数据驱动的校园生活热水系统负荷预测及需求侧响应柔性潜能分析
Data-driven load prediction and analysis of demand-side response flexibility potential of campus domestic hot water systems
摘要：
高校宿舍生活热水能耗是校园供热系统能耗的主要构成之一,预测生活热水负荷并量化其柔性潜能对优化系统运行和消纳可再生能源至关重要。本文以北京市某高校为例,基于生活热水实测数据,运用聚类分析和统计拟合的方法建立了热水需求量预测模型。验证结果表明,聚类统计模型的均方根误差空调季、供暖季和过渡季分别为4.48、4.42、4.52 m3,延时曲线最大相对误差分别为5.47%、5.00%、6.46%。在此基础上,首先,引入热水分时价格调整和学分激励措施,针对学生设计了A、B、C、D 4种需求响应策略;其次,结合调研的参与意愿数据,确定了一种基于二项分布函数的响应概率计算方法;最后,计算了热水系统柔性潜能的概率分布,并采用柔性潜能指标量化了热水系统的柔性潜能期望。结果显示:削峰量、填谷量和削峰率从策略A～D均逐渐增大,最大削峰量和削峰率分别为34 921 m3/a和62.66%;削峰填谷比在39%～53%之间,削峰率在空调季和暑假较高,在供暖季和寒假较低,过渡季介于两者之间。
Abstract：
The energy consumption of domestic hot water in university dormitories is one of the main components of the energy consumption of campus heating systems, and it is important to predict the domestic hot water load and quantify its flexibility potential to optimize system operation and absorb renewable energy. Taking a university in Beijing as an example, based on the measured data of domestic hot water, this paper uses clustering analysis and statistical fitting to establish a hot water demand prediction model. The verification results show that the root mean square error of the clustering statistical model is 4.48 m3for the air conditioning season, 4.42 m3for the heating season, and 4.52 m3for the transition season, and the maximum relative errors of the delay curves are 5.47%, 5.00% and 6.46% for each respective season. Based on this, firstly, the study presents hot water time-sharing price adjustments and credit incentives, along with four student-focused demand response strategies: A, B, C, and D. Secondly, the study also develops a response probability calculation method, utilizing a binomial distribution function and integrating data on students’ willingness to participate. Finally, the study calculates the probability distribution of the flexibility potential of the hot water system,and uses the flexibility potential index to quantify the flexibility potential expectation of the hot water system. The results show that the peak shaving volume, valley filling volume and peak shaving rate increase gradually from strategy A to D, with the highest peak shaving volume being 34 921 m3/year and the rate at 62.66%. The ratio of peak shaving to valley filling varies from 39% to 53%,and the peak shaving rate is higher in the air conditioning season and summer vacation, lower in the heating season and winter vacation, and the transition season is between the two.
Keywords:data-driven; domestic hot water; clustering analysis; demand prediction; flexibility potential; uncertainty