岑振钊, 岳学军, 王林惠, 等. 基于神经网络PID的无人机自适应变量喷雾系统的设计与试验[J]. 华南农业大学学报, 2019, 40(4): 100-108. DOI: 10.7671/j.issn.1001-411X.201811017
    引用本文: 岑振钊, 岳学军, 王林惠, 等. 基于神经网络PID的无人机自适应变量喷雾系统的设计与试验[J]. 华南农业大学学报, 2019, 40(4): 100-108. DOI: 10.7671/j.issn.1001-411X.201811017
    CEN Zhenzhao, YUE Xuejun, WANG Linhui, et al. Design and test of self-adaptive variable spray system of UAV based on neural network PID[J]. Journal of South China Agricultural University, 2019, 40(4): 100-108. DOI: 10.7671/j.issn.1001-411X.201811017
    Citation: CEN Zhenzhao, YUE Xuejun, WANG Linhui, et al. Design and test of self-adaptive variable spray system of UAV based on neural network PID[J]. Journal of South China Agricultural University, 2019, 40(4): 100-108. DOI: 10.7671/j.issn.1001-411X.201811017

    基于神经网络PID的无人机自适应变量喷雾系统的设计与试验

    Design and test of self-adaptive variable spray system of UAV based on neural network PID

    • 摘要:
      目的  针对传统植保无人机在定量喷施作业时由于飞行速度的变化造成施药不均匀以及传统控制算法无法满足无人机变量喷雾系统所需的实时性和稳定性等问题,设计一种基于神经网络PID的自适应无人机变量喷雾系统。
      方法  采用风压变送器测出无人机的飞行速度,根据速度采用脉宽调制(PWM)方法进行自适应变量喷雾,同步用流量传感器测出实际喷雾流量,融合BP神经网络PID控制算法调节喷雾流量。由MATLAB构建BP神经网络PID控制算法,并与PID、模糊PID和神经元PID对比及分析;田间试验过程中,对比分析无人机定量喷雾与随飞行速度改变的变量喷雾效果,采用水敏纸获取雾滴沉积量分布,分别从整体区域、飞行方向和喷杆方向评价沉积量分布的均匀性。
      结果  算法仿真对比试验结果表明,与PID、模糊PID和神经元PID相比,BP神经网络PID阶跃响应上升时间分别少28.57%、84.73%和31.03%,正弦跟踪平均误差分别小63.01%、87.03%和0.58%,方波跟踪平均误差分别小74.00%、79.53%和6.80%,鲁棒性强,无静差,超调量为1.20%;喷雾对比试验结果表明,本系统能够根据飞行速度自适应调节喷雾流量,实际流量与目标流量的平均偏差为8.43%,水敏纸扫描结果表明总体区域雾滴沉积量的变异系数对比定量喷雾平均降低26.25%,喷杆方向平均降低18.79%。
      结论  该研究结果可为农业航空变量喷雾技术的应用提供理论基础。

       

      Abstract:
      Objective  The change of flight speed in constant flow spraying process of traditional plant protection UAVs causes nonuniform pesticide application, and the common control algorithms cannot meet the requirements of the real-time and stability that UAVs variable spray system needs. To solve these problems, we designed a self-adaptive variable spray system of UAV based on neural network PID.
      Method  The flight speed of UAV was measured by wind pressure transmitter. According to the flight speed, we used pulse width modulation (PWM) for self-adaptive variable spray. At the same time, we measured the actual spray flow with the flow sensor and adjusted the spray flow with PID control algorithm based on BP neural network. We used MATLAB to construct PID control algorithm with BP neural network and compared with PID, fuzzy PID and neural PID control algorithms. In the field experiment, we compared and analyzed the effects of constant spray and variable spray based on changing flight speed. We used water-sensitive paper to obtain the distribution of droplet deposition, and then evaluated the uniformity of deposition distribution from the whole area, flight direction, and spray bar direction respectively.
      Result  The simulation result of algorithms indicated that comparing with PID, fuzzy PID and neuron PID, the rise time of step response for BP neural network PID was 28.57%, 84.73% and 31.03% shorter respectively, the average error of sinusoidal tracking was 63.01%, 87.03% and 0.58% lower respectively, the average error of square wave tracking was 74.00%, 79.53% and 6.80% lower respectively. Additionally, the BP neural network PID had strong robustness, 0 static error, and lower overshoot (1.20%). The comparison of the spray tests showed that this system can automatically adjust the spray flow according to flight speed. The average deviation between the actual flow and the set flow was 8.43%. Based on the testing results of the water sensitive paper, variable spray decreased the coefficient of variation of droplet deposition in the whole area by 26.25% and in the spray bar direction by 18.79% on average compared with constant spray.
      Conclusion  The research results can provide a basis for the application of variable spray technology in precision agricultural aviation.

       

    /

    返回文章
    返回