District heating system (DHS) is a major part of urban energy system. DHS consists of primary and secondary loop. Compared to the primary loop, the secondary loop is closer to the heating user side. The operation control of heating substation and its corresponding secondary loop is a challenging topic. At present, the existing control of secondary loop is a manual and intermittent regulation process of its water supply temperature regardless of the weather condition, leading to either overheating or insufficient heating on the user side. Intelligent operation control of secondary loop has become a major focus of DHS digitalization. This paper proposes a new method of secondary loop operation control via a data-driven approach by dividing the process into two parts: primary loop valve opening (PLVO) prediction and corresponding secondary loop supply temperature (SLST) prediction. The goal is to provide both a time sequence of enhanced valve opening and its expected supply water temperature. Both models are trained with multilayer perceptron (MLP) and extreme gradient boosting (XGBoost) methods with consideration of the impact of input and prediction length. A demo heating substation and its secondary loop in Zhengzhou city are selected to illustrate the effectiveness of this approach. The application results show that the enhanced PLVO strategy proposed in this paper could reduce the supply water temperature deviation from the expected SLST and adapt to the change of ambient conditions. Compared with the existing operation strategy, the proposed approach reduces the supply water temperature deviation from 12.68% to 3.18% in an overheating scenario and from 6.18% to 3.87% in an insufficient heating scenario. It indicates that the proposed PLVO strategy could overcome the issue of overheating and insufficient heating in existing DHS. The results from this study could provide a basis for flexible and energy-saving operation of decentralized heating systems demanding a delicate and dynamic control of its secondary loop.

区域供热系统（DHS）是城市能源系统的重要组成部分。DHS 由初级和次级回路组成。与一次回路相比，二次回路更靠近供热用户侧。供热变电站及其相应二次回路的运行控制是一个具有挑战性的课题。目前，现有的二次回路控制是一种不考虑天气情况的手动间歇调节供水温度的过程，导致用户侧要么过热要么供暖不足。二次回路智能运行控制已成为DHS数字化的一大重点。本文通过将过程分为两个部分，提出了一种通过数据驱动的方法进行二次回路运行控制的新方法：一次回路阀门开度 (PLVO) 预测和相应的二次回路供应温度 (SLST) 预测。目标是提供增强阀门打开的时间序列及其预期的供水温度。两种模型都使用多层感知器 (MLP) 和极端梯度提升 (XGBoost) 方法进行训练，同时考虑了输入和预测长度的影响。选择郑州市的一个示范供热变电站及其二次回路来说明这种方法的有效性。应用结果表明，本文提出的增强型PLVO策略可以减少供水温度与预期SLST的偏差，适应环境条件的变化。与现有的运行策略相比，所提出的方法将供水温度偏差从 12 减小。过热情况下为 68% 至 3.18%，供暖不足情况下为 6.18% 至 3.87%。这表明所提出的 PLVO 策略可以克服现有 DHS 中过热和加热不足的问题。这项研究的结果可以为分散供热系统的灵活节能运行提供基础，该系统需要对其二次回路进行精细和动态控制。