Artificial Neural Networks Based Power Management for a Battery/Supercapacitor and Integrated Photovoltaic Hybrid Storage System for Electric Vehicles

The exponential growth of energy demand around the world has led to a massive increase in the emission of greenhouse gases, thereby sparking the interest of the global industrial and academic sectors in finding alternative energy sources [1]. The ecological and economical motivation for improving the transportation sector has led to the adoption of renewable energy-based electric vehicles, which are projected to lead the automobile market in the upcoming years [2, 3].

The emission of greenhouse gases is a global issue with harmful consequences for the environment and human health. Introducing electric vehicles is a potential solution as they emit significantly less than traditional gasoline vehicles. This would reduce reliance on fossil fuels and greenhouse gas emissions while also creating economic benefits through job creation and sustainable energy sources. Overall, this is a crucial step towards mitigating climate change and promoting sustainable development.

The recent integration of solar energy in electric vehicles can be viewed as the next step in the latter's move away from its fossil fuel dependency [4, 5]. This integration is well suited for regions near the equator [6] as well as Mediterranean regions with an abundance of solar radiation, such as Algeria [7, 8]. The radiation can be collected utilizing photovoltaic panels mounted on top of the car or by wrapping the entire vehicle in a photovoltaic coating [9]. The harvested energy can power the auxiliary systems such as lights, air conditioning, etc. and it can even limit the critical discharge of the storage systems [10]. The concept of electric vehicles with integrated photovoltaic panels has been adopted by various car manufacturers ranging from well-established brands, namely Tesla [11], Audi, Mercedes Benz, BWM, Jaguar, and Honda [12], to the relatively small start-ups, including Stella Lux [13], Hanergy [14], Fisker [15, 16], Sono Motors [17] and Light year [18]. The Sion car introduced by the German start-up Sono Motors is the closest existing vehicle to our proposed EV model. It incorporates a total of 248 long-lasting, scratch-resistant, and efficient solar cells, that are seamlessly planted into the body of the car generating a peak power of 1.2 kW [17]. Although the Sono Sion is a promising evolution for solar electric vehicles, it still suffers from significant issues that might hinder its performance in real world applications, mainly its short driving range, which is primarily due to the use of the battery alone as an energy storage device.

The aforementioned energy storage system that is based on a battery only can have many drawbacks such as the limited life-cycle, the limited power density as well as high cost [19]. Unlike batteries, supercapacitors offer many advantages namely high power density, long life cycle along with a wide operating range. These complementary features can be combined in a hybrid storage system, which will eventually increase the driving range by supporting the battery during high power demands, as well as prolonging its lifetime [20].

The use of a hybrid energy storage system requires a robust energy management technique to provide proper energy flow in all of the electric vehicle components [21], which is accomplished using an artificial neural networks strategy. This strategy is one of the suitable methods to deal with the internal relations of a complex model because of its high nonlinearity, large amounts of data-parallel processing, and high robustness [22].

Several scholars have undertaken in-depth study on power management solutions during the past few decades [9-11]. Several academics have categorized and described the various approaches utilized by some researchers to optimize the gas emissions and fuel consumption for a hybrid electric vehicle [11-13]. Online power management strategies and offline power management strategies are the two broad categories into which these techniques can be divided. In order to determine the best answer, offline power management strategies need prior knowledge of the global information. As a result, real-time implementation of these methods is not possible.

Power management solutions have both advantages and shortcomings. They can improve energy efficiency, reduce costs, increase reliability, and support renewable energy integration. However, high implementation costs, specialized expertise, and potential system failures are disadvantages. There is room for improvement in the field, including more efficient and cost-effective technologies, better standards and regulations, and continued research and development. Despite challenges, power management solutions offer benefits for improving energy efficiency and sustainability.

One of the best techniques in this field for analyzing problems involving sequential decision-making is based artificial neural network (ANN) [14-16]. ANN typically divides a global optimization problem into smaller sub problems, then searches through various control inputs from the end state and looks at the control sets to find the minimal cost final solution, which is the best one. Unfortunately, ANN cannot be used in real-time on a vehicle because it requires prior knowledge of the full drive cycle and is very computationally costly. As a result, the findings of the ANN simulation will be employed in this work as an ideal reference control method for implementation.

This research developed an intelligent power management system for electric vehicles that balances energy flow between the battery, supercapacitor, and photovoltaic panel. The system predicts energy demand and optimizes power flow in real-time, improving energy efficiency and performance while reducing dependence on non-renewable energy sources. This reduces the vehicle's carbon footprint and contributes to a sustainable transportation system. The research also has potential applications beyond electric vehicles in smart grid, renewable energy, and energy management systems. Overall, this study offers an efficient and intelligent power management solution for electric vehicles with broader implications for energy systems [17].

This study has several motivations for researching artificial neural networks based power management in electric vehicles. Challenges in managing power in electric vehicles could be a motivation, and artificial neural networks may provide a solution. Environmental and economic benefits may also be motivating factors, as improved power management could make electric vehicles a more sustainable alternative to gasoline vehicles.

