Liberalized market designs for district heating networks under the EMB3Rs platform
António Faria, Tiago Soares, Zenaida Mourão, José Maria Cunha
LLiberalized market designs for district heatingnetworks under the EMB3Rs platform st Ant´onio Faria
INESC TEC
Porto, [email protected] nd Tiago Soares
INESC TEC
Porto, [email protected] rd Jos´e Maria Cunha
INEGI
Porto, [email protected] th Zenaida Mour˜ao
INEGI
Porto, [email protected]
Abstract —Current developments in heat pumps, supportedby innovative business models, are driving several industrysectors to take a proactive role in future district heating andcooling networks in cities. For instance, supermarkets and datacenters have been assessing the reuse of waste heat as an extrasource for the district heating network, which would offset theadditional investment in heat pumps. This innovative businessmodel requires complete deregulation of the district heatingmarket to allow industrial heat producers to provide waste heatas an additional source in the district heating network.This work proposes the application of innovative market de-signs for district heating networks, inspired by new practices seenin the electricity sector. More precisely, pool and Peer-to-Peer(P2P) market designs are addressed, comparing centralized anddecentralized market proposals. An illustrative case of a Nordicdistrict heating network is used to assess the performance of eachmarket design, as well as the potential revenue that differentheat producers can obtain by participating in the market. Animportant conclusion of this work is that the proposed marketdesigns are in line with the new trends, encouraging the inclusionof new excess heat recovery players in district heating networks.
Index Terms —District heating networks, Excess heat, Marketliberalization, Energy Market, Peer-to-peer
I. I
NTRODUCTION
Over the years, District Heating and Cooling (DHC) systemshave been proliferating in many countries [1]. In Denmark,according to EUROHEAT & POWER, 65% of citizens wereserved by Distric Heating Networks (DHNs) in 2017, account-ing for more than 30 000 km of pipelines in DHNs. MostEuropean DHC systems follow a monopolistic approach due toheat demand sparsity, the market power of a single generatingunit that often owns the DHN, the lack of DHN linking allpossible customers, and long-term return on investment. Thesereasons pull back new investors and market liberalization,which could foster the reuse of waste heat as an extra sourcein DHNs [2, 3]. In fact, DHN is a natural monopoly due tothe large infrastructure and operation costs, concerning theproduction and distribution of heating and cooling. Therefore,the heat production plants and the network are commonlyowned, operated and managed by the same company, which isthe main obstacle to the complete liberalization of the system[4]. Overall, DHC systems are heavily regulated and pricecompetitiveness for consumers is disregarded.Nevertheless, governments (through energy regulators andpolicymakers) are enforcing the liberalization of heat markets (similar to what happened in the power system), as it becomeseasier to monitor the whole process of energy systems, aimingto drag the prices down through competition, once the energyproviders are competing with each other, leading to economicbenefits for consumers [5, 6, 7, 8, 9]. Therefore, DHC mar-ket liberalization is gaining momentum in some Europeancountries, aiming to replicate and adapt the good experiencewith electricity markets, bringing their capacity to improvesystem efficiency [7, 10, 11]. This disruptive paradigm shiftwill increase competitiveness through the inclusion of newplayers in the system. That is, several agents from differentindustry sectors can play an active role in the DHC market bybuying and selling energy from different sources, increasingcompetitiveness and bringing financial benefits to everyoneinvolved [6, 12]. The authors in [10, 13] present case studiessuggesting that a large amount of heat demand can be suppliedby industries, e.g., by supplying waste/excess heat of industryprocesses to neighbouring consumers. Similarly, the authors in[14, 15] also demonstrate the benefits that external producers(taking advantage of heat pumps, waste heat and renewableheat technology) bring to the DHC system if they supply theirexcess heat to the DHN. The results would be advantageousfor all parties, bringing economic and environmental gains.On the other hand, the works in [16, 17, 18] show thebenefits of the synergies between the power and DHC systems,modeling centralized dispatches to improve the efficiency ofthe entire energy system. In addition, consumers can alsoplay an active role in the DHC system, providing demandflexibility in response to dynamic tariffs, thereby improvingmarket competition [19, 20, 21, 22].DHC markets inspired by the electricity sector, applyingconventional market designs and approaches, are growing [11,23]. An example of a running DHC market is the Open DistrictHeating project [24], operating at Stockholm’s DHN, whichencourages industrial businesses to sell their excess heat tothe DHN at a uniform price cleared in the proposed day-aheadheating market.In addition, innovative market ideas to increase competitive-ness in the DHN are emerging in the literature [25, 26, 27].One of them is the adaption of the sharing economy principleto industries and small-scale production units to supply surplusheat to the DHN [13, 28]. In this regard, different consumer-centric market designs, adapted from the power system, are a r X i v : . [ phy s i c s . s o c - ph ] J a n xpected to be replicated to the DHC system, allowing thesenew market participants to inject heat in the DHN and get extrarevenue. In order to assess several options and assumptions forthe best market design to apply in existing and new DHNs,a brand new platform (EMB3Rs) is being developed [29].This platform will empower different stakeholders (e.g., utilitycompanies, municipalities, DHN operators, excess heatingproducers, among other entities) to simulate distinct marketdesigns that can be applied to current and future DHNs.In this context, this work contributes to the literature andto the EMB3Rs platform, modelling distinct market modelsfor the negotiation of heat in DHNs considering a competitiveenvironment. More precisely, three distinct market designs aremodelled and compared, namely, the pool-based, the peer-to-peer (P2P), and the community-based market designs. Themarkets are adapted from the current and future trends inelectricity markets. Additionally, consumers preferences (e.g.,distance, losses and CO ) through product differentiation areapplied to the P2P market design, enabling consumers tochoose sources they prefer to be provided from. An illustrativeDHN based on Nordic countries is used to test the applicabilityof the proposed solution. The main contributions of the presentwork are fourfold: • To implement, analyze and compare, different marketmodels in the EMB3Rs platform; • To model new market designs for heat exchange in theDHN, namely, the pool-based, P2P, and community-basedmarket designs; • To explore competitiveness in DHC markets, enablingindustrial businesses with excess heat recovery systemsto inject excess heat in the DHN; • To improve market options for consumers by introducingproduct differentiation in the P2P market design.In addition to this introductory section, this paper is orga-nized as follows. Section two describes the EMB3Rs plat-form for the simulation of different DHC market designs.Section three presents the detailed mathematical models of theproposed market designs. Section four assesses the proposedmarket models considering an illustrative case of NordicDHNs, while section five gathers the conclusions of the study.II. EMB3R S P LATFORM FOR
DHC M
ARKET S IMULATION
This section provides an overview of the EMB3Rs platformthat will incorporate current and new market designs, adaptedto the context of DHC systems. In addition, it provides a briefreview of the actual situation of the DHC markets in the Nordiccountries.
