Coal is the major source of China's energy consumption, and this coal-dominated energy consumption structure will not change for a long time. Our country boasts the largest steel and coke output, and the coke price directly affects the quality and security of both the coke and steel industry. Current Chinese economic development has stepped into a new and important phase in which the heavy industry is fast-growing. Experience from developed countries has proved that it will be a high energy consumption period. From this view, China will face a serious shortage of coke supply. But, by scientific prediction and assessment on coke supply shortage, we can provide corporations both decision-making and strategy-framing references. This will be of great practical meaning for solving the coke supply shortage problem, realizing sustainable development of coke industry, and furthermore, for the creation of an economized society and harmonious society.

焦炭价格预测的常用方法一般有两种: 行业市场预测法和参数回归分析法。行业市场预测法对焦炭短期价格预测偏差较大, 适合长期预测。相对来说, 参数回归分析法在短期焦炭价格预测准确度上相对较高, 但由于影响因子的多样性和不宜量化性, 参数回归分析法预测误差仍然较大。这两种价格预测方法对短期焦炭价格的预测效果都不近人意, 所以本文的主要研究任务就是寻求和构建精准度更高的短期焦炭价格预测工具。

Generally, there are two ways of predicting coke price which are industrial market prediction and parametric regression analysis. The former is more effective in long-term prediction, for there is a notable deviation when used in short-term prediction. Comparatively, the latter is more accurate in short-term prediction, but the deviation still can not be ignored due to the diversity and difficulty in quantification of the influencing factors. For this reason, this article aims mainly to seek and create a more accurate tool for the short-term coke price prediction.

由于短期焦炭价格预测的影响因素较多、因素间的关系不明且不易量化, 因此很难用明确的函数关系建立预测准确度高的模型, 只能建立统计模型。在建立焦炭价格预测模型时, 要考虑市场确定因素对价格的影响, 如果还能同时考虑不确定性因素的影响, 模型也就应该具有很好的精度及普适性。北京英语翻译

Due to the complicated influencing factors, ambiguous relations among factors and difficulty in quantification, we can only build a statistical model for a better accuracy instead of using functional relations. During this process, the influence of fixed market factors should be calculated; and, the model will be of better accuracy and applicability, if the uncertain factors can also be considered. Semi-parametric model is an important statistical model developed in 80s of 20 century for solving the deficiencies of parametric model and nonparametric model. It brings in the nonparametric component representing model deviation or other system deviations. Therefore, it includes both parametric and nonparametric components, has advantages of both types of model, and is more applicable. On the other side, because it overcomes the problem of massive information loss of nonparametric model, so it can better describe realities, make full use of information and provides better accuracy. These advantages make it boast high applicable value in coke price prediction, for it just meets the demands of building a price-predicting model of coke.