**1. Introduction**

Buildings energy systems account for about one-third of the global energy consumption [1]. In China, the total energy consumption of HVAC systems is expected to account for 65% of residential buildings by 2020 [2]. Optimal control strategies have been increasingly developed in building HVAC systems [3]. The energy consumption of building energy systems can be greatly reduced by developing effective control strategies for building HVAC systems [4].

Some scholars have focused on the air-conditioning control strategies. Yordanova et al. [5] designed a fuzzy controller for temperature and humidity control. This method uses fuzzy control to ensure indoor comfort and reduce energy consumption. Wang et al. [6] proposed a direct load control strategy to optimize the distribution of set values for local and global refrigerators by setting adaptive effect functions, which saves energy while ensuring indoor comfort. Krstic [7] proposed a method based on feedback control to compensate for the input delay of any length in a nonlinear control system. Mossolly et al. [8] proposed control strategies based on energy cost and thermal space transient model constraints and they used genetic algorithms to solve problems. This optimization plan/model is suitable for building floor case studies in Beirut. Powell et al. [9] modeled the characteristics of the complexity of large-scale energy systems and used recursive neural networks to accurately predict the hourly load capacity of regional energy systems 24 h in advance. Other researchers have studied the operation and simulation of central air-conditioning systems. Wei et al. [10] obtained the operating power consumption curve based on the mathematical.

Model of the established air conditioning system equipment to determine the optimal operating plan. Chan [11] proposed a solar heating and cooling (SHC) absorption chiller of central air-conditioning system design based on the TRNSYS simulation model for a hotel building. Di fferent control strategies are loaded into the TRNSYS simulation model to evaluate the superiority of di fferent control strategies. Li et al. [12] proposed a central air-conditioning solar heating and cooling (SHC) absorption chiller system based on TRNSYS and proposed three control schemes for the solar collector circuit to determine the preferred design strategy for these systems. Xue et al. [13] proposed a fast power demand response control strategy to investigate the performance of operational dynamics and energy systems in response to strategically controlled demand response events. From the analysis of research status at home and abroad, the current research on control strategies of central air-conditioning systems focuses on the combination of intelligent algorithms and hybrid models. However, the model has insu fficient accuracy, inaccurate control, problems such as control errors and overshoot, and overly complex control models are not suitable for actual engineering control.

Judging from the progress of theoretical research at present, although there are many types of research on advanced HVAC system control technology [14,15], the control modes of building HVAC systems presently have a grea<sup>t</sup> limitation both in control methods and controlled parameters. From the control method point of view, PID control [16–18] is a kind of negative feedback control system, which is widely used in the control of HVAC systems of public buildings by using the proportional integral and di fferential method to calculate the control amount according to the system deviation. For controlled objects with inherent nonlinearity and hysteresis characteristics [19] such as the HVAC system, it is di fficult to obtain an ideal PID control e ffect due to the uncertainty and time-varying nature of external environmental disturbances. From the controlled parameters point of view, constant pressure control [20] and constant temperature control [21,22] are widely applied. However, there are significant drawbacks to the constant pressure di fference and constant temperature di fference control of the air conditioning system. On one hand, for constant pressure di fference control, there is no direct relationship between the load and pressure di fference of the HVAC system. It is not possible to use the di fferential pressure as a controlled variable to ensure that the chilled water flow changes accurately following the load change [23]. Moreover, the return temperature of the chilled water is inconsistent with the water supply temperature due to the transmission delay of the HVAC system. It is unscientific to adjust the chilled water flow rate according to the temperature di fference between the supply and return water detected at the same time as the controlled parameter. Therefore, the control mode based on water supply temperature commonly used for HVAC systems is only applicable to controlled objects or processes without time delay [24,25]. New control techniques and methods need to be adapted to meet the actual needs of the stability and rapid response of the central air conditioning system.

Aiming at the problems existing in HVAC system control technologies, a load forecast fuzzy control strategy was proposed. The predicted load obtained by the SVM method training is used as an input parameter to the controller in advance for feedforward fuzzy control, which can regulate the HVAC system in advance based on the forecast cooling load demand and overcome the shortcomings of controlled parameters. In this study, a simulation platform was established for the heat pump system in Tianjin University based on TRNSYS and MATLAB to confirm the advantages of the proposed LFF control strategy.