A. Current DHC Market Situation in Nordic Countries
The current situation of DHC markets varies on a countrybasis, as the deregulation of DHC systems has been car-ried out in different ways [30]. In Denmark, the DHN isstill a natural monopoly, as the network and heating plantsare mostly owned by energy companies, municipalities orconsumer cooperatives. The regulation dictates that the heatsupply works under non-profit rules, which means that the supplier must provide heat to consumers at marginal cost. Thisnon-profit rule benefit everyone, as any profits are distributedto consumers to reduce costs [31]. In this case, industries withexcess heat are encouraged to self-consume and only then tosell excess heat to the market, since the sale of excess heatcomes with a tax to prioritize energy efficiency [30].Similarly to Denmark, DHNs are also heavily regulated inNorway. DHNs are mostly private and municipal owned, withmandatory connections to consumers decided by the munic-ipalities, while the operator is forced to expand the network[30]. The energy price from different producers are set on acompetitive market, but prices for consumers with mandatoryconnections are regulated and cannot exceed the price ofelectric heating within the supply area [32]. Alternatively,consumers without mandatory connections are free to choosetheir heating source (e.g., electric heating or heat pump), sothe supply price will follow the electricity price [32, 33].In contrast, Sweden was one of the first European countriesto deregulate the heating market, however, that deregulationwas not as robust as expected. According to [34], the pricesof the different Swedish utility companies are not similar,meaning that these companies behave as price-makers. Thecosts are related to heating production and DHN operation,while what was expected was marginal-based pricing. On theother hand, Finish utility companies have a monopoly oncertain DHNs. Costumers have no open market to select theirDHC utility [35]. Some Finish companies have been tryingto change this paradigm, i.e., offering seasonal tariffs, butthese measures also do not shape the fair price for customers[19]. For further details on the situation of DHCs systems inEuropean countries, interested readers are referred to [30, 36].Nonetheless, the transition to sustainable, efficient and com-petitive markets is unavoidable and future DHC markets willrequire new market approaches suitable to the integration ofrenewable energy sources in DHNs [37].
B. EMB3Rs Platform Overview
The EMB3Rs platform has been designed to assess the reuseand trade of excess thermal energy in a holistic perspectivewithin an industrial process, energy system environment, or inan DHN under regulated and liberalized market environment[38]. The platform empowers industrial users and stakeholdersto investigate the revenue potential of using industrial excessheat and cold as an energy resource, based on the simulationof supply-demand scenarios. Therefore, the platform simulatesmultiple business and market models, proposing innovativesolutions in the sector.From the large variety of options, users can: ( i ) map new andexisting supply and demand users with geographic relevancyand enable their interlink; ( ii ) assess costs and benefits relatedto the excess heat and cold utilization routes, consideringexisting and new network infrastructure (e.g., DHN); ( iii )explore and assess the feasibility of new technology andbusiness scenarios; and ( iv ) compare and analyze distinctmarket models applied to the DHN to dynamically create newusiness models and identify potential benefits and barriersunder specific regulatory framework conditions.The integration of a dedicated market module in EMB3Rsplatform allows users to perform market analysis consideringmultiple existing market designs. Therefore, users can create,test and validate different market structures for selling andbuying energy in the DHN, identifying barriers and risks, aswell as regulatory framework conditions required to ensurethat the implementation of such market solutions are eco-nomically feasible. That is, the market analysis enable users(e.g. industries, supermarkets and data centers) to estimatepotential revenues from selling excess heat and cold. Thisis especially important for users who have invested (or areconsidering investing) in waste heat recovery technology toassess the potential economic and environmental savings oftheir investment. C. Market Approach for Heat Exchange
On the EMB3Rs platform, users must be able to exploredifferent market designs, from centralized to the decentralizeddesigns, allowing them to analyze the best market frameworkfor their interests, which can be economic, environmental orsocial.In this regard, three distinct market designs are adapted inthe present work to be included in the EMB3Rs platform. Theconventional pool market, the innovative P2P and community-based market designs are addressed to ensure that the plat-form’s users (e.g., industries, supermarkets and data centers)can assess their business models under different levels ofmarket decentralization for the exchange of thermal energyin DHNs. All the three market designs are inspired in theelectricity sector, and therefore, need to be adapted to theunderlying characteristics of DHC systems.The pool market follows a systemic perspective of thewhole market by applying the merit order mechanism andperforming the intersection of production and demand curves.This mechanism, known as uniform price, results in a marketclearing price that is used for the settlement of producers andconsumers. That is, each producer and consumer scheduled inthe market will receive and pay for the energy at the marketclearing price, respectively.In contrast, consumer-centric market designs (such as P2Pand community-based market designs) follow a more decen-tralized and consumer-focused perspective. The P2P marketenables producers and consumers to exchange energy directlywith each other, subject to certain specific conditions definedby consumers. In this market design, no central facilitatoris needed to verify energy exchanges. On the other hand,the community-based market requires the use of a centralentity that coordinates energy exchanges within the energycommunity, well as the imports and exports to other energycommunities and DHN players. It worth mention that thesekind of decentralized markets can empower consumers andprosumers to play a more active role in the DHN. For instance,local supermarkets are emerging thermal prosumers that canprovide and consume heat in different hours, making them a flexible player to reuse excess heat and even selling surplusheat to other consumers in the DHN.III. D
ISTRICT H EATING M ARKET D ESIGNS
The DHC market designs discussed in this work, representinsights into the future of heat exchange in DHNs. There isstill a long way to go regarding infrastructure and legislationfor the implementation of liberalized markets. In this context,the first steps in what we believe could be the DHC systemsof tomorrow are given in this work. In this way, pool, P2P andcommunity-based market approaches are addressed. Note thatfor the rest of the work, it is assumed that the heat sourcesare considered producers and the heat sinks are consumers.
