中文摘要：

准确预测空调负荷不仅对蓄能空调高效运行意义重大，而且也是新兴的冷热电三联产技术发挥技术优势的关键所在。针对同一幢建筑，分别采用了多元线性回归、季节性指数平滑法以及神经网络方法等三种典型性预测方法进行负荷预测研究，并对三种方法做了进一步改进。然后从预测精度、建模的复杂程度、工程上的可行性以及模型的其他特性（新建筑预测问题）等四个方面对负荷预测方法进行分析。结果表明：神经网络方法具有较高预测精度，而改进的季节性指数平滑法则具有较好的工程应用价值。

英文摘要：

Accurate prediction of air conditioning load is not only very important for the efficiency of thermal storage technology, but also for the cooling-heating power technology. This paper selects three typical predictive methods for carrying out load prediction of the same objective building： linear regression （LR）,seasonal exponential weight moving average （SEWMA）, and artificial neural network （ANN）. And then, some conditions for comparison relative to prediction accuracy, complexity, feasibility, as well as other specialties are put forward. The result of comparison shows： ANN is the most accurate of the three, but SEWMA can be applied in projects more easily.