A P2P-dominant Distribution System Architecture

Jip Kim; Yury Dvorkin

Source-linked AI summary

A P2P-dominant Distribution System Architecture

Jip Kim, Yury Dvorkin

arXiv:1902.03940v4Feb 11, 2019eess.SY

TL;DR

Peer-to-peer energy trading uses distribution infrastructure without fully accounting for utility revenues or network constraints. The paper proposes centralized and peer-centric architectures using DLMP-based network charges; simulations show the centralized configuration maximizes network welfare, while the peer-centric configuration incurs lower charges.

Problem
Peer-to-peer trades use distribution networks without financially accounting for utilities, and matching alone does not guarantee compliance with network limits.
Method
The architecture integrates peer matching with distribution AC optimal power flow and uses DLMPs to coordinate trades and calculate utility network usage charges.
Results
The system-centric configuration achieves maximum distribution-network welfare, whereas the peer-centric configuration favors neighboring trades and incurs lower network usage charges.
Takeaways & Limitations
The two configurations offer distinct trade-offs between welfare-maximizing coordination and peer autonomy with lower network usage charges.
Takeaways & Limitations
The proposed architecture may require extensions for capital costs, changing peer information and objectives, uncertainty, and meshed distribution networks.

Abstract

from arXiv · show

Peer-to-peer interactions between small-scale energy resources exploit distribution network infrastructure as an electricity carrier, but remain financially unaccountable to electric power utilities. This status-quo raises multiple challenges. First, peer-to-peer energy trading reduces the portion of electricity supplied to end-customers by utilities and their revenue streams. Second, utilities must ensure that peer-to-peer transactions comply with distribution network limits. This paper proposes a peer-to-peer energy trading architecture, in two configurations, that couples peer-to-peer interactions and distribution network operations. The first configuration assumes that these interactions are settled by the utility in a centralized manner, while the second one is peer-centric and does not involve the utility. Both configurations use distribution locational marginal prices to compute network usage charges that peers must pay to the utility for using the distribution network.

NOMENCLATURE · A. Sets and Indices · B. Parameters

The nomenclature defines buses, distribution lines, peers, peer trades, matched trades, and trade participants. Parameters specify network characteristics, peer and utility operating limits, prices, tariffs, voltage constraints, penetration, revenue, surplus value, and price adjustment.

A. Sets and Indices: The model indexes buses b ∈ B, distribution lines l ∈ L, peers n ∈ N, and peer trades ω ∈ Ω.Buying and selling peers form N = N^b ∪ N^s.
A. Sets and Indices: Each peer n has a trade set Ω_n, with the union across peers satisfying ∪_{n∈N} Ω_n = Ω.Matched trades are represented by Ω* ⊆ Ω.
A. Sets and Indices: For each trade ω, Λ_ω contains buying and selling prices, while b(ω) and s(ω) identify the participating buying and selling peers.The indices o(l) and r(l) denote originating- and receiving-end nodes of line l.
B. Parameters: Network parameters include bus susceptance B_b and conductance G_b, line resistance R_l, reactance X_l, and apparent-flow limit S_l.These quantities are expressed in p.u. or MVA as specified.
B. Parameters: Peer and market parameters define selling cost C_n(·), buying utility U_n(·), wholesale price C_w, tariff T_b, and standard trade size P.Peer buying and selling power limits are specified in MW.
B. Parameters: Operational parameters constrain active/reactive demand, peer purchases and sales, utility-generator real/reactive outputs, and squared voltage magnitude at each bus.Demand is measured in MW/MVAr, generator limits in MW/MVAr, and voltage limits in p.u.
B. Parameters: Additional parameters capture peer-trading penetration Γ_b, utility revenue Π_u, peer surplus value Υ_n, and price adjustment Δρ.The corresponding units are level, $, $/MW, and $/MWh.

C. Variables · I. INTRODUCTION

The paper introduces variables for modeling distribution-system power flows, peer trading, voltages, and dual prices, then motivates a P2P architecture that lets utilities support large-scale DER trading while addressing financial and operational challenges.