A. Pool Market Design
The pool market designs consists of representing the meritorder mechanism and obtain the market clearing price throughthe intersection of supply and demand curves. Thus, the markethas the goal of maximizing the social welfare, meaning thatlower offers from producers and higher offers from consumersare accepted. Mathematically, this market can be presented as: min D (cid:88) n ∈ Ω n C n P n (1a)s.t. P n ≤ P n ≤ P n p ∈ Ω n (1b) (cid:88) n ∈ Ω n P n = 0 (1c) P n ≤ n ∈ Ω c (1d) P n ≥ n ∈ Ω p (1e)where D = { P n ∈ R } n ∈ Ω n correspond to the energy tradedby each agent n . C n represents the agents’ bid price; P n , P n , represent the lower and upper bound of the agents’energy offer, respectively; Ω c represent the consumers sets, Ω p represent the producer sets. Eq (1b) set the agents offersboundaries. Eq (1c) sets the market balance, where the supplymust equal the demand. (1d) sets that the consumption is non-positive in the system, while (1e) sets that production variablefrom producers is non-negative. B. P2P Market Design
Regarding the P2P approach, it is proposed that two differ-ent peers can trade heat on a bilateral basis, without a thirdparty supervision [39]. That is, each peer n can exchange withanother peer m on an individual basis, defining the amount ofenergy to be bought or sold at a given price. This problem canbe mathematically formulated as follows: in D (cid:88) n ∈ Ω n C n P n (2a)s.t. P n = (cid:88) m ∈ Ω n P n,m n ∈ Ω n (2b) P n ≤ P n ≤ P n n ∈ Ω n (2c) P n,m + P m,n = 0 { n, m } ∈ { Ω n } (2d) P n ≤ n ∈ Ω c (2e) P n ≥ n ∈ Ω p (2f)where D = { P n ∈ R } n ∈ Ω n represents the heat traded by eachagent n . Like in the pool market, the goal is to minimize thecost associated with the agents’ transactions (2a). The totalheat traded by an agent n must equal the sum of the heatexchanges from that agent n to the other agents m (2b). Also,a reciprocity is expected in the bilateral trades (2d), where P n,m and P m,n must be symmetric.Looking at the peer-to-peer formulation, one can see thatit yields the trade between agents. Thus, a preference can beadded to each of these trades, which can be translated intoa penalty or benefit. This is called product differentiation,meaning that a certain trade can be advantageous or harmfulto the system management. In this way, the objective functionis willing to benefit or penalize the trades that deserve suchconsideration. The distance between agents, the thermal lossesand the CO emissions are preferences that can be placedwithin this scope. There is also the option where the agents canchoose the penalty that best suits their ideology. For instance,on the EMB3Rs platform, three different penalty options areprovided to the consumers. One option is the physical networkdistance between agents. For example, an agent can select thedistance penalty if he wishes to trade with the nearest neighbor.Another option is thermal losses, where an agent can selectthe thermal losses penalty if it is concerned about the systemenergy efficiency. Alternatively, the CO penalty is proposedif an agent has environmental concerns. Conventionally, theproduct differentiation is represented as: C n,m = P n,m c n,m (3)where C n,m represents the final penalty applied to the tradebetween agents n and m . P n,m represents the energy tradebetween agents n and m , and c n,m represents the initialpenalty between these agents.In order to apply product differentiation, the objectivefunction must account with the penalty from Eq (3). Thus,the objective function takes the following form: min D (cid:88) n ∈ Ω n C n P n + (cid:88) n ∈ Ω n (cid:88) m ∈ Ω n C n,m (4)where D = { P n , C n,m } ∈ R n,m ∈ Ω n . Hence, the formula-tion is completed, since equations (2b)-(2f) keep unchanged.Nevertheless, the determination of the product differentiationpenalties may follow different ways.
1) Physical Network Distance Preference:
In the distancepreference, the network distance between the selected agentsis determined. The penalty implies the sum up of all thepipes that make the path between agents. Note that Dijkstra’salgorithm [40] is used to find the shortest path between agents.Thus, the penalty associated to the network distance is givenby: c n,m = (cid:88) i ∈ Ω In,m d i,n,m /T otDist (5)where d i,n,m represents the pipe distance along the pathbetween agents n and m , while 385.08m is the total networkdistance.
2) Network Thermal Losses Preference:
The thermal lossespenalty between two agents is given by the share that eachagent has in the system losses considering the thermal flow ineach pipe. In this case, it is required to determine the thermalflow in the DHN and, therefore, the losses in each pipe. Todetermine the thermal flow and losses in the DHN based on theinitial market results, the thermal control algorithm in [41] isused. Therefore, the impact that each agent has on the thermalflow and losses of each pipeline is determined using Bialek’sdownstream looking algorithm [42]. Finally, the thermal lossespenalty for the transaction between two peers is given by: c n,m = (cid:88) i ∈ Ω In,m l i,n,m D i,n,m d i,n,m /T otLoss (6)where l i,n,m represents the thermal losses in each pipe alongthe path between agents n and m ; D i,n,m represents the n, m peer impact in each pipe of the system determined by thedownstream looking algorithm presented in [42]. In this way, afairly penalty allocation for the transaction between two agentsis achieved, accounting for the cumulative impact that suchtransaction has in the thermal losses in the system. CO Emissions Preference:
The last option proposedfor product differentiation is to penalize transactions through CO emissions. This penalty consists of penalizing peertransactions that may, consequently, emit higher emissions intothe atmosphere. The EMB3Rs platform can provide standardlevels of CO per technology, and therefore, penalties betweenagents n and m consider such levels. Here, the penalty isonly associated with the heat source. Hence, the CO penaltybetween agents n and m is given by the quotient betweenagent n emissions and the total system emissions: c n,m = E n / (cid:88) n ∈ Ω n E n (7)where E n represents the CO emissions by agent n . C. Community-based Market
