基于RF-SEGA组合算法的热分散系统运行优化模型

Operation Optimization Model of the Heat Dispersion System Based on RF-SEGA Hybrid Algorithm

中文关键词:  热分散系统  工艺优化  机器学习

Key Words:heat dispersion system  process optimization  machine learning

中文摘要:

      热分散系统是制浆造纸过程中废纸回用的重要单元，热分散处理效果对纸机运行性能和纸张产品质量有着重要影响。由于不同废纸的性能差异较大，纸浆中的胶黏物含量和尺寸大小、纤维形态等关键指标无法量化，热分散系统难以实现最优化运行，存在能源消耗较大、化学品浪费和纸浆性能波动等问题。本研究建立了一种热分散系统运行优化模型，以纸张产品及其定量为分类依据，建立优化模式数据库，采用随机森林（RF）算法和增强精英保留的遗传算法（SEGA）建立能效优化模型。实际生产数据验证结果表明，经模型优化后，热分散系统运行参数的优化幅度分别为：扭矩0~3%、温度0~4 ℃、热分散机功率0~50 kW。优化结果符合工艺调整幅度，表明优化模型具备稳定性。依据模型优化值调参4次，胶黏物累积量相比于调参前分别降低了约6.36%、5.17%、4.25%、7.82%；每吨绝干浆的综合成本分别降低了约4.62%、3.55%、4.43%、4.36%。研究结果表明，该热分散系统运行优化模型符合实际生产工况，可改善纸浆分散效果，降低热分散能耗成本。

Abstract:

      The thermal dispersion system is an important process unit for the reuse of waste paper in the pulping and papermaking process， and the effect of thermal dispersion treatment has a significant impact on the operating performance of the paper machine and the quality of paper products. Due to the performance variation of different waste papers， key indicators such as the content and size of adhesive， and fiber morphology in pulp cannot be quantified. The thermal dispersion system is difficult to achieve optimal operation， and there are some problems such as large energy consumption， chemical waste， and pulp performance fluctuations. This study established an optimization model for the operation of a thermal dispersion system， based on the types and basis weight of paper for classification to establish an optimization mode database. The energy efficiency optimization model was established using the Random Forest （RF） algorithm and the Segregative Genetic Algorithm （SEGA）. The results of actual production data verification showed that after model optimization， the optimization ranges of the operating parameters of the thermal dispersion system were： torque 0~3%， temperature 0~4 ℃， and thermal dispersion machine power 0~50 kW. The optimization results were in line with the process adjustment range， indicating that the optimized model had good stability. After parameters adjusting for 4 times according to the model optimization values， the cumulative amount of adhesive decreased by about 6.36%， 5.17%， 4.25%， and 7.82%， respectively， compared to that before parameter adjustment； the comprehensive cost per ton of dry pulp had been reduced by approximately 4.62%， 3.55%， 4.43%， and 4.36%， respectively. The results indicated that the optimized operation model of the thermal dispersion system was in line with actual production conditions， which could improve the dispersion effect of pulp and reduce the energy consumption cost of thermal dispersion process.