现存的垃圾焚烧炉在运行过程中存在较大的不稳定性，关键运行数据呈现出滞后性、强耦合及多扰动的特点，所以垃圾焚烧炉运行参数需要实时动态调节，以保障垃圾焚烧炉的稳定高效运行。本研究通过采集垃圾焚烧炉的历史运行数据，采用深度学习技术（多层LSTM神经网络），利用离线训练建立垃圾焚烧炉运行参数与热效率之间的高度非线性耦合数学模型；以垃圾焚烧炉热效率最大为目标，利用强化学习技术对垃圾焚烧炉的运行参数进行优化，以达到降低垃圾焚烧运行成本并消除人为影响因素的目的，并给出了具体的应用实例。经过优化后，每吨垃圾最大产汽量由原来的2.22吨上升到了2.5吨，热效率提高了11.1%，优化效果明显。该优化方法对于提高垃圾焚烧炉运行效率有重要的参考价值，值得进一步推广应用。

The existing waste incinerators have significant instability during operation, with key operating data showing lag, strong coupling, and multiple disturbances. Therefore, the operating parameters of waste incinerators need to be dynamically adjusted in real time to ensure stable and efficient operation. This study collects historical operating data of waste incinerator and uses deep learning technology (multi-layer LSTM neural network) to establish a highly nonlinear coupled mathematical model between the operating parameters and thermal efficiency of the waste incinerator through offline training; Then, with the goal of maximizing the thermal efficiency of the garbage incinerator, reinforcement learning technology was used to optimize the operating parameters of the garbage incinerator, in order to reduce the cost of waste incineration power generation and eliminate human factors. Specific application examples were provided.After optimization, the maximum steam production per ton of waste has increased from 2.22 tons to 2.5 tons, and the thermal efficiency has increased by 11.1%. The optimization effect is significant. This optimization method has important reference value for improving the operational efficiency of waste incinerators and is worthy of further promotion and application.

TK16