The community-based market design intends to represent amore hierarchical structure of bilateral peer trades. In general,a community is composed by members who share commoninterests or are geographically close. In this model, there is acommunity manager responsible for the community’s energyanagement. This manager supervises all the trading activitieswithin the community, as well as works as an intermediary inthe heat trade with other communities or with the main grid[43]. The mathematical formulation is presented as: min D (cid:88) n ∈ Ω n (cid:88) k ∈ Ω k C n,k P n,k − c exp,k q exp,k + c imp,k q imp,k (8a) P k,k (cid:48) + P k (cid:48) ,k = 0 , ∀ ( k, k (cid:48) ) ∈ (Ω k ) (8b) q exp,k (cid:48) = (cid:88) k ∈ Ω k P k (cid:48) ,k , ∀ k (cid:48) ∈ Ω k (8c) q imp,k (cid:48) = (cid:88) k ∈ Ω k P k (cid:48) ,k , ∀ k (cid:48) ∈ Ω k (8d) (cid:88) k ∈ Ω k P k (cid:48) ,k = q exp,k (cid:48) − q imp,k (cid:48) , ∀ k (cid:48) ∈ Ω k (8e) P n,k + q n,k + α n,k − β n,k = 0 , ∀ ( n, k ) ∈ (Ω n , Ω k ) (8f) (cid:88) n ∈ Ω n q n,k = 0 , ∀ k ∈ Ω k (8g) (cid:88) n ∈ Ω n β n,k = q exp,k , ∀ k ∈ Ω k (8h) (cid:88) n ∈ Ω n α n,k = q imp,k , ∀ k ∈ Ω k (8i) P n ≤ P n ≤ P n ( n, k ) ∈ (Ω n , ω k ) (8j)where D = { P n,k , q exp,k , q imp,k ∈ R } ( n,k ) ∈ (Ω n , Ω k . P n,k rep-resents the internal trade of agent n within its own com-munity k . (8b) represents the symmetry when communitiesexchange heat. Equation (8c) balances the exported heat bya community with other communities. The same is validfor (8d), regarding the imported heat. Also, the sum of onecommunity bilateral trades must equal the exported heat minusthe imported heat (8e). Equation (8f) sets agents’ balance,i.e., the purchase/consumption, the heat traded within thecommunity and the heat exchanged with other communitiesmust reach an equilibrium in each time period. Within acommunity, the purchase/consumption of all involved agentsmust be equal to zero (8g). Furthermore, the heat exportedby each community agent must equal the total heat exportedby the community (8h). The same holds true for the importedheat (8i). Like in the previous market designs, heat boundariesought to be kept (8j). IV. C ASE STUDY
In this section, a case study is presented considering an il-lustrative Nordic DHN with several producers and consumers.This illustrative example has been developed to assess differentmarket designs on the EMB3Rs platform. All the input dataand results of this study, including demand and supplier offersfor an entire year (from April 2018 to March 2019) areavailable at Mendeley Data [44].
A. Case Characterization
A DHN has been built considering several producers andconsumers with different characteristics and patterns.
Fig. 1. Illustrative district heating network.
Note that the DHN must ensure that the temperature iswithin the levels required by the heating demand, and that theflow rates in the DHN must be kept at a reasonable low levelin order to avoid water velocities. To this end, it is assumedthat this DHN operates similarly to most Danish DHNs, whichwork within annual averages temperatures of 77.6 ◦ C supplyand 43.1 ◦ C return [45].Figure 1 shows the schematic diagram of the DHN, where31 row houses and 4 potential producers are considered. Theconsumption of 31 row houses for a entire year (from April2018 to March 2019) has been generated considering a typicaldemand pattern taken from [46]. The price that the row housesare willing to pay for the demand in the market follows anormal distribution, in which the base price is the heat tariffin Copenhagen, Denmark [47]. In order to suppress basicconsumption needs, at least 70% of the heat demand of eachhouse must be supplied at all periods.A 15 kW industrial ammonia heat pump is located in theDHN and can provide heat at some time of the day at a certaincost. The heat pump generation profile considers a constantCoefficient of Performance (COP) of 4.8, providing hot watervia a heat exchanger at 80 ◦ C, based on [48, 49]. The costcurve of the heat pump is based on the electricity spot pricein 2018 and 2019 in DK2 area in Denmark, taken from [50].In addition, a 0.4 MW data center is included in the DHN.Commonly, data centers follow a relatively constant patternof excess heat recovery to inject into the DHN, although thetemperature of their excess heat from the condenser coolingtowers is usually between 35 ◦ C and 45 ◦ C. Thus, an industrialammonia heat pump, similar to the one referred above, wouldbe required to upgrade its heat to inject into the DHN. Thisdata center has been modeled producing 71,6 kWh on average,in which the calculus for the heat recovery profile is based on[51, 52]. To this value, it would be added the energy used inthe ammonia compressor. The cost curve for the data centersell recovered heat energy in the DHN has been modeledfollowing a normal distribution and the monthly excess heat
ABLE IDHN
DISTANCE BETWEEN AGENTS . Distance (m)Agent CHP Supermarket Data Center Heat PumpC1-C10 266,24 181,25 206,15 174,96C11-C15 190,76 20,47 168,58 199,06C16-C18 228,66 143,67 230,27 137,38C19-C25 175,25 90,26 158,21 127,01C26-C30 196,37 111,38 224,52 193,31C31 259,32 174,33 122,23 94,28SM 201,37 - 240,87 209,67 procurement costs presented in [52]. A Combined Heat andPower (CHP) unit is included in the DHN being the mainproducer in the system. This CHP is designed to provide theentire consumption of the system, being therefore the mostexpensive generation resource. The cost curve for a entire yearfollows the behavior of the natural gas spot price for years2018 and 2019, available in [53]. Note that the prices werenormalized for the Nordic context.Besides this, a supermarket with heat pump technology isincluded in the system behaving like a prosumer. That is,the supermarket may consume heat from the DHN or injectrecovered heat into the DHN, taking into account the hourof the day and the outdoor temperature. The generation andconsumption profile depends on the outdoor temperature. It hasbeen considered the outdoor temperature in Copenhagen forthe entire year (April 2018 to March 2019), available at [54].Then, the prosumer profile of the supermarket is determinedfollowing a typical COP (around 3.0) for heat recovery insupermarkets, and a typical supermarket consumption pattern,detailed in [55]. The cost curve for the supermarket to injectrecovered heat in the DHN depends on the outdoor temperatureand is based on [56].It is noteworthy that different market designs may requirethe use of different data or configurations. For example, thecommunity-based market design requires the configurationof the energy community, that is, who are the communitymembers. For the community-based market, two communitieswere created, based on the aforementioned energy resources,namely: • Community 1: Data Center and all consumers from 19 to31; • Community 2: Supermarket, Heat Hump, and consumersfrom 1 to 18.Regarding the P2P market model via product differentiation,the required data were retrieved based on the THERMOSproject tool [57]. This tool is able to provide the distance(Table I) and nominal losses (Table II) between agents, basedon the supply and return temperatures, and on the maximumheat flow in the pipelines.The CO signals for the CHP were obtained from [58],while for technologies that rely on the electricity mix wereretrieved from [59] considering the Nordic zone. Table IIIpresents the CO signals for all heat producers. TABLE IIDHN
NOMINAL LOSSES BETWEEN AGENTS . Losses (W/m)Agent CHP Supermarket Data Center Heat PumpC1-C10 17,31 16,40 17,31 14,02C11-C15 18,35 17,23 16,83 17,43C16-C18 17,90 17,12 17,73 17,58C19-C25 18,10 18,01 17,78 17,49C26-C30 17,24 16,51 17,41 16,43C31 17,39 16,99 17,05 16,66SM 18,64 - 16,87 17,86
TABLE III CO EMISSIONS BY HEAT PRODUCER . CO Signals (g/kW)CHP Supermarket Data Center Heat Pump225 225 166.1 34.6
B. Results
This section presents the main results and indicators forcomparing the different market designs. All simulations wereperformed for an entire year of market operation.
