CAJ | 학술논문

根据神经网络需要训练学习的特点，探索了神经网络PID控制器的实际应用技术，在学习阶段采用固定PID参数的控制器，在训练结束后实行实时调整PID参数的策略. 将神经网络PID控制器用于油田模拟装置的实际控制，结果表明，系统的超调量减小，准确性提高，鲁棒性增强. 为神经网络在生产过程中的工业应用提供了经验.
근거신경망락수요훈련학습적특점，탐색료신경망락PID공제기적실제응용기술，재학습계단채용고정PID삼수적공제기，재훈련결속후실행실시조정PID삼수적책략. 장신경망락PID공제기용우유전모의장치적실제공제，결과표명，계통적초조량감소，준학성제고，로봉성증강. 위신경망락재생산과정중적공업응용제공료경험.
In the paper，the research was made for the real applicable technology of the neural network. Because the neural network need to be trained and learned, thus during learning, fixed parameters were used by controller. After training, the PID parameters were adjusted to real time by the neural network. The above controller was used successfully by the oil field simulating device. The system's overshoot was reduced, accuracy and robustness was improved. The application provides the experience for the neural network on the production processing in industrial applications.