Abstract
This paper discusses a consensus-based alternating direction method of multipliers (ADMM) approach to solve the multi-area coordinated network-constrained unit commitment (NCUC) problem in a distributed manner. Due to political and technical difficulties, it is neither practical nor feasible to solve the multi-area coordination problem in a centralized fashion, which requires full access to all the data of individual areas. In comparison, in the proposed fully distributed approach, local NCUC problems of individual areas can be solved independently, and only limited information is exchanged among adjacent areas to facilitate the multi-area coordination. Furthermore, since traditional ADMM can guarantee convergence only for convex problems, this paper discusses several strategies to mitigate oscillations, enhance convergence performance, and derive good-enough feasible solutions, including: (1) a tie-line power-flow-based area coordination strategy is designed to reduce the number of global consensus variables; (2) different penalty parameters ρ are assigned to individual consensus variables and are updated via certain rules during the iterative procedure, which reduces the impact of the initial values of ρ on the convergence performance; (3) heuristic rules are adopted to fix certain unit commitment variables to avoid oscillations during the iterative procedure; and (4) an asynchronous distributed strategy is studied, which solves NCUC subproblems of small areas multiple times and exchanges information with adjacent areas more frequently within one complete run of slower NCUC subproblems of large areas. Numerical cases illustrate the effectiveness of the proposed asynchronous fully distributed NCUC approach, and we investigate key factors that would affect its convergence performance.
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig1_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig2_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig3_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig4_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig5_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig6_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig7_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig8_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs11081-018-9375-8/MediaObjects/11081_2018_9375_Fig9_HTML.gif)
Similar content being viewed by others
Abbreviations
- d, i :
-
Indices of loads and units
- g :
-
Index of global variables
- k, n/n′:
-
Indices of iterations and areas
- t, τ :
-
Indices of hours
- m, j :
-
Indices of buses
- mj :
-
Index of a transmission line connecting bus m and bus j
- \(F_{i} ()\) :
-
Fuel cost of unit i
- \(I_{i,t}\) :
-
Unit commitment variable of unit i at time t
- \(P_{i,t}\) :
-
Real power output of unit i at time t
- \(PL_{mj,t}\) :
-
Real power flow from bus m to bus j at time t
- \(ON_{i,t} /OFF_{i,t}\) :
-
ON/OFF time counter of unit i at time t
- \(SU_{i,t} /SD_{i,t}\) :
-
Startup/shutdown cost of unit i at time t
- \(Z_{g,t}\) :
-
Global consensus variable g at time t
- \(\theta_{m,t}\) :
-
Voltage angle of bus m at time t
- \(C_{{su_{i} }} /C_{{sd_{i} }}\) :
-
Startup/shutdown cost coefficient of unit i
- \(I_{i,0}\) :
-
Initial commitment status of generator i
- \(C_{{nl_{i} }}\) :
-
No-load cost of unit i
- NT :
-
Number of hours under study
- \(ON_{i,0} ,OFF_{i,0}\) :
-
Initial ON/OFF time counter of unit i
- \(P_{d,t}\) :
-
Demand level of load d at time t
- \(PL_{mj}^{max}\) :
-
Capacity limit of line connecting buses m and j
- \(P_{i}^{min} ,P_{i}^{max}\) :
-
Minimum/maximum capacity of unit i
- \(T_{on,i} /T_{off,i}\) :
-
Minimum ON/OFF time limit of unit i
- \(UR_{i} /DR_{i}\) :
-
Ramp up/down rate limit of unit i
- \(X_{mj}\) :
-
Reactance of line connecting buses m and j
- \(\varvec{B}_{n} ,\varvec{B}_{n} \left( m \right)\) :
-
Set of buses/set of buses connected to bus m of area n
- \(\varvec{D}_{n} \left( m \right)\) :
-
Set of loads located at bus m of area n
- GL :
-
Set of global variables
- \(\varvec{I}_{n} ,\varvec{P}_{n}\) :
-
Vector of unit commitment/real power output variables for generators located in area n
- \(\varvec{L}_{n}\) :
-
Set of tie-lines in area n
- N, T :
-
Set of areas/time periods