1) General Results:
Table IV presents the social welfareand the revenue achieved by each agent over the simulatedyear. For the pool market, the achieved results are the same asthe Full P2P, so these are not discussed in detail. As expected,the Full P2P market design is the one presenting the bestsolution, since there are no limitations on heat exchangesbetween agents, opposite to what happens in P2P with productdifferentiation where penalties (consumer preferences) areconsidered. Note that social welfare represents the objectivefunction without penalties, i.e., once the objective is defined,the penalties are removed and all heat transactions are kept.Within the P2P markets, the P2P with distance as productdifferentiation (P2P Distance) is the one achieving the lowestsocial welfare (65,9% compared to Full P2P), since it is theone that most penalizes the transactions between agents. P2PCO2 is the one reaching the social welfare closest to the FullP2P (more than 99.8%). The Full P2P and the community-based are the market models supplying more load, reaching90% of the total load demand. Other models have a smallerdelivery capacity and the minimum is reached for the P2PDistance where only 70% of the entire load demand is met.Although the community performs the poorest social welfare(63% compared to the Full P2P), it is worth stressing thatit is the market that allocates the most load. In terms of heatproduction, the CHP and the data center are the ones producingthe most heat throughout the year. The CHP has the largestthermal energy producing capacity and is the most expensiveresource. Thus, it is often used to cover the remaining energydemand, which other producers cannot cover. On the otherhand, the high dispatch of the data center is related to itshigh nominal capacity and low bid price offered in the market.The CHP shows a drop of about 45 % in production at P2PDistance when compared to the Full P2P, which is linked to
ABLE IVA
GENTS ’ REVENUE BY MARKET DESIGN
Revenue (C)Full P2P P2P Distance P2P Losses P2P CO CommunitySocial Welfare 175250 115560 166422 175040 110407CHP 89328 69179 78094 85115 185057Supermarket 5615 6162 5813 5352 6093Data Center 85090 77614 84670 86931 77452Heat Pump 6610 13413 5338 7007 14113Load 361893 281928 340338 359446 366479
TABLE VA
GENTS ’ DISPATCHED HEAT BY MARKET DESIGN
Dispatched Heat (kW)Full P2P P2P Distance P2P Losses P2P CO CommunityLoad 682941 532850 642078 678188 687215CHP 217191 120623 180546 205486 275674Supermarket 39937 43255 43255 38173 42758Data Center 411472 338954 408897 419155 336219Heat Pump 14341 30018 11522 15372 32564 the fact that it is the producer that is more distant from theconsumers.It is worth mentioning that the heat pump reaches highdispatched heat levels and, consequently, high revenue in theP2P Distance and Community-based markets. The heat pumpis located very close to the consumption points, which helpsto explain the heat pump performance in the market designthat considers the distance between agents. With respect tothe community-based, the heat pump results are related tothe community structure. The heat pump is a member ofCommunity 2, where only the supermarket compete to meetthe demand. As the supermarket behaves as a prosumer, theheat pump or imported heat are often the only available heatsources for that community, leading to a higher market sharefor the heat pump. As the heat pump and the data center are thetwo sources with the lowest CO emissions, these are also theonly agents presenting an increase in the heat supplied (1.8%and 6.7%, respectively), when comparing the P2P CO withthe Full P2P.
