To confirm the effectiveness of the DDQN technique, this paper has established a HIL testing platform as proven in Fig. 9 to conduct HIL exams on the proposed vitality administration technique. The HIL testing platform is primarily composed of the car ECU controller PowerECU-57A produced by Shandong Hydrogen Exploration New Power Know-how Co., Ltd. in China, and the HIL testing cupboard manufactured by Nationwide Devices in america. The HIL testing course of is illustrated in Fig. 9: The ECU communicates with the NI real-time simulation machine utilizing analog voltage indicators for knowledge transmission. By using the corresponding Matlab/Simulink plugins, PowerECU-Toolbox for the ECU and NI-VeriStand Blocks for the NI real-time simulation machine, applicable analog I/O communication modules are added to the tractor mannequin and vitality administration technique mannequin constructed on this paper. Utilizing the goal language compiler (TLC) information akin to the ECU and NI real-time simulation machine, together with the Matlab/RTW code generator, the tractor mannequin and vitality administration technique mannequin are compiled into C code. The compiled code is then burned into the ECU and NI real-time simulation machine utilizing PowerBOOT V1.10 and NI-VeriStand 2020, respectively. Lastly, software program configuration and knowledge monitoring of the ECU and NI real-time simulator are carried out utilizing Energy-CAL V1.32 and NI-VeriStand 2020.
Beneath plowing circumstances, the specified car velocity monitoring efficiency throughout HIL testing is proven in Fig. 10. The typical error between the specified car velocity and the present car velocity stays at 0.028 km/h, indicating that the simulation mannequin can precisely change knowledge in the course of the HIL testing.
Beneath plowing circumstances, the iterative outcomes of the DDQN algorithm are proven in Fig. 11. The iterative rewards and common iterative rewards start to converge in the course of the 74th and 81st iteration coaching processes, respectively. After 100 iterations of coaching, the reward curve tends to stabilize, verifying the reliability of the algorithm’s iterative outcomes.
As proven in Fig. 12a, below the plowing situation, the driving motor energy is concentrated between 40 and 50 kW, with a median load energy of 40.90 kW and a complete energy consumption of roughly 34.09 kWh. As illustrated in Fig. 12b, below the management of the DDQN technique, the engine begins as much as cost the battery inside roughly 15 s, and thereafter, the battery energy stays under 23.44 kW. Beneath the management of the PF technique, the battery energy equals the driving motor energy a number of occasions, indicating that in the course of the PF technique management course of, the engine begins and stops a number of occasions to cost the battery.
As proven in Fig. 13a, below the management of the DDQN technique, the engine energy is concentrated between 23 and 52 kW, with a median energy of 35.62 kW and a complete energy consumption for electrical energy technology of roughly 29.70 kWh. Beneath the management of the PF technique, the engine energy is concentrated between 47 and 58 kW, with a median energy of 36.19 kW and a complete energy consumption for electrical energy technology of roughly 30.17 kWh. As illustrated in Fig. 13b, the engine working factors chosen by each the PF technique and the DDQN technique are concentrated close to the OOL curve. Nevertheless, below the management of the DDQN technique, the engine’s working vary is wider.
As proven in Fig. 14a, below the management of the PF technique, the remaining SOC is roughly 60.04%, whereas below the management of the DDQN technique, the remaining SOC is about 60.75%. The distinction in remaining SOC between the 2 methods is roughly 1.18%. Beneath the management of the DDQN technique, the pattern of the SOC curve is smoother in comparison with the PF technique, with no steady important will increase or decreases in SOC. As illustrated in Fig. 14b, the equal gas consumption of the PF technique and the DDQN technique is roughly 13.84 L and 12.40 L respectively, and the diesel consumption is roughly 8.27 L and seven.89 L respectively. In comparison with the PF technique, the DDQN technique reduces equal gas consumption by 10.40% and diesel consumption by 4.59%.
Beneath the rotary tillage situation, the specified car velocity monitoring impact of the HIL take a look at is proven in Fig. 15, the place the common error between the anticipated car velocity and the present car velocity stays at 0.030 km/h.
Beneath rotary tillage circumstances, the iteration outcomes of the DDQN algorithm are proven in Fig. 16. The iteration reward and common iteration reward started to converge on the 68th and 73rd iteration coaching processes, respectively. After 100 iteration trainings, the reward curve stabilized, verifying the reliability of the algorithm’s iteration outcomes.
As proven in Fig. 17a, below rotary tillage circumstances, the driving motor energy is concentrated between 40 and 50 kW, with a median load energy of 42.10 kW, and the overall energy consumption is roughly 35.09 kWh. As proven in Fig. 17b, below the management of the DDQN technique, the battery energy variation is much like that in plowing circumstances. Roughly 11 s after the engine begins, it costs the battery, and thereafter the battery energy stays under 29.08 kW. Equally, below the management of the PF technique, there are additionally a number of cases the place the battery energy is the same as the driving motor energy.