The structure of the paper is as follows: The case study is briefly introduced, followed by a depiction of the studied system in section 2. The physical modeling of the various components of the proposed electric vehicle's propulsion chain, including the motor equations, is detailed in section 3. Section 4 outlines the artificial neural networks-based power management, and section 5 presents the simulation results and subsequent discussion. Finally, section 6 summarizes the principal findings of the research and its potential future developments.

3.1 Modeling of the battery

The modeling of batteries is an important aspect of battery engineering and design. There are various approaches to modeling batteries, including analytical models, equivalent circuit models, and electrochemical models.

Analytical models are based on mathematical equations that describe the behavior of the battery. These models are relatively simple and can provide useful insights into the behavior of the battery under certain conditions. However, they do not provide a detailed understanding of the underlying electrochemical processes.

Equivalent circuit models are more complex than analytical models and are based on a network of resistors, capacitors, and current sources that represent the various components of the battery. These models can provide a good approximation of the battery's behavior under different conditions and are often used in battery management systems.

Electrochemical models are the most detailed models of batteries and are based on the fundamental electrochemical processes that occur within the battery. These models can provide a detailed understanding of the battery's behavior under different conditions and can be used to optimize battery design and performance.

The choice of battery modeling approach depends on the specific application and the level of detail required. Analytical models are useful for simple applications, while electrochemical models are required for more complex applications that require a detailed understanding of the battery's behavior.

The simplified electrical model of the battery is illustrated in Figure 2 [27, 28].

2.png

Figure 2. Battery model

where,

${{V}_{Bat}}={{E}_{Bat}}+{{R}_{Bat}}{{I}_{Bat}}$              (1)

The battery capacity C_Bat can be calculated as follows [29]:

${{C}_{Bat}}={{C}_{0}}\frac{1.67(1+0.005\ \Delta T)}{1+0.67\left( \frac{{{I}_{Bat}}}{{{I}_{0}}} \right)}$                  (2)

The state of charge of the battery is given as:

$SO{{C}_{Bat}}=\frac{Q}{{{Q}_{n}}}$             (3)

where,

$Q={{I}_{Bat}}\times t$             (4)

For the battery control, the latter is connected to the DC bus via a bi-directional buck-boost DC-DC converter, as shown in Figure 3.

3.png

Figure 3. Battery control using a conventional PI controller

3.2 Modeling of the supercapacitor

Supercapacitor is an electrical energy storage device, which is also referred to as an ultracapacitor or electrochemical capacitor, stores electric energy in an electric field. Modeling of a supercapacitor involves developing a mathematical model that represents the electrical behavior of the device.

To manage the remaining battery voltage and regulate battery charging, a battery control system is implemented. The exchange of power between the battery and the DC bus is possible through a bidirectional buck-boost converter that can be controlled.

The process of modeling ultracapacitors (UCs) involves creating a series RC circuit that comprises an equivalent capacitance (C_s) and a series resistor (R_s). The series resistor represents the energy loss that occurs during the charging and discharging of the capacitor. To factor in the impact of cell-balancing intervention and energy loss that results from capacitor self-discharge, an equivalent resistance (R_p) is connected in parallel to the series RC circuit. This model is illustrated in Figure 4 [30].

4.png

Figure 4. Supercapacitor model

The electrical behavior of a supercapacitor can be described by a set of differential equations that govern the charge and discharge of the device. These equations relate the voltage, current, and capacitance of the supercapacitor, and take into account the internal resistance and capacitance of the device, as well as any external loads connected to the device.

Generally, the modeling of a supercapacitor is important for predicting its electrical and thermal behavior, and for optimizing its design and performance. By using these models, we can better understand the behavior of supercapacitors under different operating conditions and design them for specific applications, such as in energy storage systems for electric vehicles, renewable energy systems, and other high-power applications.

5.png

Figure 5. Supercapacitor electrical model with DC-DC converter

The supercapacitor, also referred to as an ultra-capacitor; is characterized by low series resistance, significant equivalent capacitance and a large number of charge/discharge cycles permitting a prolonged service duration. A capacitor stores energy via a static charge, as contrasted with an electrochemical reaction [30]. Figure 5 shows the equivalent average electrical model, a DC-DC buck-boost converter and a choke filter are associated with the supercapacitor tank as a means to adapt the voltage levels between the DC-DC converter and supercapacitor [31].

The power management strategy is formulated by assigning specific roles to each component of the hybrid power source. The supercapacitor is tasked with fulfilling the peak power demands, while the battery helps in meeting the transient power demand by regulating its state-of-energy. The photovoltaic panel system, with its slow dynamics, controls the state-of-charge of the battery.