2) Average Dispatched Heat and Successful Participationin the Market:
In addition to the general results, two keyperformance indicators (namely, the Average Dispatched Heat(ADH) and the Successful Participation in the Market (SPM)),were introduced. ADH represents the amount of heat that isdispatched from a source on average, i.e., the mean percentageof dispatched heat from the total capacity of the source.The values are presented in percentage (%) and determinedthrough:
ADH ( n ) = (cid:80) Tt =1 P n,t P n,t T , ∀ n ∈ { Ω p } (9)where P n,t represents the heat dispatched by source n in timeperiod t and P n,t represents the maximum capacity of source n in time period t .Regarding the SPM, it indicates the level of participation TABLE VIA
NNUAL AND SEASONAL INDEX OF AVERAGE DISPATCHED HEAT FOREACH HEAT PRODUCER AND MARKET DESIGN . YearCHP Supermarket Data Center Heat PumpFull P2P 72% 97% 62% 25%P2P Distance 71% 100% 51% 64%P2P Losses 71% 98% 61% 14%P2P CO
72% 96% 63% 28%Community 30% 100% 51% 91%SummerCHP Supermarket Data Center Heat PumpFull P2P 84% 97% 48% 1%P2P Distance 83% 100% 29% 36%P2P Losses 87% 98% 48% 1%P2P CO
84% 96% 49% 4%P2P Community 34% 100% 31% 92%WinterCHP Supermarket Data Center Heat PumpFull P2P 60% 98% 76% 50%P2P Distance 58% 100% 73% 92%P2P Losses 54% 98% 76% 28%P2P CO
60% 97% 77% 53%Community 26% 100% 71% 90% by an agent n in the market, which is given by: SP M ( n ) = (cid:80) Tt =1 P articipation ( n,t ) T × , ∀ n ∈ { Ω p } (10)where P articipation n,t is a binary variable indicatingwhether a source n is or not dispatched in the market, in timeframe t .In addition to the annual results, seasonal results are alsopresented, once the sources and loads have seasonal behaviors.As one can see in Table VI, the heat dispatched is generallyhigher in the winter, which is linked to lower external tem-peratures, hence larger levels of heat demand are required.However, the CHP presents lower ADH in the winter whencompared to the summer period. This is connected to thehigher bidding prices offered by this resource in that period ofthe year, which enhances other resources participation in themarket. Also note that the supermarket is the resource withthe highest ADH, being fully dispatched most of the time. Itis also noteworthy that the heat pump is less dispatched in thesummer than in the winter, not only due to the increase of thebid offer, but also due to the lower production capacity duringthis season.Regarding the SPM indicator, the results clearly point toa high successful participation of the supermarket and datacenter in all market designs. When it comes to the data center,these results are justified by its steady heat production and lowoffer price, being one of the first sources that all consumerswant to exchange with. It is important to highlight the contrastexhibited between SPM and ADH in relation to the CHP, sincein the summer there is less heat demand that can be met byother agents with better offers, thus reducing this agent overallparticipation. ABLE VIIA
NNUAL AND SEASONAL INDEX OF SUCCESSFUL PARTICIPATION IN THEMARKET FOR EACH HEAT PRODUCER AND MARKET DESIGN . YearCHP Supermarket Data Center Heat PumpFull P2P 36% 91% 89% 26%P2P Distance 61% 100% 100% 64%P2P Losses 37% 99% 100% 16%P2P CO
35% 88% 90% 28%Community 81% 100% 100% 92%SummerCHP Supermarket Data Center Heat PumpFull P2P 13% 83% 93% 1%P2P Distance 56% 100% 99% 37%P2P Losses 15% 99% 100% 1%P2P CO
12% 75% 95% 4%Community 71% 100% 100% 93%WinterCHP Supermarket Data Center Heat PumpFull P2P 60% 95% 85% 51%P2P Distance 66% 100% 100% 93%P2P Losses 59% 98% 100% 31%P2P CO
59% 93% 86% 54%Community 91% 100% 100% 91%
3) Fairness Indicators:
Fairness indexes are also assessedin this work. The methodology of [60, 61] was followed toevaluate the resource allocation in each market design. Theseindicators are not meant to measure quantities, but rather toassess the relationships between the different agents and theimpact that each of them brings to the whole system. To do so,Quality of Service (QoS), Quality of Experience (QoE) andMin-Max Indicator (MiM) were determined. QoS representshow all the agents impact the heat distribution in the system,i.e., if all involved agents trade the same amount of heat,then the QoS would be equal to 100%. This index assessesthe equilibrium in the system. QoE points out the consumersatisfaction related to the heating price when trading withother agents. The MiM indicator stands for the fairness inthe prosumers and consumers field, where the ratio betweenthe minimum and maximum values for each time period iscalculated. If all the consumers trade the same amount of heat,then this index equals 100%. Table VIII gathers the fairnessindicators results.As one can see, in general, the market modules present aQoS around 20%, meaning that there are agents with largercapacities when compared to other. This discrepancy leadsto lower levels of QoS. When looking at community 2, thisindex is even lower which is related to the heat pump impactin this community. For most of the year, this player is incharge of supplying the whole community, creating a hugeimpact, attracting a large part of the exchange within thecommunity. The QoE, related to the user viewpoint, presentssimilar values for all P2P designs. When analysing the com-munities, these values are substantially lower, due to the fewercompetitiveness existing in each community. Therefore, agentsare compelled to exchange with players who do not offer pricesas favorable as their competitors at certain times, as in the
TABLE VIIIF
AIRNESS INDICATORS FOR EACH MARKET MODEL
QoS QoE MiMFull P2P 21% 78% 4%P2P Distance 17% 83% 4%P2P Losses 21% 79% 5%P2P CO2 20% 79% 4%Community Com 1 Com 2 Com 1 Com 2 Com 1 Com 226% 14% 48% 23% 2% 16%
P2P market models. The low values presented by MiM pointto the significant difference between the heat values that areexchanged among the different agents.
4) Supermarket Individual Analysis:
The supermarket is theonly prosumer in the system, which means that it is the onlyplayer capable of behaving as a producer or consumer in dif-ferent periods of time, being important to analyze its individualtrades with other peers. When the supermarket is behaving as aproducer, it is able to sell heat to the loads. Figure 2 depicts thecumulative heat trade over the year between the supermarketand the loads for each of the considered P2P market designs.More precisely, figure 2 points to a steady supply to allconsumers by the supermarket in the Full P2P design, whichwas expected, since there are no preference constraints for anyheat consumer. On the other hand, the product differentiationeffect is clear in the P2P Distance and P2P Losses, sinceconsumer preferences (namely, distance and losses) encouragetrading with closest peers. Thus, the consumers (C11-C15) arestrongly encouraged to trade with the supermarket, as it isone of the closest producers. In fact, most of the supermarketheat production goes directly to these consumers (about 59.2%and 73.9% for P2P Distance and P2P Losses, respectively),supplying other consumers with residual heat, or not at all.In the P2P considering the CO signals, there are no majorfluctuations once the CO emissions value of the supermarket(225 g/kW) is similar to that of the CHP and Data Center,and much higher than that of the Heat Pump. In this way,the differentiation criterion is minimal relative to the CHPand Data Center with consumers giving priority to trade withthe Heat Pump. More precisely, as both the supermarket andthe Heat Pump have a low capacity to influence the system,the changes in the exchanges between the supermarket andthe consumers are relatively small compared to the Full P2Pmarket design.Notwithstanding, there are periods in which the supermarketdoes not have sufficient self-generation of heat and needsto consume from the DHN, behaving as a consumer in themarket. In this case, Figure 3 depicts the annual percentageof heat supplied by the heat producers to the supermarket. Ingeneral, the supermarket is mainly supplied by CHP and thedata center, since these agents have a large thermal capacity.As the supermarket is closer to the CHP, when consideringthe distance criteria (P2P Distance), the heat supplied by thisresource, reaches its peak. Hence, as the data center is thefarthest resource from the supermarket, the heat exchangereaches its low. The same line of thought is true for the P2P ig. 2. Cumulative annual heat exchange of the supermarket as a heatproducer in the P2P designs. Losses. Conversely, as the heat pump is the resource withthe lowest CO emissions, this resource reaches its maximumwhen considering the P2P CO market design. Fig. 3. Cumulative annual heat exchange of the supermarket as a heatconsumer in the P2P designs.
