Design and implementation of active frequency control algorithm for virtual power plant based on flexible resource rapid response

At present, the power generation resources in the power system show a clean and decentralized development trend. These distributed generation resources are small in scale and lack of special regulation system. Aiming at the uncertain and intermittent output characteristics of photovoltaic distributed generation virtual power plant(VPP), this paper puts forward the scheme of active frequency response intelligent control, completes the hardware and communication design, algorithm design and software implementation. Finally, the actual engineering practice is carried out in three photovoltaic VPP in Zhongshan City, Guangdong Province, which verifies the feasibility of relevant schemes and algorithms, and provides a certain basis for the next research.


Introduction
Under the current situation, the power generation resources in the power system show a clean and decentralized development trend. These distributed generation resources are small in scale and scattered in layout. At present, there is no special regulation system, which can not fully meet the requirements of the rapid promotion of power marketization.
Virtual power plant (VPP), as an independent controllable system containing flexible loads and a variety of distributed generators, organically integrates distributed generators, loads, energy storage devices, converters and monitoring and protection devices, which can flexibly switch between grid connected operation and isolated island operation modes, so as to greatly improve the reliability of power supply security and main network friendliness. [1]There are a lot of flexible resources existed in the VPP, such as various energy storage equipment, new energy power generation equipment, etc. This kind of equipment responds quickly. By adjusting its power output, the active power of the VPP can quickly respond to the frequency fluctuation of the external network, so that the VPP can provide active frequency support to the outside world. [2] In the photovoltaic distributed generation VPP, its power output is random and intermittent, which is not conducive to maintaining the active power balance of the photovoltaic distributed generation VPP, and even serious frequency oscillation. [3] In this case, how to adjust effectively and control the frequency of photovoltaic distributed generation VPP intelligently needs further exploration. The purpose is to quickly and effectively realize the frequency adjustment to ensure the stable and reliable operation of the system. [4] ICPEPT

Design of active frequency control algorithm for photovoltaic distributed generation VPP
The project adopts the cluster primary frequency regulation control algorithm in the VPP based on feedback optimization. The structure of feedback optimization is shown in Figure 1. Fig. 1 Feedback optimization control structure Specifically, after the VPP collects the frequency deviation of the parallel node, the active power gap in the current state is calculated according to the deviation and the reported regulation characteristics of the VPP, and the optimal power flow solution algorithm based on second-order cone relaxation is used to dynamically solve the active power allocation problem. After the active power is distributed, each controlled equipment is controlled in place. At this time, the power gap is calculated again according to the measurement information, the above optimization problem is corrected, and the iteration converges to the optimal solution.

Active frequency response control process of VPP
When the power grid frequency changes beyond the manually defined active frequency response dead zone, the fast control device of the VPP forms the control target command according to the active frequency response active frequency droop characteristics (realized by setting the frequency and active power broken line function), and can distribute the control command of each unit according to the control constraint conditions and the operating conditions of the unit. It shall be forwarded to the monitoring system and the unit for execution through fast Ethernet. [5] The schematic diagram of active frequency response control process ( Fig. 2) is as follows: IOP Publishing doi:10.1088/1742-6596/2108/1/012020 3 1) Algorithm 1 -Initial value of FM target command: when the power grid frequency changes beyond the manually defined dead zone of active frequency response, the system forms the initial value of control target command according to the active frequency droop characteristics of active frequency response (realized by setting the broken line function of frequency and active power).
2) Algorithm 2 -FM target command execution value: coordinate control logic judgment according to control constraints and AGC control target value to form control target command execution value.
3) Algorithm 3 -Active control allocation command for each unit: combined with the operation information of the unit, the active control command of each unit is formed through the definition of control command allocation strategy, and transmitted to the monitoring system through fast Ethernet or directly distributed to each unit for execution. 4)Performance index algorithm: It mainly includes theoretical contribution power, actual contribution power, primary frequency modulation contribution rate, correct frequency modulation action rate, and real-time frequency modulation response.

Initial value algorithm for FM target command
The active power variation is realized by the given active power frequency droop characteristic curve function (Fig. 3

Instruction allocation algorithm for target value unit
The optimal control algorithm for the characteristics of VPP, on the premise of meeting the control target value of active frequency response coordinated with AGC and various safety constraints of power grid and equipment, and combined with the operation status of the unit, allocates the overall control target command, which can provide the following distribution strategies (consistent adjustment direction, smooth adjustment and optimization strategy): Proportional distribution method: P1=P2= …=Pn= P/n(n is the number of controllable units) (4) Constraints: single unit control difference ΔPn meets the climbing rate requirements. Similarity margin distribution method: P1=P10+(P-P0)/n (5) … Pn=Pn0+(P-P0)/n (6) Constraints: single unit control difference ΔPn meets the climbing rate requirements.

Real time index algorithm of active frequency response
The adjustment of active frequency response frequency step disturbance process of photovoltaic power station is shown in the following figure (Figure 4).  Fig. 4 Adjustment diagram of active frequency response frequency step disturbance process of photovoltaic power station 1) Load response lag time of active frequency response The load response lag time of active frequency response refers to the time required for the VPP from the grid frequency crossing the dead zone of active frequency response to the load change of the VPP.
2) Adjustment and stabilization time of active frequency response t0.9: When the power grid frequency changes beyond the dead zone of the active frequency response of the VPP, the load adjustment amplitude of the active frequency response of the VPP shall reach 90% of the maximum load adjustment amplitude of the theoretical calculated active frequency response within the required time; ts:When the power grid frequency change exceeds the active frequency response dead band of the VPP, within the required time, the average deviation between the actual output of the VPP and the response target of the VPP shall be within ± 1% of the rated active output of the VPP.
3)Active frequency response performance index of photovoltaic power station [5] Load response lag time of active frequency response：thx≤2s Adjustment and stabilization time of active frequency response：t0.9≤5s、ts≤15s

Engineering implementation of algorithm
At present, photovoltaic power generation accounts for the largest proportion of distributed power generation in Zhongshan, Guangdong. The real-time monitoring networking schemes adopted by photovoltaic power stations are different according to the type of grid connected inverter and the power response time of inverter. The intelligent control sub station of the VPP is deployed in Changhong Photovoltaic Station, Galanz Photovoltaic Station and Midea Photovoltaic Station to complete self-discipline control.
According to the actual situation of field equipment of Changhong Photovoltaic Station, Galanz Photovoltaic Station and Midea Photovoltaic Station, two equipment transformation schemes in Photovoltaic Station are designed as follows: 1) After fast control communication transformation, the inverter is directly connected to the intelligent control substation of the VPP to realize the fast power control of the photovoltaic inverter. Support steady-state AGC / AVC control of photovoltaic power station and primary frequency regulation of photovoltaic power station.
2) The inverter is connected to the intelligent control substation of the VPP through the existing station monitoring system to realize the conventional power control of the photovoltaic inverter. Support steady-state AGC / AVC control of photovoltaic power station.