- \(\varvec{U}_{n} ,\varvec{U}_{n} \left( m \right)\) :
-
Set of units located in area n/set of units located at bus m in area n
- \(\varvec{\theta}_{n}\) :
-
Vector of voltage angles for buses located in area n
- \(\varvec{\lambda}_{n}\) :
-
Vector of Lagrangian multipliers corresponding to area n
References
Ahmadi-Khatir A, Conejo AJ, Cherkaoui R (2014) Multi-area unit scheduling and reserve allocation under wind power uncertainty. IEEE Trans Power Syst 29(4):1701–1710
Anese ED, Zhu H, Giannakis GB (2013) Distributed optimal power flow for smart microgrids. IEEE Trans Smart Grid 4(3):1464–1475
Arroyo M (2006) A computationally efficient mixed integer linear formulation for the thermal unit commitment problem. IEEE Trans Power Syst 21(3):1371–1378
Bakirtzis AG, Biskas PN (2003) A decentralized solution to the DC-OPF of interconnected power systems. IEEE Trans Power Syst 18(3):1007–1013
Baldick R, Kim BH, Chase C, Luo Y (1999) A fast distributed implementation of optimal power flow. IEEE Trans Power Syst 14(3):858–864
Biskas PN, Bakirtzis AG, Macheras NI, Pasialis NK (2005) A decentralized implementation of DC optimal power flow on a network of computers. IEEE Trans Power Syst 20(1):25–33
Boyd S, Parikh N, Chu E, Peleato B, Eckstein J (2011) Distributed optimization and statistical learning via the alternating direction method of multipliers. Found Trends Mach Learn 3(1):1–122
Chang T, Liao W, Hong M, Wang X (2015) Asynchronous distributed ADMM for large-scale optimization—part II: linear convergence analysis and numerical performance. arXiv e-prints: 1509.02604
Chang T, Hong M, Liao W, Wang X (2016) Asynchronous distributed ADMM for large-scale optimization—part I: algorithm and convergence analysis. IEEE Trans Signal Proc 64(12):3118–3130
Cohen G (1980) Auxiliary problem principle and decomposition of optimization problems. J Optim Theory Appl 32(3):277–305
Conejo AJ, Aguado JA (1998) Multi area coordinated decentralized DC optimal power flow. IEEE Trans Power Syst 13(4):1272–1278
Damci-Kurt P, Küçükyavuz S, Rajan D, Atamtürk A (2016) A polyhedral study of production ramping. Math Program 158(1):175–205
Doostizadeh M, Aminifar F, Lesani H, Ghasemi H (2016) Multi-area market clearing in wind-integrated interconnected power systems: a fast parallel decentralized method. Energy Convers Manag 113:131–142
Eckstein J (2017) A simplified form of block-terative operator splitting and an asynchronous resembling the multi-block alternating direction method of multipliers. J Optim Theory Appl 173(1):155–182
Erseghe T (2014) Distributed optimal power flow using ADMM. IEEE Trans Power Syst 29(5):2370–2380
Erseghe T, Zennaro D, Anese ED, Vangelista L (2011) Fast consensus by the alternating direction multipliers method. IEEE Trans Signal Process 59(11):5523–5537
Feizollahi MJ, Costley M, Ahmed S, Grijalva S (2015) Large-scale decentralized unit commitment. Electr Power Energy Syst 73:97–106
Fougner C, Boyd S (2015) Parameter selection and pre-conditioning for a graph form solver. arXiv eprint: 1503.08366
Gade D, Hackebeil G, Ryan SM, Watson J-P, Wets RJB, Woodruff DL (2016) Obtaining lower bounds from the progressive hedging algorithm for stochastic mixed-integer programs. Math Program 157(1):47–67
Guo J, Hug G, Tonguz O (2016) Enabling distributed optimization in large-scale power system. arXiv e-prints: 1605.09785
Hu B, Wu L, Marwali M (2014) On the robust solution to UC with load and wind uncertainty correlations. IEEE Trans Power Syst 29(6):2952–2964
ISONE, NYISO (2011) Inter-Regional Interchange Scheduling (IRIS) analysis and options. https://www.iso-ne.com/static-assets/documents/pubs/whtpprs/iris_white_paper.pdf
Iutzeler F, Bianchi P, Ciblat P, Hachem W (2013) Asynchronous distributed optimization using a randomized alternating direction method of multipliers. In: Proceedings of the 52nd IEEE conference on decision and control
Kargarian A, Fu Y, Li Z (2015) Distributed security-constrained unit commitment for large scale power systems. IEEE Trans Power Syst 30(4):1925–1936
Kargarin A, Fu Y (2014) System of systems based security-constrained unit commitment incorporating active distribution grids. IEEE Trans Power Syst 29(5):2489–2498
