Movement of a Solar Electric Vehicle Controlled by ANN-based DTC in Hot Climate Regions

Vehicle autonomy presents the most complex problem for modern commercialized solar electric vehicle (SEV) propulsion systems. The power supplied by electric vehicles’ batteries is limited by the state of charge, the type of battery, and its level of technological development. This study’s aim was to resolve the problem of energy variation at several velocities and under different road topology conditions. Several works related to the use of fuzzy logic confirm that classical regulators have such advantages over fuzzy regulators as short processing times and mathematical precision. Therefore, the hybrid power source is presented as the best solution for energy management, and it is composed of a solar panel (PV) and a nickel metal hydride battery. The PV system is connected to the SEV via a boost converter that is controlled using the maximum power point tracking technique. In this paper, we used an intelligent PI regulator for direct torque control, which introduced a certain degree of intelligence into the regulation strategy. Indeed, this approach of associating the PI regulator with the fuzzy rules-composed supervisor allowed us to take advantage of both the PI’s mathematical precision and the adaptability, flexibility, and simplicity of fuzzy linguistic formalism. Because of its dynamic capabilities, an adaptive PI regulator was substituted to achieve high speeds and a satisfactorily vigorous performance while quickly compensating for the disturbances that were expected to possibly take place on the regulation chain. The present study’s results confirm that the proposed control approach increased the utility of SEV autonomy under several speed variations. Moreover, the industry’s future offerings must take the option of hybrid power management into consideration during this type of vehicle’s manufacturing phase.


Introduction
Due to its effect on every aspect of our lives, from heating, air conditioning, food preparation, transportation, and power production, energy is one of humankind's most vital necessities. Both traditional and renewable sources of energy themselves cannot be utilized to do research, as they are converted into heat, mechanical, or electrical energy (Ghosh et al. 2009).
Photovoltaic (PV) energy is, as a matter of fact, the cleanest technology used in the world. Its main disadvantage, however, is its poor efficiency. A variety of solutions to this have been proposed, including hybrid systems (Ammari et al. 2017), optimization (Kumar et al. 2016), and control algorithms (Javed et al. 2018), which were developed to optimize the performance of the PV generator.
Algeria has one of the world's highest solar capacities with 2,000 h per year across the entire country (but over 3,900 h in the highlands and the Sahara). The global horizontal irradiation is 3,000 Wh/m 2 annually in the north and more than 5,000 Wh/m 2 annually in the south (i.e., the Sahara) ).
* Corresponding author: asma.amal70@yahoo.fr Because they guarantee energy supply stability and increase device reliability, energy storage systems are important for power system operations, and they come in widely varying shapes and sizes. Nickel metal hydride (NiMH) batteries have dominated in automotive applications since the 1990s due to their overall performance and their offering of the best available combination of power energy density, thermal performance, lifecycle, and low maintenance cost (Mebarki et al. 2017).
The artificial neural network-based direct torque control (ANN-DTC) strategy is one kind of highperformance driving technology used for AC motors due to its simple structure, and its ability to achieve a fast response of flux and torque has attracted growing interest in recent years. Using PI-controlled ANN-DTC without a hysteresis band can effectively reduce torque and flux ripple, enhance the system's robustness, and improve its dynamical performance. A DC-DC converter is used with a control strategy to ensure that a solar electric vehicle's (SEV) energy requirements are met. This paper presents an SEV using a hybrid PV-battery energy storage power system, which required that it use ANN-based direct torque under speed variation. In this study, the nickel metal hybrid batteries would charge using a PV system; when the battery's state of charge fell below the reference state of charge, the PV would distribute its power between the battery and the SEV. The battery would then discharge, feeding the proposed propulsion system when the battery's state of charge became higher than the reference state of charge (Jia-Min et al. 2014;Gasbaoui et al. 2012). This paper deals with the behavior of the SEV's movement in hot weather and at different speeds. The SEV was equipped with two 3.3 HP induction motors, and an ANN-DTC was proposed to minimize current ripple, flux, and electromagnetic torque in both. To evaluate the performance of the proposed system, a simulation was performed on all the system's operating conditions. This paper is organized as follows. Section 2 presents the SEV model. Section 3 deals with the specifics of the ANN-DTC. Section 4 contains a description of the system and the control. Section 5 presents, and then discusses, the results of the simulation. Finally, a conclusion is reached.

Solar Electric Vehicle Model
The resistive forces that apply to the solar car's movement are ( Fig. 1): • the rolling resistance force "Ftire" due to the friction of the tires on the road; • the aerodynamic drag force "Faero" caused by the friction of the body moving through the air; and • the climbing force "Fslope" that is needed on a sloped road.
• Fig. 1. The forces acting on a vehicle moving along a slope.
The globally resistive force "Fr" is the sum of all the previously mentioned resistive forces, as shown in Equation 1 (Yang and Lo 2008;Gao and He 2008): (1) The rolling resistance force is described as: (2) The aerodynamic resistance force is described as: ( 3) Where "m" is the vehicle's total mass, "f" is the rolling resistance force constant, "g" is the acceleration due to gravity, "ρair" is air density, "Cd" is the aerodynamic drag coefficient, "Af" is the frontal surface area of the vehicle, "v" is the vehicle's velocity, and "β" is the road slope angle. The values of these parameters are shown in Table 1. The weight parameter of the SEV can be obtained adding the weight of the specific components (PV panels, battery pack, electric motor, and inverter). The solar vehicle considered in this work has separate induction motors.

