A Distributed Economic Dispatch Mechanism to Implement Distribution Locational Marginal Pricing Zhao Yuan Mohammad Reza Hesamzadeh Department of Electric Power and Energy Systems KTH Royal Institute of Technology Stockholm, Sweden yuanzhao@kth.se Abstract—We address the challenges of power flow computa- tion and network operator coordination to implement distribu- tion locational marginal pricing (DLMP) in this paper. Compared with other dynamic pricing schemes, DLMP can give clearer economic signals regarding distributed energy resources (DERs) investment, demand side response, congestion management and network reinforcement. Without neglecting the power loss of distribution network, the second-order cone AC optimal power flow (SOPF) model is used here to calculate DLMP. A distributed economic dispatch mechanism based on the modified Benders decomposition and distributed generation cost (DGC) is proposed to reduce the dispatch complexity in facing high penetration of DERs. The key contribution is that we take the tie-line power flow as the complicating variable to formulate the modified Benders decomposition algorithm. The concept of DGC is proposed to reallocate the global dispatch cost to economically incentivize the regional network operators for coordination. The distributed economic dispatch mechanism is implemented in GAMS grid computing platform. Numerical results show that SOPF can give accurate power flow and DLMP results. The fast convergence of the proposed distributed dispatch is guaranteed by the convexity of the SOPF model and efficient grid computing technique. Index Terms—Distribution locational marginal pricing, opti- mal power flow, Benders decomposition, distributed economic dispatch, GAMS grid computing. NOMENCLATURE Indices: n Index for buses or nodes from 1 to n max . l Index for branches or lines from 1 to l max . i Index for iterations from 1 to i max . j Index for networks from 1 to j max . Parameters: A nl ,B nl Node to line incidence matrix. X l Reactance of branch l. R l Resistance of branch l. G n Shunt conductance at bus n. B n Shunt susceptance at bus n. p min gn ,p max gn Lower and upper bounds of p gn . q min gn ,q max gn Lower and upper bounds of q gn . Variables: p gn Active power production at bus n. q gn Reactive power production at bus n. P dn Active power demand at bus n. Q dn Reactive power demand at bus n. p s l Active power flow at the sending end of branch l. q s l Reactive power flow at the sending end of branch l. p o l Active power flow loss through branch l. q o l Reactive power flow loss through branch l. v n Voltage magnitude at bus n (lower case). V n Voltage magnitude square at bus n (upper case). V s l Voltage magnitude square at the sending end of branch l. V r l Voltage magnitude square at the receiving end of branch l. θ l Voltage phase angle difference of branch l. θ s l Voltage phase angle at the sending end of branch l. θ r l Voltage phase angle at the receiving end of branch l. I. I NTRODUCTION The planning and operation of power system is significantly being reshaped by the large-scale integration of DERs [1]– [4]. DERs can improve system reliability [2], promote energy efficiency (e.g. combined heat and power (CHP) potentially can result in system efficiency up to 90%) [2] and reduce greenhouse gas emissions [5]. Due to these benefits, there are currently various support schemes such as tax reduction, feed- in tariff and subsidy for DERs in many countries [6]. Dynamic pricing schemes in place of flat pricing in distribution network have been proposed to activate the potential flexibilities from DERs. These dynamic pricing approaches mainly include time of use (ToU) [7], critical peak pricing (CPP) [8] and real time pricing (RTP) [9]. However, none of these pricing schemes provides locational signals to the growing prosumers [10]. [3] concludes that absence of DLMP pushes the use of network charges to conveys locational signals both in the short- and long- run. Distribution networks are always built in a way to satisfy peak load which only happens during short period of the whole year [4]. A power grid always with over- invested distribution network to eliminate any possible occur of congestion can not be called smart at least in an economic point of view. Authors in [3] design a new framework for distribution net- work use-of-system charging according to the impacts on net- work costs from network users. Authors in reference [11] point out that applying nodal pricing to distribution network makes sense as DERs are transforming distribution network to be operated like transmission network. Reference [11] also shows that nodal pricing gives higher revenue for DERs by reflecting the contributions of DERs to reductions of line losses and