Kekatos V, Giannakis GB (2013) Distributed robust power system state estimation. IEEE Trans Power Syst 28(2):1617–1626
Kim BH, Baldick R (1997) Coarse-grained distributed optimal power flow. IEEE Trans Power Syst 12(2):932–939
Kraning M, Chu E, Lavaei J, Boyd S (2013) Dynamic network energy management via proximal message passing. Found Trends Optim 1(2):1–54
Li Z, Shahidepour M, Wu W, Zeng B, Zhang B, Zheng W (2015) Decentralized multiarea robust generation unit and tie-line scheduling under wind power uncertainty. IEEE Trans Sustain Energy 4(4):1377–1388
Loukarakis E, Bialek JW, Dent CJ (2015) Investigation of maximum possible OPF problem decomposition degree for decentralized energy markets. IEEE Trans Power Syst 30(5):2566–2578
Magnusson S, Weeraddana PC, Fischione C (2015) A distributed approach for the optimal power flow problem based on ADMM and sequential convex approximations. IEEE Trans Control Netw Syst 2(3):238–253
Mota JFC, Xavier JMF, Aguiar PMQ, Markus P (2013) D-ADMM: a communication efficient distributed algorithm for separable optimization. IEEE Trans Signal Proc 61(10):2718–2723
Ostrowski J, Anjos MF, Vannelli A (2012) Tight mixed integer linear programming formulations for the unit commitment problem. IEEE Trans Power Syst 1(27):39–46
Phulpin Y, Begovic M, Petit M, Ernst D (2008) On the fairness of centralised decision-making strategies in multi-TSO power systems. In: Proceedings of the power systems computation conference (PSCC08), Glasgow, UK
PJM, MISO (2014) Interchange Optimization Workshop. https://www.pjm.com/-/media/committees-groups/stakeholder-meetings/pjm-miso-joint-common/20140418/20140418-miso-pjmjcm-interchange-optimization-presentation.ashx
PJM, NYISO (2014) Coordinated Transaction Scheduling (CTS) between NYISO & PJM - Third Joint Meeting. http://www.nyiso.com/public/webdocs/markets_operations/committees/bic_miwg/meeting_materials/2013-04-02/CTS%20NY_PJM%2003282013%20FINAL.pdf
Senjyu T, Shimabukuro K, Uezato K, Funabashi T (2003) A fast technique for unit commitment problem by extended priority list. IEEE Trans Power Syst 18(2):882–888
Shaw JJ (1995) A direct method for security-constrained unit commitment. IEEE Trans Power Syst 10(3):1329–1342
Takapoui R, Moehle N, Boyd S, Bemporad A (2015) A simple effective heuristic for embedded mix-integer quadratic programming. arXiv preprint arXiv:1509.08416
Wang Y, Wu L, Wang S (2016) A fully-decentralized consensus-based ADMM approach for DC-OPF with demand response. IEEE Trans Smart Grid. https://doi.org/10.1109/tsg.2016.2532467
Watson J-P, Woodruff DL, Strip DR (2011) Progressive hedging innovations for a class of stochastic mixed-integer resource allocation problems. Comput Manag Sci 8(4):355–370
Wei E, Ozdaglar A (2013) On the O(1/k) convergence of asynchronous distributed alternating direction method of multipliers. In: Proceedings of the IEEE global conference on signal and information processing (GlobalSIP)
Wu L (2011) A tighter piecewise linear approximation of quadratic cost curves for unit commitment problems. IEEE Trans Power Syst 26(4):2581–2583
Wu L (2016) Accelerating NCUC via binary variable-based locally ideal formulation and dynamic global cuts. IEEE Trans Power Syst 31(5):2491–2492
Wu L, Shahidepour M (2010) Accelerating the bender decomposition for network-constrained unit commitment problems. Energy Syst 1:339–376
Wu H, Zhai Q, Guan X, Gao F, Ye H (2012) Security-constrained unit commitment based on a realizable energy delivery formulation. Math Probl Eng 2012:1–22
Yan W, Wen L, Chung CY, Wong KP (2011) Decomposition-coordination interior point method and its application to multi-area optimal reactive power flow. Electr Power Energy Syst 33:55–60
Zhang R, James K (2014) Asynchronous distributed ADMM for consensus optimization. In: Proceedings of the 31st international conference on machine learning (ICML-14)
Author information
Authors and Affiliations
Corresponding author
Appendix: Illustration of proposed distributed NCUC via a two-area four-bus system
Appendix: Illustration of proposed distributed NCUC via a two-area four-bus system
In this “Appendix”, the two-area four-bus system shown in Fig. 10a is used to describe the detailed formulation of the NCUC problem (1) and the tie-line power flow based decomposition strategy proposed in Sect 2.2.
