农机化研究

针对丘陵山地榨菜直播作业中因地形复杂、土壤条件多变而导致的播种精度不足、施肥不均匀、喷水量不稳定等问题，创新性地设计了一种基于BP神经网络PID控制的榨菜直播机自适应控制系统，通过多传感器实时采集作业地形坡度、土壤湿度、播种机前进速度等参数，构建播种、施肥、喷水、开沟和行走控制5大执行环节的协同控制模型。采用“前馈神经网络预测+反馈PID调节”的混合控制策略，利用BP神经网络强大的非线性映射能力实现PID控制器参数的在线自整定与多目标优化；引入了基于地形识别的动态优先级调度算法，可根据坡度变化实时调整各子系统控制指令的优先级分配，有效解决陡坡工况下的动力协调与执行冲突问题，确保直播机的作业稳定性与安全性。田间试验结果表明：在坡度不超过25°的山地条件下，播种合格率最高达93.1%,综合平均值为91.5%,施肥均匀性变异系数≤6.8%,喷水量误差控制在±7.2%以内，榨菜直播机作业效率较传统直播机提升26.3%。所设计的BP-PID自适应控制系统能够有效解决丘陵山地复杂环境下的非线性、时变性的控制难题，显著提升榨菜直播机的作业精度、地形适应性与综合效能。

To address issues such as insufficient sowing precision, uneven fertilisation, and unstable water application arising from complex terrain and variable soil conditions during direct seeding operations for mountainous pickled mustard greens, an adaptive control system of mustard direct seeding machine based on BP neural network PID control has been innovatively designed. The multiple sensors was employed to collect real-time parameters including terrain gradient, soil moisture, and machine forward speed, constructing a collaborative model for five key operational segments: sowing, ferti-lisation, water spraying, ditching and walking control. The hybrid control strategy of ' feedforward neural network prediction + feedback PID regulation ' is adopted, and the online self-tuning and multi-objective optimization of PID controller parameters were realized by using the powerful nonlinear mapping ability of BP neural network. A dynamic priority sche-duling algorithm based on terrain recognition was introduced, which could adjust the priority allocation of control instructions of each subsystem in real time according to the change of slope, effectively solve the problem of dynamic coordination and execution conflict under steep slope conditions, and ensure the operation stability and safety of the live broadcaster. The results of field experiment showed that under the condition of mountain slope not exceeding 25 °, the sowing qualified rate was up to 93.1%, the comprehensive average value was 91.5%, the coefficient of variation of fertilization uniformity was less than 6.8%, the error of water spraying was controlled within ± 7.2%, and the operation efficiency of mustard direct seeding machine was 26.3% higher than that of traditional direct seeding machine. The designed BP-PID adaptive control system could effectively solve the nonlinear and time-varying control problems in the complex environment of hilly and mountainous areas, and significantly improve the operation accuracy, terrain adaptability and comprehensive efficiency of the mustard direct seeding machine.