C. Variables: The model defines peer purchases and sales, utility generation, trade transfers, line flows, nodal voltages, and squared distribution-line currents.These variables cover electrical quantities exchanged among peers, utilities, buses, and distribution lines.
C. Variables: Dual variables represent forward/backward line-flow limits, active-power balance, and reactive-power balance constraints.The listed dual variables are associated with distribution-line and bus-level network constraints.
I. INTRODUCTION: DER deployment can reduce utility revenue while creating bidirectional flows, voltage fluctuations, and volatile nodal injections that complicate distribution-network operations.These effects undermine utility financial viability and challenge infrastructure designed for conventional power flows.
I. INTRODUCTION: Utilities’ tariff increases can further encourage DER adoption, creating a self-fueling utility death spiral that motivates changes to distribution-system interactions.The paper presents this cycle as an urgent techno-economic problem.
I. INTRODUCTION: P2P architecture is proposed as a coordination mechanism for heterogeneous DERs that can respect distribution-network physical limits while enabling resource sharing.Its value proposition differs from conventional economies-of-scale and scope models by monetizing under-used resources through a sharing economy.
I. INTRODUCTION: Existing P2P research includes bilateral contract networks, DC-flow models, and system-centric or peer-centric matching, but limitations remain in network physics and utility cost representation.The cited DC approximation does not capture losses, voltage regulation, or reactive-power support, while matching approaches differ in centralization and flexibility.
I. INTRODUCTION: The paper aims to accommodate large-scale P2P interactions among small-scale DERs while shifting utilities from volumetric electricity sales toward service-based revenue.The proposed platform is technology agnostic and can use blockchains or other decentralized technologies.
I. INTRODUCTION: Network usage charges are intended to encourage P2P transactions that improve distribution-system performance and offset utility revenue losses from customer-end DER deployment.The charges create an additional utility revenue stream linked to supporting electricity trading by other parties.

II. DISTRIBUTION SYSTEM WITH THE P2P PLATFORM · A. Current distribution system architecture

The paper frames the P2P platform as requiring interfaces with the distribution system, contrasting current utility-operated electricity and payment flows with proposed architectures. The current model uses regulated volumetric rates but does not adequately capture DER-related costs, revenue effects, or network conditions.

II. DISTRIBUTION SYSTEM WITH THE P2P PLATFORM: The P2P platform requires generic interfaces with the rest of the distribution system regardless of the peer-matching mechanism.These interfaces account for different volumes of electricity supplied through the P2P platform relative to the current architecture.
A. Current distribution system architecture: Small-scale consumers are charged flat or time-of-use volumetric rates that are typically regulated to recover the utility’s operating and capital costs.Operating costs include electricity supply costs and are proportional to rate T_b and active power demand D_p^b.
A. Current distribution system architecture: The current rate structure uniformly allocates operating and capital costs among customers according to electricity consumption.This assumes costs are recovered through volumetric consumption-based charges.
A. Current distribution system architecture: Customers with DERs can reduce or eliminate utility-supplied consumption, but the current approach does not adequately capture their incurred costs or reduced utility revenue.Because rate T_b combines different operating and capital costs, DER effects on operating costs cannot be accurately itemized.
A. Current distribution system architecture: Time-of-use rates capture temporal demand fluctuations only coarsely, usually through two intra-day intervals such as peak and off-peak periods.Their temporal granularity is limited despite recognizing within-day demand variation.
A. Current distribution system architecture: Current tariffs also lack spatial granularity because they do not reflect distribution-network peculiarities and are typically set at municipality boundaries.Efforts to increase spatio-temporal tariff granularity face limitations involving advanced metering infrastructure and socio-economic implications.
A. Current distribution system architecture: In the current architecture, the sole utility operates the distribution network, supplies customers, and collects electricity payments.Figure 1 compares this current arrangement with mixed and P2P architectures; solid arrows denote energy flows and dashed arrows financial flows.

B. Distribution system architecture with the P2P platform · III. P2P TRADING WITH NETWORK USAGE CHARGE · A. Peer matching process

The paper presents mixed and fully peer-to-peer distribution architectures that integrate peer interactions with utility network operations through operating-condition-based network usage charges. It defines system-centric and peer-centric matching routines, differing in whether the utility or peers control matching and trade negotiation.

B. Distribution system architecture with the P2P platform: The mixed architecture preserves utility supply for some consumers while enabling P2P interactions among stand-alone DERs and consumers.The P2P platform matches producing and consuming peers and sets their electricity price.
B. Distribution system architecture with the P2P platform: Network usage charges are collected from participating peers and incorporated into P2P price formation to recover utility operating costs.The charges are computed from operating conditions and do not recover long-term capital costs.
B. Distribution system architecture with the P2P platform: The P2P architecture is a special mixed-architecture case in which the utility supplies no electricity and the P2P platform satisfies all customer demand.This case sets Γb = 1 and leaves the utility supporting network operations.
III. P2P TRADING WITH NETWORK USAGE CHARGE: The architecture couples peer matching with distribution-network operations through matching routines and network usage charges.The matching routines are described in Section III-A, while Sections III-B and III-C introduce the charges.
A. Peer matching process: The system-centric configuration assigns the utility responsibility for welfare-maximizing peer matching.This configuration is contrasted with autonomous peer-driven matching.
A. Peer matching process: The peer-centric configuration performs matching autonomously from the utility and is driven by peer preferences and choices.Peers can negotiate trade prices in this configuration.
A. Peer matching process: In the peer-centric configuration, all trades have standard size pω = P, and multiple parallel trades can connect one buyer and one seller.The number of parallel edges reflects the seller’s generation capacity and buyer’s maximum demand.

