Abstract: In order to stabilize the price of fresh milk and ensure the supply of fresh milk, combined with the characteristics of modern urban agriculture, taking Tianjin as an example, prewarning mechanisms for monthly and annual price fluctuations of fresh milk were constructed. Monthly prices were predicted based on seasonal adjustment method and Holt double parameter exponential smoothing method, and a prewarning mechanism of annual price fluctuation of fresh milk was established based on BP neural network model. The results showed that combined forecast of monthly price was ideal, the price of fresh milk from January to June in 2022 would be maintained at about 4.34-4.38 yuan/kg, the price fluctuation of fresh milk were seasonal, and the price fluctuation of fresh milk at the end of each year was prone to positive light alarm or even positive heavy alarm. BP neural network model had a high prediction accuracy for annual price fluctuation of fresh milk. The test result in 2019 was 0.101 5, which was close to the actual value of 0.106 7 in 2019 and showed no alarm state. This model could be used to warn annual price fluctuation of fresh milk. In order to stabilize the price of fresh milk, we should establish a scientific and effective market information collection, analysis and release mechanism, strengthen the prewarning and monitoring of fresh milk price fluctuation, establish and optimize the price negotiation mechanism of fresh milk.

Key words: fresh milk price, prewarning, BP neural network, seasonal adjustment method, Holt double parameter exponential smoothing method

摘要： 为稳定生鲜乳价格、保障生鲜乳供给，结合现代都市型农业特点，以天津市为例，构建生鲜乳价格波动预警机制。运用季节调整法和Holt双参数指数平滑法对生鲜乳月度价格进行组合预测并进行预警；利用BP神经网络模型，建立生鲜乳年度价格波动预警机制。研究结果表明，生鲜乳月度价格组合预测结果比较理想，2022年1—6月价格基本维持在434~4.38元/kg，生鲜乳价格波动存在比较明显的季节性，每年年末生鲜乳价格波动易出现正向轻警甚至正向重警；对于生鲜乳年度价格波动，BP神经网络模型预测精度较高，2019年测试结果为0.101 5，与2019年实际值0.106 7相近且均显示为无警状态，可以用此模型对生鲜乳年度价格波动进行预警。为稳定生鲜乳价格，应建立科学有效的市场信息收集、分析、发布机制，加强生鲜乳价格波动预警监测，建立并优化生鲜乳价格协商机制。

关键词: 生鲜乳价格, 预警, BP神经网络, 季节调整法, Holt双参数指数平滑法