Zhu et al. in [12] obtained better EV load forecasting results with LSTM. While the implemented LSTM model provides adequate forecasting of fast charging demand, it is not validated using conventional AI methods such as Multi-Layer Perceptron (MLP). ...
Article
Full-text available
As the transition to electric mobility is accelerating, EV fleet charging loads are expected to become increasingly significant for power systems. Hence, EV fleet load forecasting is vital to maintaining the reliability and safe operation of the power system. This paper presents a new multiple decomposition based hybrid forecasting model for EV fleet charging. The proposed approach incorporates the Swarm Decomposition (SWD) into the Complete Ensemble Empirical Mode Decomposition Adaptive Noise (CEEMDAN) method. The multiple decomposition approach offers more stable, stationary, and regular features of the original signals. Each decomposed signal is fed into artificial intelligence based forecasting models including multi-layer perceptron (MLP), long short-term memory (LSTM) and bidirectional LSTM (Bi-LSTM). Real EV fleet charging data sets from the field are used to validate the performance of the models. Various statistical metrics are used to quantify the prediction performance of the proposed model through a comparative analysis of the implemented models. It is demonstrated that the multiple decomposition approach improved the model performance with an $R^2$ value increasing from 0.8564 to 0.9766 as compared to the models with single decomposition.