Artificial Neural Network-Based Direct Torque Controller
The basic DTC strategy is based on the determination of instantaneous space vectors in each sampling period regarding desired flux and torque references. The reference speed is compared to the measured one, and the error thus obtained is applied to the speed regulator PI, whose output yields the reference torque ).
An artificial neural network (ANN) is based on learning by decreasing errors during a training period; once training has been completed, the objective functions should be optimized (Devi et al. 2016;Siva et al. 2017;Chill and Lin 2006;Sasikumar and Harinandan 2014).
The induction motor's stator flux can be estimated using the following equations: The electromagnetic torque Tem can be described as follows: (8) The neural network used in this study was based on three inputs: flux, torque, and sector. The network employed was a 3-5-3 feedforward network with a log sigmoid transfer function at its first layer, a hyperbolic tangent sigmoid transfer function at its second, and a linear transfer function at its third. All three neural networks were trained perform to at 0.00001 msec. A Levenberg-Marquardt backpropagation algorithm was used to train the neural networks, updating weights and bias values according to the Levenberg-Marquardt optimization (Fig.  2).
The neural network inputs resulted in an error between the estimated flux value and its reference value; the difference between the estimated electromagnetic torque, the torque reference, and the position of the flux stator vector was represented by the number of the corresponding sector. The neural network output layer was composed of three neurons.

System and Control Description
The SEVs included PV panels to generate electricity and feed the NiMH batteries and induction motors. Batteries were included in case the PV panels failed to supply the induction motors. A bidirectional converter was used to transfer and regulate the energy traveling from the generators (i.e., the PV panels and the batteries) to the induction motors. The two induction motors were coupled with the left and right rear wheels.
Power generated from the PV array was transmitted to the motor and the battery via a DC-DC boost converter and controlled using incremental conductance maximum power point tracking (MPPT) algorithm. (Gasbaoui et al. 2013). A bidirectional DC-DC converter was used to mitigate power disturbances between the battery and the PV panel. The two rear induction motors were controlled by an ANN-DTC using an adaptive fuzzy speed controller (Allaoua et al. 2013;Zhang et al. 2003). The electronic differential system ensured robust vehicle control on different road topologies. The basic principles of energy management employed in this study are shown in Fig. 3. A block diagram of the ANN-DTC and the differential electronics (Moreno et al. 2006) is shown in Fig. 4.

Nickel Metal Hydride Battery power responses.
To characterize the SEV's behavior, simulations were performed using a model (Fig.4) in a MATLAB/SIMULINK environment. The simulation was divided into two phases. The first phase comprised a performance test of the SEV controlled by an ANN-DTC, and the second phase tested the performance of the NiMH batteries and PV panels under different conditions (Tahiri et al. 2019).
The topology test was subdivided into three phases (Fig. 5). In the first, the SEV traveled along a straight road at 20 km/h. The solar irradiance was 1,000 W/m 2 , and the temperature was 25°C. The second phase saw the SEV accelerate to 40 km/h, and the temperature was increased steadily to 50°C, while the solar irradiance was ramped down from 1,000-190 W/m 2 , with MPPT continuously tracking the maximum power and solar irradiance. The final phase saw the SEV traveling in the hottest region. The SEV decelerated to 30 km/h, the solar irradiance was restored to 1,000 W/m 2 , and the temperature was increased to 50°C. The road topologies at each phase are illustrated in Fig. 5, and the speed constraints at each phase are shown in Table 2. As described above, the SEV traveled along a straight road during the first phase at 20 km/h under a solar irradiance of 1,000 W/m 2 and a temperature of 25°C (Fig. 6). The second phase saw an increase in temperature, but the SEV's linear speed was not affected, demonstrating the same robustness of both the control and the driving wheels. Finally, as the SEV traveled in the hottest region, the temperature and solar irradiance reached 50°C and 1,000 W/m 2 , respectively, and the vehicles decelerated to 30 km/h (Fig. 7). Figure 8 and Table 3 note that the effect of the disturbance was ignored by the adaptive fuzzy PI controllers, affording better performance in both overshoot control and tracking error.     Figure 9 reflects the relationship between the vehicle's change in speed and the distance traveled in the different phases; for this, a new fitting formula was developed: The SEV required a current of 12.40 A in the first phase for the tow induction motors, yielding a driving force of 74.30 N.
In the second phases, the effect of acceleration with 40 km/h on the directly road has been obtained. The wheels' driving force increased, and the current demand doubled from the current braking phase, causing the PV panel and NiMH battery to exploit 80% of their capacity to achieve the power necessary to negotiate the road's condition, which can be interpreted physically by the increase in the vehicle's overall resistive torque (Table 4). However, the linear speeds of the two induction motors remained similar, and the lowering of the road's gradient did not influence the torque controller. In the third phase, the demands on both the current and driving force increased, indicating that the vehicle entered a recharging phase, which can be explained by the subsequent decrease in the current demand and driving force (Figs. 10 and 11). Finally, the current increased to 31.32 A, and the driving force increased to 83.20 N.   In the first phase, the vehicle's resistive torque was 20.54 Nm, and power propulsion system resisting more than 66% from nominal motor torque. During the breaking phase (Phase 2), the rear wheels built up to fit the traction chain demand, which induced a resistive torque of 27.50 Nm. Our results confirm that the traction chain under acceleration exerted more force than in the first phase; the vehicle, while decelerating, required only half the energy needed for acceleration. Table 5 summarizes the abovementioned results. The battery was sufficient to supply the SEV with power in the beginning, accelerating, and decelerating phases, indicating that the peak power of the NiMH batteries must be greater than or equal to the peak power of the electric motors. Table 6 lists this battery's parameters.
The battery's power production depended solely on the electronic differential assigned to the state of the acceleration/deceleration driver, which can be illustrated by the current (see Fig. 12).
The state of the NiMH battery's charge changed during the driving cycle; it seems that the state of charge (SOC) decreased rapidly at the beginning, and it ranged between 67.93% and 70% throughout every phase, from beginning to end (Table 7).
Battery power was controlled by the bidirectional converter, which accurately distributed power during every phase. The buck boost converter was not only a robust converter that guaranteed the transmission of voltage, but it also demonstrated excellent battery recharge characteristics while decelerating. These advantages were noted when the speed increased from 20 km/h to 40 km/h and the demanded voltage remained stable at 450 V (Fig. 13).