As proven in Fig. 18a, below the management of the DDQN technique, the engine energy is concentrated between 23 and 51 kW, with a median energy of 34.17 kW, and the overall energy consumption for electrical energy technology is roughly 28.48 kWh. Beneath the management of the PF technique, the engine energy is concentrated between 47 and 60 kW, with a median energy of 35.04 kW, and the overall energy consumption for electrical energy technology is roughly 29.21 kWh. As proven in Fig. 18b, the engine working factors chosen by each the PF technique and the DDQN technique are concentrated close to the OOL curve. Just like the outcomes obtained in plowing circumstances, the engine working vary is wider below the management of the DDQN technique.
As proven in Fig. 19a, the remaining SOC below the PF technique management is roughly 53.1%, whereas below the DDQN technique management, it’s roughly 51.46%, with a distinction of about 3.09% between the 2 methods. Just like the outcomes obtained in plowing circumstances, the pattern of the SOC curve below the DDQN technique management is gentler in comparison with the PF technique. As proven in Fig. 19b, the equal gas consumption below the PF technique and DDQN technique is roughly 13.40 L and 12.09 L respectively, with diesel consumption of roughly 7.97 L and seven.59 L respectively. In comparison with the PF technique, the DDQN technique reduces equal gas consumption by 9.78% and diesel consumption by 4.77%.
Beneath transportation circumstances, the specified car velocity monitoring efficiency throughout HIL testing is proven in Fig. 20, with the common error between the anticipated car velocity and the present car velocity maintained at 0.070 km/h.
Beneath transportation circumstances, the iterative outcomes of the DDQN algorithm are proven in Fig. 21. The iterative rewards and common iterative rewards start to converge in the course of the 66th and 72nd iterative coaching classes respectively. After 100 iterations of coaching, the reward curve tends to stabilize, verifying the reliability of the algorithm’s iterative outcomes.
As proven in Fig. 22a, below transportation circumstances, the driving motor energy is concentrated between 50 and 105 kW, with a median load energy of 65.63 kW, and the overall energy consumption is roughly 29.19 kWh. As proven in Fig. 22b, below the management of each the DDQN technique and the PF technique, the utmost charging energy of the battery is 70.92 kW and 58.90 kW respectively, and the pattern of battery energy change is mainly constant.
As proven in Fig. 23a, below the management of the DDQN technique, the engine energy is concentrated between 23 and 76 kW, with a median energy of 53.61 kW, and the overall energy consumption for electrical energy technology is roughly 23.84 kWh. Beneath the management of the PF technique, the engine energy is concentrated between 50 and 80 kW, with a median energy of 53.48 kW, and the overall energy consumption for electrical energy technology is roughly 23.78 kWh. As proven in Fig. 23b, the engine working factors chosen by each the PF technique and the DDQN technique are concentrated close to the OOL curve. In comparison with the plowing and rotary tillage circumstances, the engine working vary below each management methods has elevated.
As proven in Fig. 24a, the remaining SOC below the PF technique management is roughly 56.35%, whereas below the DDQN technique management, it’s roughly 55.51%, with a distinction of about 1.49% between the 2 methods. As proven in Fig. 24b, the equal gas consumption below the PF technique and DDQN technique is roughly 6.48 L and 5.86 L respectively, with diesel consumption of roughly 6.14 L and 6.18 L respectively. In comparison with the PF technique, the DDQN technique reduces equal gas consumption by 9.57% however will increase diesel consumption by 0.65%.
The HIL take a look at outcomes for the DDQN technique and PF technique below totally different working circumstances are proven in Fig. 25.
Beneath plowing circumstances, the DDQN technique saved 1.18% of SOC, diminished equal gas consumption by 10.40%, and decreased diesel consumption by 4.59% in comparison with the PF technique.
Beneath rotary tillage circumstances, the DDQN technique consumed 3.09% extra SOC, diminished equal gas consumption by 9.78%, and decreased diesel consumption by 4.77% in comparison with the PF technique.
Beneath transportation circumstances, the DDQN technique consumed 1.49% extra SOC, diminished equal gas consumption by 9.57%, and elevated diesel consumption by 0.65% in comparison with the PF technique.
Based mostly on the evaluation of the HIL take a look at knowledge outcomes, it may be concluded that below the three working circumstances of plowing, rotary tillage, and transportation, the DDQN technique consumes as much as 3.09% extra remaining SOC and will increase diesel consumption by a most of 0.65% in comparison with the PF technique. Due to this fact, it may be inferred that below the premise of comparable vitality consumption, the DDQN technique considerably reduces equal gas consumption in comparison with the PF technique.