1) System-centric configuration: · 2) Peer-centric configuration:

The system-centric configuration centrally matches peers by maximizing system-wide welfare under trading and physical constraints, while the peer-centric configuration decentralizes matching according to individual preferences. Peer-centric price adjustment produces stable matches through iterative consensus among peers.

1) System-centric configuration:: The system-centric objective maximizes aggregate peer welfare by subtracting producers’ costs from consumers’ utilities.The formulation optimizes the difference between consumer utility functions and producer cost functions.
1) System-centric configuration:: Producer sales and consumer purchases are constrained by physical generation limits and minimum–maximum purchasing bounds.Consumer bounds model elastic demand and can be converted to equality by fixing the demand parameter.
1) System-centric configuration:: The system-centric formulation computes total sold and received power over each peer’s trade set and restricts every transfer to nonnegative values.Its optimization returns the set of optimal matches, Ω∗.
1) System-centric configuration:: System-wide welfare maximization can sacrifice individual preferences, including consumer cost minimization or producer profit maximization.Peers may act strategically to increase their individual welfare.
2) Peer-centric configuration:: The peer-centric configuration decentralizes matching according to individual preferences, allowing peers to negotiate, accept, or reject trades independently of the utility.Peer decisions can incorporate bounded rationality and privacy considerations.
2) Peer-centric configuration:: Peer-centric matching seeks a stable producer–consumer match with no incentive to deviate unless producer availability or consumer demand changes.Full substitutability enables decentralized price adjustment using local peer information and limited communication.
2) Peer-centric configuration:: The peer-centric policy selects trades optimal for each peer’s preferences while modeling consumer elasticity, willingness to adjust consumption, and traded power totals.Consumer utility can be modified to represent different peer trade preferences.
2) Peer-centric configuration:: Algorithm 1 iteratively adjusts trade prices until buyer–seller prices converge, then returns jointly accepted trades as Ω∗.The process initializes prices at zero, updates prices for trades accepted by buyers but rejected by sellers, and seeks consensus among peers.

B. Network usage charges

The architecture uses DLMP-based network usage charges to incorporate distribution-network conditions into P2P trading, incentivizing supportive transactions and penalizing unfavorable ones. These charges recover operating costs but may not fund future network expansion without capital-cost components.

Network constraints: P2P trade selections are not guaranteed to satisfy distribution-network limits, potentially overloading distribution-system assets.The matching process must therefore consider network constraints.
DLMP-based charges: DLMP-based network usage charges link P2P interactions with distribution-network operations by incentivizing trades that facilitate operations and penalizing unfavorable trades.The charges reflect network conditions and are intended to influence trade selection accordingly.
DLMP calculation: The second-order-cone AC OPF model derives DLMP components accounting for energy demand, line congestion, nodal voltage, and power losses.The resulting DLMPs reflect demand supplied by both the P2P platform and utility and internalize binding network constraints.
Charge allocation: The network usage charge for each trade is equally split between its seller and buyer because both are assumed to benefit equally from using the distribution network.The factor of 2 implements this equal allocation.
Expansion limitation: DLMP-based total network usage charges recover operating costs but may generate insufficient revenue for further distribution-system expansion.Including capital costs in DLMPs is proposed to support expansion.

C. Coordination between the P2P platform and utility

The P2P platform and utility coordinate through network usage charges to comply with distribution network limits, using different approaches for system-centric and peer-centric configurations. In the system-centric configuration, co-optimization integrates P2P-platform and utility decisions and requires peers to share operational and preference data.

Coordination across configurations: Network usage charges enable coordination between P2P-platform interactions and utility operations while complying with distribution network limits.The coordination approach varies between system-centric and peer-centric configurations.
System-centric configuration: Under the system-centric configuration, the P2P platform and utility co-optimize their respective decisions.This arrangement is similar to a pool market design.
System-centric configuration: The system-centric co-optimization requires peers to share consumption, production, cost- and utility-function characteristics, and other preferences with the P2P platform.These data-sharing requirements follow from the similarity to a pool market design.

1) System-centric configuration:

The system-centric configuration co-optimizes peer-to-peer trading and distribution-network operations under P2P and network constraints. After solving the co-optimization, DLMPs determine network usage charges, and the P2P platform settles peer transactions.

