1 ΗΜΕΡΙΔΑ ΣΤΟ ΠΛΑΙΣΙΟ ΤΟΥ ΕΡΕΥΝΗΤΙΚΟΥ ΠΡΟΓΡΑΜΜΑΤΟΣ ENDECON (ΘΑΛΗΣ) Microgrids The Building Blocks of Smartgrids Prof. Nikos Hatziargyriou National Technical University of Athens, Greece
2 Microgrids are electricity distribution systems containing loads and distributed energy resources, (such as distributed generators, storage devices, or controllable loads) that can be operated in a controlled, coordinated way, either while connected to the main power network and/or while islanded. Microgrids (CIGRE WG C6.22) EU Microgrids (ENK5-CT ) and MOREMICROGRIDS (PL019864)
3 Benefits by Criteria & Recipient Economic Benefits Consumer Local Market Value Aggregation Platform Value DSO Micro- Source Network Hedging Value Peak Load Shaving Voltage Regulation Energy Loss Reduction Reliability Enhacement Technical Benefits GHG Reduction Environmental Benefits Identification of Microgrid benefits is a multi-objective and multi-party coordination task
4 Who will develop a Microgrid? Who will own or operate it? Investments in a Microgrid can be done in multiple phases by different interest groups: DSO, energy supplier, end consumer, IPP (individual power producer), etc. The operation of the Microgrid will be mainly determined by the ownership and roles of the various stakeholders. Three general models: DSO owns and operates the distribution grid and also fulfils the retailer function of selling electricity to end consumers. (DSO Monopoly) ESCO are the actors that maximize the value of the aggregated DG participation in local liberalized energy markets (Liberalized Market) Consumer(s) own and operate DG to minimize electricity bills or maximize revenues (Prosumer Consortium)
5 Technical Challenges Use of different generation technologies (prime movers) Presence of power electronic interfaces Small size (challenging management) Relatively large imbalances between load and generation to be managed (significant load participation required, need for new technologies, review of the boundaries of microgrids) Specific network characteristics (strong interaction between active and reactive power, control and market implications) Protection and Safety / static switch Communication requirements
6 Market Challenges coordinated, but decentralised energy trading and management market mechanisms to ensure efficient, fair and secure supply and demand balancing development of islanded and interconnected price-based energy and ancillary services arrangements for congestion management secure and open access to the network and efficient allocation of network costs alternative ownership structures, energy service providers new roles and responsibilities of supply company, distribution company, and consumer/customer
7 Microgrids Hierarchical Control MicroGrid Central Controller (MGCC) promotes technical and economical operation, interface with loads and micro sources and DMS; provides set points or supervises LC and MC; MC and LC Controllers: interfaces to control interruptible loads and micro sources Centralized vs. Decentralized Control
8 Centralized & Decentralized Control The main distinction is where decisions are taken Centralized Control implies that a Central Processing Unit collects all the measurement and decides next actions. Decentralized Control implies that advanced controllers are installed at each node forming a distributed control system. Choice of approach depends on DG ownership, scale, plug and play, etc.
9 Centralized Control Participation in Energy Markets Prosumer/Consortium Model Microgrid serving its own needs using its local production, when financially beneficial. MGCC minimises operation costs based on: Prices in the open power market Forecasted demand and renewable power production Bids of the Microgrid producers and consumers. Technical constraints IEEE PES General Meeting July 2011, Detroit, Michigan USA, Microgrid Control Panel Session
10 Liberalized Energy Model Microgrid buys and sells power to the grid via an Energy Service provider MGCC maximizes value of the Microgrid, i.e. maximizes revenues by exchanging power with the grid based on similar inputs based on: The market prices for buying and selling energy to the grid (same prices for end-users of the Microgrid) Demand and renewable production forecasting The offers of the DG Centralized Control Participation in Energy Market The technical constraints for interconnection and the DG
