基于FUSION模型的碳基固废热解产物产率预测

低温热解是清洁转化碳基固废、实现汇碳和减排的成熟有效方法之一。通过建立预测碳基固废热解产物产率的数学模型可以极大缩短科研探索时间，优化调控热解反应过程。本研究以80组热解实验数据为样本，首先对神经网络（ML）、支持向量机（SVM）和线性回归（LR）模型进行训练和测试，分析机器学习的有效性，然后将三种模型通过算法融合，建立具有自适应性的FUSION模型。最后，利用实验数据对该模型进行进一步的训练和测试，形成适合预测碳基固废热解产物的数据模型。融合模型能够有效解决单一模型在预测碳基固废热解产物分布过程中，受热解交互作用影响，预测精度波动的问题。同时，该模型预测值精度较高，预测值与实验值的相对误差<2%。

Products prediction of carbon-based solid waste pyrolysis based on FUSION model

Low-temperature pyrolysis is a promised method for cleanly converting carbon-based solid waste, achieving carbon sequestration and reducing emissions. By establishing a mathematical model for predicting the yields of carbon-based solid waste pyrolysis products, the time for scientific research and exploration can be greatly shortened, and the pyrolysis reaction process can be optimized and controlled. In this paper, 80 sets of pyrolysis experimental data as samples were used. The neural network (ML), support vector machine (SVM) and linear regression (LR) models were firstly trained and tested to analyze the effectiveness of machine learning, and then the three models through algorithm fusion, an adaptive FUSION model (FUSION) was established. At last, the model was further trained and tested with experimental data to form a data model suitable for predicting carbon-based solid waste pyrolysis products. The FUSION model can effectively solve the problem that a single model is affected by the interaction of pyrolysis in the process of predicting the distribution of carbon-based solid waste pyrolysis products, and the prediction accuracy fluctuates. Meanwhile, the predicted value of this model has high accuracy, and the relative error between the predicted value and the experimental value is less than 2%.