System-centric configuration: The system-centric formulation maximizes OP2P + ODist.This objective is labeled max OP2P + ODist (11a).
System-centric configuration: The co-optimization enforces P2P constraints and distribution-network constraints.The P2P constraints are given in Eq. (4b)–(4f), while the network constraints are given in Eq. (7b)–(7j).
System-centric configuration: After solving Eq. (11), DLMPs are computed using Eq. (8).The resulting DLMPs are then used by the P2P platform to compute network usage charges under Eq. (9).
System-centric configuration: The P2P platform computes network usage charges and settles transactions among peers.Settlement occurs after the DLMPs have been computed.

2) Peer-centric configuration: · IV. CASE STUDY · A. 15-bus distribution test system

The peer-centric configuration iteratively negotiates peer matches without sharing individual-peer information with the utility. In the 15-bus case study, it favors neighboring-bus trades, producing lower network usage charges than the system-centric configuration.

2) Peer-centric configuration:: The peer-centric configuration updates peers’ cost and utility functions using the trade price ρω and the updated value of cn.These updates support continuation of the iterative matching procedure.
2) Peer-centric configuration:: The iterative procedure continues until a stable peer match is found, accommodating peer negotiation without sharing individual-peer information with the utility.This distinguishes the peer-centric procedure from the system-centric configuration.
IV. CASE STUDY: The case study evaluates a 15-bus distribution system and a realistic, urban-scale 141-bus distribution feeder using Julia JuMP.The code and input data are available in.
A. 15-bus distribution test system: In the 15-bus system, producers at nodes 1 and 12 have capacities of 2MW and 0.4MW, incremental costs of $50/MWh and $10/MWh, and total load is 1.63 MW.With Γb = 100% for all b ∈ B, the utility does not supply electricity and only operates the distribution network.
A. 15-bus distribution test system: The two configurations sell roughly the same capacity but allocate sold power differently among consumers.The peer-centric configuration favors neighboring-bus transactions, while the system-centric configuration enables trades among electrically remote nodes.
A. 15-bus distribution test system: Producer 1 sells exclusively to nodes 2 and 13 under the peer-centric configuration, which favors transactions among directly adjacent buses.The system-centric configuration produces a more diverse set of P2P interactions among electrically remote nodes.
A. 15-bus distribution test system: The different trade allocations produce different distribution-network utilization and network usage charges; the peer-centric configuration incurs lower charges because it does not fully utilize line capacity.Fig. 6 compares line loading and nodal voltage magnitudes under both configurations.

B. 141-bus urban-scale distribution feeder · V. CONCLUSION

The 141-bus study evaluates system- and peer-centric P2P configurations as P2P penetration varies, while the conclusion summarizes their matching and DLMP-based coordination approaches. It also identifies future work on dynamic peer information, strategic behavior, and long-term planning.

B. 141-bus urban-scale distribution feeder: The experiment adds 9 DERs to a 141-bus distribution system with a total load of 11.98MW and varies P2P penetration parameter Γb from 0 to 0.6.The utility operates the distribution network and supplies residual demand.
B. 141-bus urban-scale distribution feeder: As P2P penetration increases, the study compares line loading and voltage magnitudes under system-centric and peer-centric configurations.The comparison is presented for both configurations in Fig. 7.
B. 141-bus urban-scale distribution feeder: The system-centric configuration yields higher network usage charges because it spreads P2P resource use across the distribution network to maximize global welfare.Average network usage charges are compared across configurations and P2P platform participation levels.
B. 141-bus urban-scale distribution feeder: The peer-centric configuration produces lower, and sometimes negative, network usage charges by favoring network-friendly P2P trades without accounting for welfare maximization.This contrasts with the system-centric configuration’s welfare-maximizing allocation.
V. CONCLUSION: The proposed architecture internalizes both centralized system-centric matching and decentralized peer-centric matching among small-scale DERs.The system-centric process is welfare-maximizing, whereas the peer-centric process lets peers reflect preferences and match decentrally.
V. CONCLUSION: DLMPs are used in both configurations to coordinate the distribution system and P2P trades.The supplied conclusion passage states that DLMPs coordinate the distribution system for both configurations.
V. CONCLUSION: Future work should model dynamically changing peer utility and cost functions, additional peer and distribution-system information, and strategic peer behavior.The conclusion identifies these extensions as directions for future research.
V. CONCLUSION: Long-term planning methods should account for decentralized peer decision-making and include system-expansion capital costs in network usage charges.The proposed architecture changes operating principles, motivating these planning modifications.

APPENDIX

The appendix specifies how distribution locational marginal prices are computed and identifies the optimization variables and functional-expression parameters used in that derivation.

DLMP computation: DLMPs are computed for the origin node o(l) of line l, as derived in.The appendix introduces the computation before presenting its expressions.
DLMP formulation: The appendix identifies o(l) as dual variables of the optimization in Eq. (7) and defines parameters A1, A2, A3, A4, and A5 through functional expressions.These variables and parameters support the DLMP formulation.