Looking at the community-based market design (Figure 4),one can see that as a consumer, the supermarket is compelledto import about 80% of the heat, the remaining 20% beingsupplied by the community itself (heat pump). As a heatproducer, all production is shared with the community itself,and no heat is exported.
Fig. 4. Supermarket heat exchange in the Community design.
V. C
ONCLUSION
District heating still has a long way to go, especiallyregarding the way heat is exchanged and the infrastructureneeded for this transformation. Within this scope, new mar-ket models for district heating have been proposed in thiswork, encouraging direct heat exchange between peers. Thenetwork characteristics and impact on heat exchange werealso assessed through product differentiation, giving to the peers and network operators the possibility to define and testcriteria that best fits their interests. All markets designs weresimulated, compared and incorporated in the market moduleof the EMB3Rs platform.The results point to the feasible implementation of this typeof market structure in DHNs. The Full P2P model presents thebest results, since it disregards any limitations of the DHNfor the heat exchanges between the different players. Thiswork, also proves that it is possible to impact the way heatis distributed according to preferences that may be associatedwith distance, minimizing losses or mitigating CO emissions.As an example, analyzing the market design of P2P Distance,one can see that the supermarket can increase by 500% theheat supply to closest consumers when compared to the FullP2P market design. In addition, the Community-based marketdesign also reveals the possibility to divide agents into com-munities, allowing them to manage their own community andexchange heat with other communities, through heat importor export. Overall, if looking at the equilibrium between theagent participation in the market, the quality indicators do notshow a balanced system. This is linked to the different heattechnologies and prices, that change over the year accordingto several factors as the weather. The MiM also highlights thispoint, as a low value for this indicator means a big differencebetween the maximum and minimum heat traded amongst theagents.Future work will focus on full network thermal characteri-zation and comparison with the main findings here presented.Also larger networks will be explored in order to test thesolutions in a real-world like environment.A CKNOWLEDGEMENTS
This work is supported by the European Union’s Horizon2020 through the EU framework Program for Research andInnovation, within the EMB3Rs project under the agreementNo. 847121. In addition, we would like to thank Tiago Sousafor the insight comments that allowed us to improve the paper.D
ATA A VAILABILITY
Datasets related to this article can be found athttp://dx.doi.org/10.17632/ydbcpb73t2.1, an open-source on-line data repository hosted at Mendeley Data (Ant´onio, Tiago,Zenaida, Jos´e, 2020). R
EFERENCES [1] B. V. Mathiesen, N. Bertelsen, N. C. A. Schneider,L. S. Garc´ıa, S. Paardekooper, J. Z. Thellufsen, andS. R. Djørup, “Towards a decarbonised heating andcooling sector in Europe: Unlocking the potentialof energy efficiency and district energy,”
AalborgUniversitet , p. 98, 2019. [Online]. Available: https://vbn.aau.dk/ws/portalfiles/portal/316535596/Towardsa decarbonised H C sector in EU Final Report.pdf[2] U. Persson and S. Werner, “Heat distribution and the fu-ture competitiveness of district heating,”
Applied Energy ,vol. 88, no. 3, pp. 568–576, 2011.3] S. Buffa, M. Cozzini, M. D’Antoni, M. Baratieri, andR. Fedrizzi, “5th generation district heating and coolingsystems: A review of existing cases in Europe,”
Re-newable and Sustainable Energy Reviews , vol. 104, no.October 2018, pp. 504–522, 2019.[4] D. Magnusson, “Who brings the heat? – From municipalto diversified ownership in the Swedish district heatingmarket post-liberalization,”
Energy Research & SocialScience , vol. 22, pp. 198–209, dec 2016.[5] P. Westin and F. Lagergren, “Re-regulating district heat-ing in Sweden,”
Energy Policy , vol. 30, no. 7, pp. 583–596, 2002.[6] P. S¨oderholm and L. W˚arell, “Market opening and thirdparty access in district heating networks,”
Energy Policy ,vol. 39, no. 2, pp. 742–752, 2011.[7] W. Liu, D. Klip, W. Zappa, S. Jelles, G. J. Kramer,and M. van den Broek, “The marginal-cost pricing fora competitive wholesale district heating market: A casestudy in the Netherlands,”
Energy , vol. 189, p. 116367,2019.[8] J. Zhang, B. Ge, and H. Xu, “An equivalent marginalcost-pricing model for the district heating market,”
En-ergy Policy , vol. 63, pp. 1224–1232, 2013.[9] M. Wissner, “Regulation of district-heating systems,”
Utilities Policy , vol. 31, pp. 63–73, 2014.[10] A. Gebremedhin and B. Moshfegh, “Modelling and opti-mization of district heating and industrial energy system- An approach to a locally deregulated heat market,”
International Journal of Energy Research , vol. 28, no. 5,pp. 411–422, 2004.[11] K. Gulzar, S. Sierla, V. Vyatkin, N. Papakonstantinou,P. G. Flikkema, and C. W. Yang, “An auction-based smartdistrict heating grid,”
IEEE International Conference onEmerging Technologies and Factory Automation, ETFA ,pp. 1–8, 2015, Luxembourg.[12] V. B¨urger, J. Steinbach, L. Kranzl, and A. M¨uller, “Thirdparty access to district heating systems - Challenges forthe practical implementation,”
Energy Policy , vol. 132,no. October 2018, pp. 881–892, 2019.[13] M. Marinova, C. Beaudry, A. Taoussi, M. Tr´epanier,and J. Paris, “Economic assessment of rural districtheating by bio-steam supplied by a paper mill in Canada,”
