From computable to perceivable: Integrating engineering methods into energy economics education with MATLAB
从可计算到可感知：工科方法融入能源经济课程的教学设计与实践——以基于MATLAB的“能源时间序列分析与预测”为例

摘要

工科方法进入经济学课堂容易简化为软件应用、算法相加，却忽略了经济问题的根本还在于人的体验、立场、选择。面对能源时间序列课程中的公式抽象、模型“黑箱”、预测结果遥不可及等问题，以能源经济专业必修课“能源时间序列分析与预测”为例，进行MATLAB支持的“可视—可调—可互动—可换位”设计。把“趋势、滞后、误差、反馈”的关系转化为图像是可视的；把样本窗口、模型参数、预测时间、评价权重转化为变量是可调的；学生可分别从能源安全、企业成本、用户可否承受、环境公平等立场“换位”调整权重，观察“最优模型”、“政策判断”随“目标”而变。由此，工科的“问题分解、系统建模、仿真检验、迭代改进”，也不只是提高计算效率，而成为降低认知门槛、展现利益差异、促进立场理解的媒介。我们对连续3学年的匿名成绩的描述性分析中：可获样本的整评平均分从82.83、82.94提到84.55；实验分项分均值从82.80提到90.40；最近一学年期末分均值提高到84.73，离散度也明显减小了。这些结果是从教学设计中预期到的“先改进做实验，再更上层楼”的机理，但受到班级的影响、考核比重的影响、对见的样本不完全一致的限制，只能是证实性的成果。课程改革的作用不在于要把“经济学”工科化，而是把抽象的能源关系看得见、调得动、可谈、可对话，而且让学生能在选择中模型体会技术背后人的位置。

Abstract

Introducing engineering methods into economics education is more likely to focus on teaching students how to use softwares and accumulate algorithm, ignoring the fact that an economic question in the end needs to reflect on human experience, position and choice. To tackle the abstract formula, black-box model and socially detached forecast often in energy time-series education, we develop a MATLAB-aided teaching design for economics class with visualization, adjustability, interactivity and perspective-taking features. Trends, lags, errors and feedback relationships are transformed to be directly displayed; sample windows, model weights, forecast span and judgement weight are to be directly adjustable. Students then adjust these weights from the perspectives of energy security, corporate cost, consumer affordability, and environmental justice, and see how the preferred model and policy judgement change with objective. Engineering practices, such as decomposition of problems, modelling of systems, simulation-based validation, debugging, and improvement, thus do not only yield computational techniques but also provide ways of lowering cognitive barriers, revealing conflicting interests, and facilitating perspective-taking. Descriptive evidence from anonymized grades for 3 academic years exhibits moderate improvement of overall performance, a stronger improvement of experimental performance and laterimprovement of final examination performance accompanied by lower dispersion, consistent with a mechanism in which interactive experiments improve participation first and then facilitate knowledge transfer, although it cannot be interpreted causally because the composition of cohorts, the relative weights of assessments and completeness of data vary across years. Our purpose is not to engineer economics, but to make abstract energy relations visible, adjustable, discussable and make students grasp the human conditions of model choice.