The centralized formulation of the NCUC problem for the two-area four-bus system in Fig. 10a is presented below.
Objective function (A1) minimizes the total operation cost of the entire system, which is composed of two areas S1 and S2. Constraints (A2)–(A10) correspond to the operation and security requirements of area S1. Specifically, constraints (A2)–(A7) are local operation constraints for the generator in S1 and constraints (A8)–(A9) are local network security constraints, both of which include only local variables in area S1. On the other hand, constraint (A10) is a coupling constraint which includes the tie-line power flow variable PL23,t. Similarly, for area S2, constraints (A11)–(A16) are local generator operation constraints, (A17) are local network security constraints, and (A18) are coupling constraints which include the tie-line power flow variable PL23,t.
As shown in Fig. 10b, the original system can be decomposed into two independent subsystems after duplicating bus 2/3 as bus 2′/3′ in the adjacent subsystem S2/S1. The NCUC problems for the two subsystems are as follows, where \(\widetilde{PL}_{{23^{\prime},t}}\) and \(\widetilde{PL}_{{2^{\prime}3,t}}\) are respectively the power flow variables of the same tie-line at time t in S1 and S2 (i.e., \(\widetilde{PL}_{{23^{\prime},t}} = \widetilde{PL}_{{2^{\prime}3,t}}\)).
NCUC problem for S1 | NCUC problem for S2 |
---|---|
\(\hbox{min} \mathop \sum \limits_{t = 1}^{T} \left( {F_{1} \left( {P_{1,t} } \right) + C_{nl1} \cdot I_{1,t} + SU_{1,t} + SD_{1,t} } \right)\) | \({ \hbox{min} }\mathop \sum \limits_{t = 1}^{T} \left( {F_{2} \left( {P_{2,t} } \right) + C_{nl2} \cdot I_{2,t} + SU_{2,t} + SD_{2,t} } \right)\) |
s.t. Constraints (A1)–(A12) | s.t. Constraints (A14)–(A23) |
\(PL_{12,t} - \widetilde{PL}_{{23^{\prime},t}} = P_{d1,t},\quad \forall t \in \varvec{T}\) | \(\widetilde{PL}_{{2^{\prime}3,t}} - PL_{34,t} = P_{d2,t},\quad \forall t \in \varvec{T}\) |
\(\widetilde{PL}_{{23^{\prime},t}} = \frac{{\left({\theta_{2,t} - \theta_{{3^{\prime},t}}} \right)}}{{X_{23}}},\quad \forall t \in \varvec{T}\) | \(\widetilde{PL}_{{2^{\prime}3,t}} = \frac{{\left({\theta_{{2^{\prime},t}} - \theta_{3,t}} \right)}}{{X_{23}}},\quad \forall t \in \varvec{T}\) |
\(PL_{23}^{min} \le \widetilde{PL}_{{23^{\prime},t}} \le PL_{23}^{max},\quad \forall t \in \varvec{T}\) | \(PL_{23}^{min} \le \widetilde{PL}_{{2^{\prime}3,t}} \le PL_{23}^{max},\quad \forall t \in \varvec{T}\) |
\(\widetilde{PL}_{{23^{\prime},t}} = Z_{1,t},\quad \forall t \in \varvec{T}\) | \(\widetilde{PL}_{{2^{\prime}3,t}} = Z_{1,t},\quad \forall t \in \varvec{T}\) |
Indexing \(\widetilde{PL}_{{23^{\prime},t}}\) as the second power flow variable in S1 at time t (i.e., \(\widetilde{\varvec{PL}}_{{S_{1},t}} \left(2 \right) = \widetilde{PL}_{{23^{\prime},t}}\)) and \(\widetilde{PL}_{{2^{\prime}3,t}}\) as the second power flow variable in S2 at time t (i.e., \(\widetilde{\varvec{PL}}_{{S_{2},t}} \left(2 \right) = \widetilde{PL}_{{2^{\prime}3,t}}\)), the mapping G t (n,w) = g between the global variables Zg,t and the duplicated local variables \(PL_{n,t} \left( w \right)\) defined in Section II.B can be represented as G t (S1,2) = 1 and G t (S2,2) = 1 for each time \(t \in \varvec{T}\). To match the consensus-based ADMM from (7), the coupling constraints \(\widetilde{PL}_{{23^{\prime},t}} = Z_{1,t}\) in NCUC problem 1 and \(\widetilde{PL}_{{2^{\prime}3,t}} = Z_{1,t}\) in NCUC problem 2 are relaxed into the corresponding objective functions, as shown in the following subproblems 1 and 2.
