mmc capacitor voltage balancing in nearest level control
TRANSCRIPT
MMC CAPACITOR VOLTAGE BALANCING IN NEAREST LEVEL
CONTROL
Mattia Ricco
Aalborg University, Energy Technology Department, Pontoppidanstæde 111, 9000 Aalborg, Denmark
Outline
Introduction
Sorting Methods for NLC
Proposed Solutions:
β’ Sorting Networks
β’ Capacitor Voltage Mapping Strategy
Conclusions
14-06-2017
Structure of MMC
Phase C
Larm
Larm
#2
n
#1
#2
n
#1Phase B
Larm
Larm
#2
n
#1
#2
n
#1
Phase A
Larm
Larm
Idc
SM2,u
SMN,u
SM1,l
SM2,l
SMN,l
SM1,u
ia
ib
icVau
Val
S1
S2
VC
VSM
iSM C
D1
D2
Rarm
Rarm
Lg
Lg
Lggrid
iau
ial
Vdc/2
Vdc/2
Advantages: β’ Reduced harmonics β’ High modularity β’ Scalability β’ Low switching losses β’ High reliability β’ Fault tolerance Challenges: β’ Circulating Current Control β’ Loss balance among SMs β’ Capacitor Voltage Balancing β’ Complex control
3 14-06-2017
4
Control scheme of MMC
SM Selection
Sorting Method
Vref
VauC
Nau
iau
Outer Current Control
ialiau
CVB
ia
+ - --
Nal
lowerupper
ValC
N
ial
Switching pulses
N
NN
NLC
NLC
* / * /
Vdc
1/2
+ +Vref,l Vref,u
Averaging Control
Circulating Current Control
++
Ξ£VC
Maximum Sampling Period Ts:
πππ π =1
ππ ππ ππ1 N = number of SMs
f1 = 50 Hz
ππππππ β€ πππ π SM
Selection
Sorting Method
Vref
VauC
Nau
iau
Outer Current Control
ialiau
CVB
ia
+ - --
Nal
lowerupper
ValC
N
ial
Switching pulses
N
NN
NLC
NLC
* / * /
Vdc
1/2
+ +Vref,l Vref,u
Averaging Control
Circulating Current Control
++
Ξ£VC
14-06-2017
Outline
Introduction
Sorting Methods for NLC
Proposed Solutions:
β’ Sorting Networks
β’ Capacitor Voltage Mapping Strategy
Conclusions
14-06-2017
6
Sorting Methods for NLC Bubble Sorting Algorithm
Advantage: β’ Easy to implement.
Drawbacks: β’ The execution time is not fixed and it depends on the input list. β’ The execution time increases when N grows.
14-06-2017
Max/Min Approach
Advantage: β’ Easy to implement. β’ The sorting method is avoided. β’ Less execution time than Bubble sorting algorithm.
Drawbacks: β’ If more than one SM has to be inserted, for example during faults, the
max/min method needs more sampling periods to insert the required SMs. This affects the control dynamic.
7
Sorting Methods for NLC Tolerance Band
Advantage: β’ Reduced switching frequency.
Drawbacks: β’ A sorting method is required.
14-06-2017
Outline
Introduction
Sorting Methods for NLC
Proposed Solutions:
β’ Sorting Networks
β’ Capacitor Voltage Mapping Strategy
Conclusions
14-06-2017
9
Sorting Networks
14-06-2017
0
β’ The Sorting Networks are typically good solutions for FPGA: o Even-Odd Sorting Network o Bitonic Sorting Network
β’ They are faster than the sorting algorithms due to the parallel construction.
N = 4
Bitonic Sorting Networks with N = 8 Compare & Swap Operator
M. Ricco, L. Mathe and R. Teodorescu, "FPGA-based implementation of sorting networks in MMC applications," 2016 18th European Conference on Power Electronics and Applications (EPE'16 ECCE Europe), Karlsruhe, 2016, pp. 1-10.
10
Sorting Networks
14-06-2017
β’ The Sorting Networks are typically good solutions for FPGA: o Even-Odd Sorting Network o Bitonic Sorting Network
β’ They are faster than the sorting algorithms due to the parallel construction.
M. Ricco, L. Mathe and R. Teodorescu, "FPGA-based implementation of sorting networks in MMC applications," 2016 18th European Conference on Power Electronics and Applications (EPE'16 ECCE Europe), Karlsruhe, 2016, pp. 1-10.
Outline
Introduction
Sorting Methods for NLC
Proposed Solutions:
β’ Sorting Networks
β’ Capacitor Voltage Mapping Strategy
Conclusions
14-06-2017
12
Capacitor Voltage Mapping Strategy
14-06-2017
M. Ricco, L. Mathe and R. Teodorescu, "New MMC capacitor voltage balancing using sorting-less strategy in nearest level control," 2016 IEEE Energy Conversion Congress and Exposition (ECCE), Milwaukee, WI, 2016, pp. 1-8.
13
Capacitor Voltage Mapping Strategy
πππΆπΆππππππ and πππΆπΆππππππ are derived from the design parameters.
14-06-2017
14
Capacitor Voltage Mapping Strategy
βππ = πππΆπΆππππππ β πππΆπΆππππππ
ππ
14-06-2017
16
Capacitor Voltage Mapping Strategy
β’ Inside a voltage range the capacitor voltage values are considered to be identical
β’ M FIFO Memory Voltage Resolution β’ Memory depth equal to N β’ The position is stored in the cell
14-06-2017
πππΆπΆππππππ = 14 V ΞV = 0,5 V
17
Simulation Results
14-06-2017
Bubble Sorting Algorithm
CVMS with M = 8
CVMS with M = 16
CVMS with M = 32
CVMS with M = 64
CVMS with M = 128
Parameter Value
DC-link Voltage πππ·π·πΆπΆ 200 kV
SM Capacitor πΆπΆ 36 Β΅F
Arm Inductance πΏπΏππππππ 50 mH
Arm Resistance π π ππππππ 1 Ξ©
Number of SM ππ 16
Sampling frequency πππ π 10 kHz
Table: MMC parameters
Parameter Value Grid frequency ππππππππππ 50 Hz
Grid Voltage ππππππππππ 121.2 kV
Grid Inductance πΏπΏππππππππ 16.7 mH
Grid Resistance π π ππππππππ 0.52 Ξ©
Table: grid parameters
Outline
Introduction
Sorting Methods for NLC
Proposed Solutions:
β’ Sorting Networks
β’ Capacitor Voltage Mapping Strategy
Conclusions
14-06-2017
20
Conclusions
Two solutions have been proposed for the capacitor voltage balancing algorithm: β’ Sorting Networks β’ Capacitor Voltage Mapping Strategy
The simulation results for the CVMS have been shown.
The HIL results have also been presented.
14-06-2017