11 CIGRE Benchmark: LV network with 20 kv multiple feeders Off-load TC kv in 5 steps 0.4 kv 3+N 20/0.4 kv, 50 Hz, 400 kva u k =4%, r k =1%, Dyn11 3 Ω Single residencial consumer 3Φ, I s =40 A S max =15 kva S 0 =5.7 kva Residential load 3x250 A 80 Ω 80 Ω 3+N Overhead line 4x120 mm 2 Al XLPE twisted cable Pole-to-pole distance = 35 m 80 Ω 80 Ω 80 Ω Twisted Cable 3x70mm2 Al XLPE mm 2 AAAC Possible neutral bridge to adjacent LV network 2 Ω 3x160 A 3+N 200 m Underground line 3x150 mm 2 Al + 50 mm 2 Cu XLPE cable 3Φ, I s =40 A S max =20 kva S 0 =11 kva 80 Ω 1Φ, I s =40 A Phase: a S max =8 kva S 0 =4.4 kva 3x160 A 4x35 mm2 Al conductors 80 Ω 4x1Φ, I s =40 A Phase: abcc S max =25 kva S 0 =13.8 kva 3+N Overhead line 4x50,35,16 mm 2 Al conductors Pole-to-pole distance = 30 m 3Φ, I s =63 A S max =30 kva S 0 =16.5 kva 80 Ω 4x16 mm2 Al conductors 4x50 mm2 Al conductors Group of 4 residences 4 x 3Φ, I s =40 A S max =55 kva S 0 =25 kva 80 Ω Appartment building 5 x 3Φ, I s =40 A 8 x 1Φ, I s =40 A S max =72 kva S 0 =57 kva Single residencial consumer 3Φ, I s =40 A S max =15 kva S 0 =5.7 kva Industrial load 3x1Φ, I Appartment building s =40 A 80 Ω 80 Ω Phase: abc 1 x 3Φ, I s =40 A S max =20 kva 6 x 1Φ, I s =40 A S 0 =11 kva S max =47 kva S 0 =25 kva 80 Ω 50 Ω Workshop 3Φ, I s =160 A S max =70 kva S 0 =70 kva Commercial load 4x1Φ, I s =40 A Phase: abbc S max =25 kva S 0 =13.8 kva 80 Ω 80 Ω 80 Ω 80 Ω 2x1Φ, I s =40 A Phase: ab S max =16 kva S 0 =8.8 kva 1Φ, I s =40 A Phase: c S max =8 kva S 0 =4.4 kva 4x35 mm2 Al conductors
12 CIGRE Benchmark Off-load TC kv in 5 steps 0.4 kv 20 kv 3+N 20/0.4 kv, 50 Hz, 400 kva u k =4%, r k =1%, Dyn11 3 Ω Study Case LV Feeder with DG Flywheel storage Rating to be determined Single residencial consumer 3Φ, I s =40 A S max =15 kva S 0 =5.7 kva Group of 4 residences 4 x 3Φ, I s =40 A S max =50 kva S 0 =23 kva Wind Turbine ~ 3Φ, 15 kw Photovoltaics 1Φ, 4x2.5 kw Fuel Cell 3Φ, 30 kw ~ 3+N+PE Circuit Breaker Possible sectionalizing CB Appartment building 1 x 3Φ, I s =40 A 6 x 1Φ, I s =40 A S max =47 kva S 0 =25 kva 3+N+PE 3+N+PE 1+N+PE Circuit Breaker instead of fuses 3+N+PE 80 Ω 10 Ω 30 Ω 30 Ω 4x6 mm 2 Cu 20 m 80 Ω 4x16 mm 2 Cu 30 m 4x25 mm 2 Cu 20 m 80 Ω 80 Ω 4x16 mm 2 Cu 30 m 80 Ω 3+N Overhead line 4x120 mm 2 Al XLPE twisted cable Pole-to-pole distance = 35 m 80 Ω 80 Ω 80 Ω 3x70mm2 Al XLPE mm 2 AAAC Twisted Cable 10 Ω 80 Ω 3+N+PE 4x6 mm 2 Cu 20 m 80 Ω 3+N+PE Possible neutral bridge to adjacent LV network 2 Ω 30 m 3x50 mm 2 Al +35mm 2 Cu XLPE Appartment building 5 x 3Φ, I s =40 A 8 x 1Φ, I s =40 A S max =72 kva S 0 =57 kva ~ ~ Other lines ~ Single residencial consumer 3Φ, I s =40 A S max =15 kva S 0 =5.7 kva Photovoltaics 1Φ, 3 kw Microturbine 3Φ, 30 kw
13 Simulation Tool of MG Market 1 3 Participation Security level
14 Residential Feeder with DGs Prosumer Consortium: Cost Reduction % 27% reduction in CO 2 emissions Liberalized Market: Cost reduction 18.66% kw Load & Power exchange with the grid (residential feeder) Hour Pow er exchanged w ith the grid Load Pattern
15 Demand Side Bidding Two Demand Side Bidding options Consumers bid for supply of high and low priority loads. The total demand is found by the aggregation of DSBs and DG bids Option A Consumers offer to shed low and high priority loads at fixed prices according to their bids and paid for this service. The total demand is assumed known beforehand Option B A typical demand shed bid formulation Price (Ect/kWh) Power (kw) 3
16 DSB Results Single day Total energy shed 232 kwh (7.27%) Total cost reduction 34.79% for option A Cost reduction 31.44% for option B. Reducing average prices for the Microgrid consumers 16 Average energy prices Price (Ect/kWh) Hour Market price Microgrid Average price Microgrid Average non-critical shed
17 Decentralized Control - MultiAgent System for Microgrids Autonomous Local Controllers Distributed Intelligence Reduced communication needs Open Architecture, Plug n Play operation DNO MO FIPA organization Java Based Platforms Agent Communication Language Grid Level Agent... Management Level Microgrid Microgrid Agent Agent Agent Microgrid MGCC LC LC LC LC... Field Level Agent Agent Agent Agent
18 The Agent Physical entity that acts in the environment or a virtual one Reactive Cognitive or Intelligent Partial representation of the environment partial representation of the environment Memory Agents communicate cooperate with each other autonomy Environment Perception Agents have a certain level of autonomy possesses skills High level communication The agents have a behaviour and tends to satisfy objectives using its resources, skills and services Typical example an ant colony Typical example the human society
19 Agent Communication A significant characteristic of agents. The Agent Communication Language allows the interaction and the knowledge sharing. One significant part of the agent communication is the auction algorithm According to the fundamental principles of economic theory, fair bidding leads to optimal solutions The auction algorithm is an important tool for agent applications.
