参与电网调频的动态可控负荷及其调频策略综述与展望OA

Dynamic controllable loads(DCLs)are emerging as dispatchable frequency-regulation resources with increasing demand-side flexibility in new power systems.This study proposes an integrated framework that links the device-level response mechanisms of major DCL categories to system-level dynamic demand control(DDC),encompassing flexibility quantification,control strategy organization,and engineering deployment.Within a unified description,the framework explains how heterogeneous loads with different power ratings,response speeds,and duty constraints can be aligned with inertia support,primary frequency control,and secondary frequency control across multiple time scales. First,DCLs are organized into industrial,commercial,and residential classes.Typical representatives—such as electrolytic aluminum,power-to-hydrogen units,commercial HVAC systems,data centers,5G base stations,electric vehicles,and thermostatically controlled appliances—are analyzed.For each type,adjustment range,comprehensive response time(from frequency disturbance to power completion),device response time(to control signals),and sustainable duration are identified.DCL response spans from millisecond-level power-electronics control to minute-level aggregated comfort-constrained adjustment,and from kilowatt-level end users to hundred-megawatt industrial facilities,enabling differentiated participation in frequency services. A DDC strategy framework is proposed that includes three control approaches:rule-based,optimization-based,and intelligent control.Rule-based DDC relies on local response rules and is well-suited to decentralized,low-communication implementations for fast frequency containment.Optimization-based DDC formulates reserve scheduling and power allocation as mathematical programming problems(including deterministic and robust formulations)for individual devices and aggregated clusters,aiming to coordinate their participation in secondary frequency regulation under technical and economic constraints.Intelligent DDC introduces data-driven models and reinforcement learning to handle uncertainties and nonlinearities in DCL behavior,enabling adaptive decision-making and multi-agent coordination among heterogeneous resources.These three categories are compared in terms of modeling requirements,computational complexity,communication needs,and achievable control performance. Policy documents,technical specifications,and engineering projects related to DCL-based frequency services are further analyzed.Typical technical norms are summarized by minimum ramp rate,response accuracy,and allowable comprehensive response time for controllable loads.Demonstration projects involving electrolytic aluminum plants,commercial building HVAC systems,and vehicle-to-grid clusters are compared with emerging pilots for power-to-hydrogen units,data centers,5G base stations,and residential thermostatic loads.By mapping these cases to Technology Readiness Levels(TRL)for dynamic demand response,the study indicates that only a few DCL types,such as electrolytic aluminum and electric vehicles,have reached large-scale commercial application. The conclusions are drawn as follows.Multi-layer aggregation models can consistently represent heterogeneous DCLs from local devices through distribution networks to transmission-level frequency dynamics.Low-latency control schemes under realistic communication and computation conditions incorporate model reduction,control–communication co-design,and event-triggered mechanisms.Multi-scenario,multi-technology DDC architectures exploit virtual power plants,integrated energy systems,and emerging digital infrastructures to coordinate diverse rule-based,optimization-based,and intelligent controllers.This paper provides a reference for engineering applications of DCLs.