Bulletin of Science, Technology & Society , vol. 28, no. 2,pp. 159–173, 2008.[14] S. Syri, H. M¨akel¨a, S. Rinne, and N. Wirgentius, “Opendistrict heating for Espoo city with marginal cost basedpricing,” , pp. 1–5, 2015, Lisbon, Portugal.[15] L. Brand, A. Calv´en, J. Englund, H. Landersj¨o, andP. Lauenburg, “Smart district heating networks - A sim-ulation study of prosumers’ impact on technical parame-ters in distribution networks,”
Applied Energy , vol. 129,pp. 39–48, 2014.[16] X. Liu and P. Mancarella, “Modelling, assessment andSankey diagrams of integrated electricity-heat-gas net-works in multi-vector district energy systems,”
Applied Energy , vol. 167, pp. 336–352, 2016.[17] J. Huang, Z. Li, and Q. H. Wu, “Coordinated dispatch ofelectric power and district heating networks: A decentral-ized solution using optimality condition decomposition,”
Applied Energy , vol. 206, no. September, pp. 1508–1522,2017.[18] S. Lu, W. Gu, J. Zhou, X. Zhang, and C. Wu, “Coordi-nated dispatch of multi-energy system with district heat-ing network: Modeling and solution strategy,”
Energy ,vol. 152, pp. 358–370, 2018.[19] D. F. Dominkovi´c, M. Wahlroos, S. Syri, and A. S. Ped-ersen, “Influence of different technologies on dynamicpricing in district heating systems: Comparative casestudies,”
Energy , vol. 153, no. March, pp. 136–148, 2018.[20] H. Li, J. Song, Q. Sun, F. Wallin, and Q. Zhang, “Adynamic price model based on levelized cost for districtheating,”
Energy, Ecology and Environment , vol. 4, pp.15–25, 2019.[21] S. Djørup, K. Sperling, S. Nielsen, P. A. Østergaard,J. Z. Thellufsen, P. Sorknæs, H. Lund, and D. Drysdale,“District heating tariffs, economic optimisation and localstrategies during radical technological change,”
Energies ,vol. 13, no. 5, p. 1172, 2020.[22] S. Bhattacharya, V. Chandan, V. Arya, and K. Kar,“Fairness based demand response in DHC networks usingreal time parameter identification,” , pp. 32–37, 2016.[23] A. Paˇz˙erait˙e and M. Krakauskas, “Towards liberalizeddistrict heating market : Kaunas city case,”
Managementof Organizations: Systematic Research
Renewable and Sustainable Energy Reviews , vol. 42, pp.56–65, 2015.[26] I. Moshkin and A. Sauhats, “Solving district heating opti-mization problems in the market conditions,” ,pp. 1–6, 2016, Riga, Latvia.[27] D. Valeriy and K. Dmytro, “Functional structure of thelocal thermal energy market in district heating,” , vol. 1, pp. 343–346,2019.[28] M. Karlsson, A. Gebremedhin, S. Klugman, D. Henning,and B. Moshfegh, “Regional energy system optimization- Potential for a regional heat market,”
Applied Energy
ClimateXChange
StatisticsNorway
Technology Analysis & Strategic Management ,vol. 26, no. 10, pp. 1228–1241, 2014.[34] M. ˚Aberg, L. F¨alting, and A. Forssell, “Is Swedishdistrict heating operating on an integrated market? -Differences in pricing, price convergence, and market-ing strategy between public and private district heatingcompanies,”
Energy Policy , vol. 90, pp. 222–232, 2016.[35] S. Paiho and F. Reda, “Towards next generation districtheating in Finland,”
Renewable and Sustainable EnergyReviews , vol. 65, pp. 915–924, 2016.[36] S. Werner, “International review of district heating andcooling,”
Energy , vol. 137, pp. 617–631, 2017.[37] P. Sorknæs, H. Lund, I. R. Skov, S. Djørup, K. Skytte,P. E. Morthorst, and F. Fausto, “Smart energy markets -future electricity, gas and heating markets,”
Renewableand Sustainable Energy Reviews , vol. 119, no. March, p.109655, 2020.[38] EMB3Rs, “User-driven energy-matching & businessprospection tool for industrial excess heat/cold reduction,recovery and redistribution,” Horizon 2020, Nº 847121.[Online]. Available: https://cordis.europa.eu/project/id/847121[39] E. Sorin, L. Bobo, and P. Pinson, “Consensus-basedapproach to peer-to-peer electricity markets with productdifferentiation,”
IEEE Transactions on Power Systems ,vol. 34, no. 2, pp. 994–1004, 2019.[40] E. W. Dijkstra, “A note on two problems in connexionwith graphs,”
Numerische mathematik , vol. 1, no. 1, pp.269–271, 1959.[41] Y. Cao, W. Wei, L. Wu, S. Mei, M. Shahidehpour,and Z. Li, “Decentralized operation of interdependentpower distribution network and district heating network:a market-driven approach,”
IEEE Transactions on SmartGrid , vol. PP, no. c, p. 1, 2018.[42] J. Bialek, “Tracing the flow of electricity,”
IEE Proceed-ings: Generation, Transmission and Distribution , vol.143, no. 4, pp. 313–320, 1996.[43] T. Sousa, T. Soares, P. Pinson, F. Moret, T. Baroche, andE. Sorin, “Peer-to-peer and community-based markets: A comprehensive review,”
Renewable and SustainableEnergy Reviews , vol. 104, no. January, pp. 367–378,2019.[44] A. S. Faria, T. Soares, J. M. Cunha, andZ. Mour˜ao. (2020) District heating database.http://dx.doi.org/10.17632/ydbcpb73t2.1.[45] H. Gadd and S. Werner, “Achieving low return tempera-tures from district heating substations,”
Applied Energy ,vol. 136, pp. 59–67, 2014.[46] M. Brand, “Heating and domestic hot water systems inbuildings supplied by low-temperature district heating,”
Technical University of Denmark ,pp. 1–9, 2013, Odense, Denmark.[49] S. Fukano, T. Kudo, and N. Arata, “Ammonia heat pumppackage using waste heat as source,”
Energy CO CO refrigeration system in supermarketsto district heating network,” EIB project carbonfootprint methodologies: Methodologies for thessessment of project GHG emissions andemission variations co IEEE Transactions on Power Systems , vol. 34,no. 5, pp. 3994–4004, 2019.[61] R. Jain, D. Chiu, and W. Hawe, “A quantitative measureof fairness and discrimination for resource allocationin shared computer systems,” arXiv: Networking andInternet ArchitecturearXiv: Networking andInternet Architecture