Subproblem 1 | Subproblem 2 |
---|---|
\(\hbox{min} \mathop \sum \limits_{t = 1}^{T} \left[ {F_{1} \left( {P_{1,t} } \right) + C_{nl1} \cdot I_{1,t} + SU_{1,t} + SD_{1,t} + \lambda_{1t} \left( {PL_{{23^{\prime},t}} - Z_{1t} } \right) + \rho \left( {PL_{{23^{\prime},t}} - Z_{1t} } \right)^{2} } \right]\) | \(\hbox{min} \mathop \sum \limits_{t = 1}^{T} \left[ {F_{2} \left( {P_{2,t} } \right) + C_{nl2} \cdot I_{2,t} + SU_{2,t} + SD_{2,t} + \lambda_{2t} \left( {PL_{{2^{\prime}3,t}} - Z_{1t} } \right) + \rho \left( {PL_{{2^{\prime}3,t}} - Z_{1t} } \right)^{2} } \right]\) |
s.t. Constraints (A1)–(A12) | s.t. Constraints (A14)–(A23) |
\(PL_{12,t} - \widetilde{PL}_{{23^{\prime},t}} = P_{d1,t},\quad \forall t \in \varvec{T}\) | \(\widetilde{PL}_{{2^{\prime}3,t}} - PL_{34,t} = P_{d2,t},\quad \forall t \in \varvec{T}\) |
\(\widetilde{PL}_{{23^{\prime},t}} = \frac{{\left({\theta_{2,t} - \theta_{{3^{\prime},t}}} \right)}}{{X_{23}}},\quad \forall t \in \varvec{T}\) | \(\widetilde{PL}_{{2^{\prime}3,t}} = \frac{{\left({\theta_{{2^{\prime},t}} - \theta_{3,t}} \right)}}{{X_{23}}},\quad \forall t \in \varvec{T}\) |
\(PL_{23}^{min} \le \widetilde{PL}_{{23^{\prime},t}} \le PL_{23}^{max},\quad \forall t \in \varvec{T}\) | \(PL_{23}^{min} \le \widetilde{PL}_{{2^{\prime}3,t}} \le PL_{23}^{max},\quad \forall t \in \varvec{T}\) |
Finally, the consensus-based ADMM procedure (i.e., Algorithm 1 in Sect. 2.2) for this two-area NCUC problem can be expressed as follows:
-
1.
We initialize λ1,t, λ2,t and Z1,t.
-
2.
S1 solves subproblem 1 and S2 solves subproblem 2 in parallel.
-
3.
S1 sends its \(\widetilde{PL}_{{23^{\prime},t}}\) solution to S2, and S2 sends its \(\widetilde{PL}_{{2^{\prime}3,t}}\) solution to S1.
-
4.
Both areas S1 and S2 update Z1,t via equation \(Z_{1,t} = \frac{{\left({\widetilde{PL}_{{23^{\prime},t}} + \widetilde{PL}_{{2^{\prime}3,t}}} \right)}}{2},\quad \forall t \in \varvec{T}\).
-
5.
S1 updates λ1,t and S2 updates λ2,t via:
$$\begin{aligned} \lambda_{1,t} =& \lambda_{1,t} + \rho \cdot \left({\widetilde{PL}_{{23^{\prime},t}} - Z_{1,t}} \right),\quad \forall t \in \varvec{T} \hfill \\ \lambda_{2,t} =& \lambda_{2,t} + \rho \cdot \left({\widetilde{PL}_{{2^{\prime}3,t}} - Z_{1,t}} \right),\quad \forall t \in \varvec{T} \hfill \\ \end{aligned}$$ -
6.
If the stopping criterion (7e) is satisfied, we terminate; otherwise, we return to Step (2).
Rights and permissions
About this article
Cite this article
Wang, Y., Wu, L. & Li, J. A fully distributed asynchronous approach for multi-area coordinated network-constrained unit commitment. Optim Eng 19, 419–452 (2018). https://doi.org/10.1007/s11081-018-9375-8
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11081-018-9375-8