Solar Photovoltaic Panel
The solar PV field was composed of six parallel of three series of connected cells. The PV field's total power was 8.01 kW. Table 8 lists the PV panel's parameters.
The voltage of photovoltaic panels took on a high value during the first phase due to the low speed (v = 20 km/h), decreasing slowly in Phase 2 when the SEV accelerated to 40 km/h and stabilizing at 140 V ( Fig. 14  [a]). In first phase, the SEV moved slowly (20 km/h), consuming minimal current (10 A) and power (2 kW). As the vehicle's speed increased, (40 km/h), current consumption peaked at 25 A before the regulator stabilized consumption to 7 A. When the speed decreased, the current's maximum value was recorded at 30 A ( Fig.  14 [b]). Fig. 14 (c) shows that the first phase (between t = 0.5s and 1.5 s), characterized by a 20-km/h speed and 1,000-W/m 2 irradiation, caused a decrease in power due to the induction motor's consumption outpacing the PV panel's production. Between t = 0.15 s and 2.5 s, power decreased due to low solar irradiance (190 W/m 2 ). However, by t = 2.5 s, the power began rising until t = 3 s due to increasing irradiance (1,000 W/m 2 ), stabilizing from between t = 3 s and 4.5 s. The duty cycles increased during times of high consumption (i.e., Phase 2) and decreased when the vehicle moved more slowly, such as in Phase 3 (Fig. 14 [d]). Table 9 presents the variation of temperature, solar radiation, PV voltage, and PV power output during the different phases. The PV panels generated 7.00 kW when the vehicle was moving in a hot region (50°C) from t = 3-4.5 s at 30 km/h under 1,000 W/m 2 irradiance. The maximum power recorded in the first phase was 8.00 kW.       Figure 15 summarizes the results in two parts. The first part is the discharge mode, when the power of the battery took a positive value and covered vehicle consumption due to low solar PV power output; this began at t = 0-2 s. At t = 2.5 s, the solar PV began to feed the vehicle as the battery power took a negative value as it began to charge.

Conclusions
The vehicle energy policy outlined in this paper has demonstrated that hybrid power sources based on a solar PV panel and a NiMH battery controlled by a buck boost DC-DC converter, used in an SEV that utilized two rear driving wheels for motion, can be improved using an artificial neural network-based direct torque control strategy when the developed power depended on the driver's reference speed.
The road topologies and variations in temperature and solar radiation did not affect the performance of the proposed system. The boost DC-DC converter, bidirectional bridge converter output voltage, and the control strategy demonstrated satisfactorily dynamic characteristics vis-à-vis the SEV's propulsion system.  The proposed adaptive fuzzy PI speed controller model improved the wheels' speed control with high accuracy on a flat road. This hybrid power source increased the SEV's autonomy. This study enables the prediction of a hybrid power source's dynamic behavior under different speed variation conditions, which is considered as a foundation for control and power management for two and four rearwheel drive SEVs.