${{C}_{SC}}\frac{d{{V}_{SC}}}{dt}=-{{I}_{SC}}$           (5)

The choke filter model is written as follows:

${{L}_{SC}}\frac{d{{I}_{SC}}}{dt}={{V}_{SC}}-{{u}_{m}}$             (6)

One possible way of describing the average model is as follows:

${{u}_{m}}={{m}_{SC}}{{V}_{dc}}$             (7)

${{i}_{m}}={{m}_{SC}}{{I}_{SC}}$           (8)

The supercapacitor tank can be modelled as follows [31].

3.3 The photovoltaic model

The PV model generally consists of several crucial parameters, such as the short-circuit current (I_sc), open-circuit voltage (V_oc), maximum power point voltage (V_mp), maximum power point current (I_mp), and fill factor (FF). These parameters can be used to calculate the efficiency of the PV device, as well as its performance under different operating conditions [32].

The PV model is important for designing and optimizing PV systems for different applications, such as grid-tied or off-grid systems, as well as for predicting the energy yield of a PV system in a given location and climate. It can also be used to evaluate the performance of different PV technologies and to identify areas for further research and development.

There are several different PV models that can be used, ranging from simple empirical models to more complex physical models that take into account the detailed physics of the solar cell. The choice of model depends on the level of accuracy required and the available data and resources.

The modeling of solar photovoltaic has an extreme influence on optimal power delivery. A proper photovoltaic model ought to be highly sensitive to the change in irradiation and temperature levels, in an effort to replicate the exact I-V characteristics in a time-varying scenario.

As depicted in Figure 6, the photovoltaic cell comprises a diode and a current source that are connected in anti-parallel with a series resistance.

6.png

Figure 6. Equivalent circuit of a solar cell with series and shunt resistance

The relationship between current and voltage in a single-diode cell can be expressed as follows [31, 32]:

${{I}_{PV}}={{I}_{ph}}-{{I}_{0}}\left( \exp \left( \frac{q\left( {{V}_{PV}}+{{R}_{s\bmod }}{{I}_{PV}} \right)}{AKT} \right)-1 \right)$           (9)

7.png

Figure 7. Current-voltage and power-voltage characteristics of the photovoltaic panel

From Figure 7, showing the photovoltaic characteristics; it can be noticed that these characteristics are not linear, thus, the presence of a controller is mandatory, in order to extract the maximum power obtained from the panel, leading to the optimal panel performance at different irradiation and temperature levels [33].

The implementation of the maximum power point tracking (MPPT) controller with a DC-DC boost converter is shown in Figure 8. The controlled boost converter serves as an interface between photovoltaic (PV) panels and the loads they are connected to. It transforms any input voltage within its operating range into a stable output voltage that is suitable for powering the load.

The proposed smart controller adapts the duty cycle of the boost converter based on input voltage and loading conditions such that it outputs a constant output voltage. A prototype system has been developed to verify the applicability and the integration of the proposed controller into automotive field.

8.png

Figure 8. MPPT Control with a boost converter

12.png

Efficient energy management strives to maximize the efficiency of powertrain components while preserving a sufficient quantity of energy in storage devices. Furthermore, energy consumption must be managed without compromising vehicle performance, which can only be accomplished by implementing an effective control method [42].

The control algorithm focuses on operating each vehicle component with optimal energy, as well as recovering as much potential energy as possible during braking while keeping the state of charge of the energy storage systems within a predetermined range. The suggested energy management flowchart, shown in Figure 12, explains how the driving modes are selected based on the previously mentioned driving modes.

The work reported in this paper describes a hybrid energy system for integrated photovoltaic vehicles that incorporates a battery as the primary storage device and a supercapacitor as a backup. This configuration improves the solar vehicle by reducing peak current effects on the batteries. Thus, extending battery life, and limiting DC bus voltage changes. A power management strategy based on neural networks was put forward to maintain DC bus voltage stability regardless of motor speed or solar irradiation and to enhance the driving range. The effectiveness of the proposed energy management strategy was evaluated by means of MATLAB simulation. The results demonstrate that combining a supercapacitor with a battery in solar cars can extend the limited driving range while also increasing the speed.

However, the effectiveness of ANNs for power management of an EV depends on various factors, such as the quality of input data, the architecture of the neural network, and the training method used. Additionally, the use of ANNs requires significant computational resources, which may be a limitation for onboard implementation in real-time.

While ANNs have shown promising results for power management of EVs, further research is needed to fully understand their potential and limitations in this application. Nonetheless, the use of ANNs for power management of EVs is a promising avenue that could significantly improve the efficiency and range of these vehicles.

In conclusion, the limitations and future research directions of power management using artificial neural networks with multiple energy sources should be considered. Limitations may include data availability, methodology, and ANN architecture. Future research could include the incorporation of other energy storage devices, emerging technologies, advanced ANN architectures, data processing techniques, and optimization algorithms. Additionally, exploring the impact of different driving conditions, traffic scenarios, and climate conditions on the power management system's performance could be beneficial. Addressing these limitations and exploring future research directions can contribute to the development of more efficient and sustainable power management systems for electric vehicles.