基于数据驱动的聚合空调特性建模及控制策略

Data-driven-based aggregate air conditioning loads′external modeling and load tracking control strategy

随着越来越多的可控负荷接入电网,空调负荷(air conditioning loads,ACLs)由于其响应速度快、调节潜力大、具有蓄冷储热等特征可作为柔性负荷参与平抑可再生能源引起的波动。在实际对ACLs进行聚合的过程中,室外温度、ACLs的温度和ACLs的数量等外部条件的变化会对聚合模型造成一定的影响。但现有大多数对ACLs的控制方法未考虑聚合ACLs外部特性模型(aggregate ACLs′external model,AAEM)的时变特征。文中提出一种基于数据驱动的AAEM及相应的负荷跟踪控制策略,考虑到基于ACLs热力学特性的ACLs的成本函数,采用神经网络对时变聚合ACLs的AAEM进行快速计算,降低AAEM的计算复杂程度。基于数据驱动的AAEM负荷跟踪控制策略有效减少了ACLs的开关状态切换次数及其调用总成本。基于实际用户的仿真结果验证了该方法的有效性。

As more and more controllable loads are connected to the power grid,air conditioning loads(ACLs)can be used as flexible loads to participate in load shaving and accommodate the fluctuation caused by renewable energies because of its characteristics of fast response,large potential adjustment power,cold storage and heat storage.In the actual aggregated process of ACLs,the changes of external conditions such as outdoor temperature,temperature of ACLs and number of ACLs will have a certain effect on the aggregated model.However,the time-varying characteristic of the aggregate ACLs′external model(AAEM)is not be considered in the most of existing control methods.A data-driven-based AAEM and its corresponding load tracking control strategy are proposed.Based on the cost function of individual ACLs considering thermodynamic characteristics,the data-driven-based AAEM is established by using neural network.By this way,computational complexity of calculation AAEM is reduced.The load tracking control strategy considering the data-driven-based AAEM can effectively reduce the number of times of switching and the total cost.The simulation results verify the effectiveness of this method.