ESE 680 -- Modern Electrical Energy Infrastructure Lecture 3 Power Flow Modeling and Optimization (by an amateur) U. Topcu 1 ESE 680 -- Modern Electrical Energy Infrastructure Sample power networks IEEE 14-bus case U. Topcu 2 ESE 680 -- Modern Electrical Energy Infrastructure Sample power networks A distribution network from the service area of Southern California Edison U. Topcu 2 ESE 680 -- Modern Electrical Energy Infrastructure Ohm’s law resistance = V R= I potential difference (voltage) current = Extension to AC circuits: impedance = complex voltage complex current Z = R + jX Resistance: opposition to the passage of an electric current U. Topcu 3 Reactance: opposition to a change in current or voltage (e.g., due to inductors or capacitors) ESE 680 -- Modern Electrical Energy Infrastructure Ohm’s law Complex form of the law: V = IZ Phasor form of voltage, current, and impedance: V = |V |ej(ωt+φV ) I = |I|ej(ωt+φI ) Z = |Z|ejθ V = IZ |V |ej(ωt+φV ) = |I||Z|ej(ωt+φI +θ) ⇔ Last equality hold for all t. Hence... |V | = |I||Z| φV = φI + θ U. Topcu 4 ESE 680 -- Modern Electrical Energy Infrastructure Impedance Admittance Impedance: Z = R + jX Admittance: Y = Z −1 = G + jB Conductance Susceptance Re-write Ohm’s law in terms of admittance: I =YV U. Topcu 5 ESE 680 -- Modern Electrical Energy Infrastructure Kirchhoff’s Current Law (KCL) Let I1 , . . . , IN be the currents flowing into or out from a node/bus (with a proper sign convention). Then, N � The algebraic sum of currents in a network of conductors meeting at a point is zero. Ij = 0. j=1 U. Topcu 6 ESE 680 -- Modern Electrical Energy Infrastructure Admittance matrix KCL at node 1: Re-write with the convention that the admittance between two buses not connected be zero: Re-arrange: U. Topcu 7 ESE 680 -- Modern Electrical Energy Infrastructure Admittance matrix I1 y1 + y12 + y13 + y14 I2 −y21 = I3 −y31 I4 −y41 −y12 y2 + y21 + y23 + y24 −y32 −y42 −y13 −y23 y3 + y31 + y32 + y34 −y43 −y14 V1 V2 −y24 V3 −y34 y4 + y41 + y42 + y43 V4 In general: Note: Y is symmetric. U. Topcu 8 ESE 680 -- Modern Electrical Energy Infrastructure Electric power P = word done by electric charges moving through an electric potential (voltage) difference per unit time = (charge flow per unit time) * (voltage difference) = IV S 2 V =I R= R 2 Q (by Ohm’s law) P Generalization to complex power: S = V I∗ = P + jQ reactive power real power • • When voltage and current are in phase, no reactive power. Capacitors generate and inductors consume reactive power. U. Topcu 9 ESE 680 -- Modern Electrical Energy Infrastructure Reactive power voltage support (mentioned in the last lecture in an even sloppier way) R Vr = Vs R + jωL 1 Vr = Vs (1 − ω 2 LC) + jωL R |Vr | v.s. capacitance |Vs | U. Topcu 10 ESE 680 -- Modern Electrical Energy Infrastructure Power balance equations and “power flow problem” Power injection into a bus k: Sk = We also know: Ik = N � ∗ Vk I k Yjk Vjk j=1 Can re-write by eliminating the currents: I k = Vk N � ∗ ∗ Yjk Vjk j=1 • Variables for each bus: • Voltage magnitude and phase • Current magnitude and phase • Real and reactive power Generator bus: real power and voltage magnitude specified • Load bus: Real and reactive power specified • U. Topcu 11 Power flow problem: Given some of the variables, find the rest subject to... • the network constraints above, and • additional performance/safety/reliability constraints. ESE 680 -- Modern Electrical Energy Infrastructure Typical (additional) constraints & objectives Constraints: • Physical limitations of the generators • Transmission lines can only carry certain amount of current • Transformers have limited ratio or phase shifting capability and they heat up • Performance: bounds on voltage magnitudes and angles • Limits on reactive power sources • Limited rate of change of almost everything • ... Objectives: • Total generation cost • Losses (difference between total generation and use) • “Time” to reach acceptable network configurations U. Topcu 12 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow (a formulation of it) x : all variables, including power injections, Sk voltages, Vk current, Ik minimize over x cost(x) subject to f (x) ≤ 0 g(x) ≤ 0 Sk = Vk Ik∗ for all k: Ik = N � j=1 U. Topcu 13 Yjk Vjk eliminating the current Sk = V k N � ∗ ∗ Yjk Vjk j=1 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow: X U. Topcu 14 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow: X U. Topcu 14 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow: X U. Topcu 14 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow: X U. Topcu 14 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow: X U. Topcu 14 ESE 680 -- Modern Electrical Energy Infrastructure Optimal power flow (a formulation of it) cost(x) minimize over x Ax ≤ b subject to Cx = d Sk = V k N � pick linear, quadratic, convex, etc. affine constraints (due to operation, performan ce, safety, etc.) network for all k constraints ∗ ∗ Yjk Vjk j=1 How to “deal” with the network constraints (usually the bottleneck)? • approximate • relax U. Topcu 15 ESE 680 -- Modern Electrical Energy Infrastructure Approximations Re-write the bus power injections: Case 1 (for transmission lines): resistance << reactance → Gkj ≈ 0 Case 2: θj − θk small for neighboring buses j and j → cos(θj − θk ) ≈ 1 sin(θj − θk ) ≈ θj − θk Case 3: Voltage magnitude at every bus is equal or close to some base voltage (normalized to 1) All three simplifications together lead to the so-called DC OPF. U. Topcu 16 ESE 680 -- Modern Electrical Energy Infrastructure Another formulation of optimal power flow (for tree networks) loss minimization real and reactive power injections, real and reactive power flows, losses, voltage magnitudes underlying physical laws and network U. Topcu 17 ESE 680 -- Modern Electrical Energy Infrastructure Idea behind the solution technique Nonconvex: Quadratically-constrained, quadratic program minimize subject to c(x) convex constraints on x Convex relaxation: Second order cone program minimize subject to q(x) = 0 c(x) convex constraints on x q(x) ≤ 0 convex, quadratic function search on the boundary of the cone When does the relaxation give solutions to the original problem? • In practice, quite often. search inside the cone SDP relaxation: Bai et al. ’08 SOCP relaxation: Farivar et al. ’11, Taylor et al. ’11 Exactness conditions: Lavaei et al. ’10, Zhang et al. ’11, Bose et al. ’11 Any sufficient exactness certificates? • Only U. Topcu in special cases. One such certificate explained in the next slide. 18 ESE 680 -- Modern Electrical Energy Infrastructure A priori exactness certificate Theorem: For tree networks, the solution to the relaxation solves the original problem, if • [non-decreasing resistance-to-reactance ratio] rij rjk ≤ for each (i, j, k) s.t. (i, j), (j, k) ∈ E xij xjk • relatively outdated results −1 yij = rij + xij i resistance reactance admittance [unidirectional real power flow in the lossless case] Pij ≥ 0 for each (i, j) ∈ E Proof idea: Involved algebraic manipulations ↓ Lagrange multipliers for q(x) ≤ 0 are strictly positive at optimality ↓ Complementary slackness to show q(x) = 0 at optimality U. Topcu 19 ESE 680 -- Modern Electrical Energy Infrastructure Exactness certificate as a design guide tree network • non-decreasing resistance-toreactance ratio • unidirectional real power flow in the lossless case valid for most distribution circuits constrains the placement of distributed generation and storage substation 47-bus distribution circuit from SCE’s service area U. Topcu { • { Exactness conditions: relatively outdated results 20 ESE 680 -- Modern Electrical Energy Infrastructure Credits: [1] http://www.eeh.ee.ethz.ch/uploads/tx_ethstudies/modelling_hs08_script_02.pdf [2] http://people.ee.ethz.ch/~bacher/publications/thun_paper.pdf (very accessible overview) [3] http://home.eng.iastate.edu/~jdm/ee552/PowerFlowEquations.pdf [4] J. L. Kirtley, “Electric Power Principles: Sources, Conversion, Distribution, and Use,” Wiley, 2010. [5] http://arxiv.org/pdf/1208.4076.pdf U. Topcu 21 ESE 680 -- Modern Electrical Energy Infrastructure