20 Behaviour, objectives, resources, skills and services Behavior Objectives Resources Skills Services Competitive Collaboration Maximize profit Minimize cost Available Fuel Energy Stored in a Battery Load Curtailment Black Start Yellow pages Data Storage
21 Agents Implementation with Java-Jade Jade is a Java based platform for agent implementation. It is compatible with FIPA requirements FIPA is the Foundation for Intelligent Physical Agents Jade provides a set of libraries that allow the implementation of the agents.
22 Model of agent platform Agent Platform AGENT AGENT MANAGEMENT SYSTEM DIRECTORY FACILITATOR Provides Yellow Pages Services MESSAGE TRANSPORT SYSTEM
23 Implementation The implementation of the system requires a network of computers capable to host Java Applications and communicate over TCP/IP (internet protocols) Main Host DF AMS Agent Agent Agent TCP/IP Computer Agent Agent Agent Computer Agent Agent Agent Computer Agent Agent Agent
24 MICROGRIDS Project (FP5) Large Scale Integration of Micro-Generation to Low Voltage Grids, PROJECT N : NNE GREAT BRITAIN UMIST URENCO 14 PARTNERS, GREECE PORTUGAL EDP INESC SPAIN LABEIN 7 EU COUNTRIES UMIST URENCO ARMINES EDF SMA ISET SMA ISET NTUA PPC /NAMD&RESD GERMANOS GERMANY NETHERLANDS EMforce INESC EDP LABEIN CENERG ICCS / NTUA GERMANOS FRANCE EDF Ecole des Mines de Paris/ARMINES CENERG PPC/NAMD&RESD Budget: 4.5M
25 MORE MICROGRIDS Advanced Architectures and Control Concepts for more Microgrids, Contract : SES6-PL Budget: 8M
26 Microgrids Pilots Østkraft AGRIA PIG FARM 26
27 Pilot Microgrid in Kythnos Supply of 12 buildings (EC projects MORE, PV-Mode, More Microgrids)
28 Advanced Sunny Island inverters, to deal with islanded mode control Intelligent Load Controllers Settlement of 12 houses Generation: 5 PV units connected via standard grid-tied inverters. A 9 kva diesel genset (for back-up). Storage: Battery (60 Volt, 52 kwh) through 3 bidirectional inverters operating in parallel. Flexible Loads: 1-2 kw irrigation pumps in each house
29 The Kythnos System House
30 Goals of Kythnos Experiment The goal of the experiment is to test the agent based control system in a real test site in order to increase energy efficiency. The main objective is to test the technical challenges of the MAS implementation. The technical implementation is based on intelligent load controllers and the Jade Platform The algorithm regarding the increase of the energy efficiency is quite simple and focuses in the limitation of the pump operation.
31 The MAS tries to increase the energy efficiency. The steps are the following: 1.The system decides the available energy that can be used by the pumps. 2.The houses decide how to share this energy.
32 Step 1: The agents identify the status of the environment Step 2: The agents negotiate on how the share the available energy PV Diesel Batteries System House Wi-Fi
33 Intelligent Load Controllers In each house an ILC is installed: Windows CE 5.0 Intel XscaleTM PXA255 64MB of RAM 32MB FLASH Memory Java VM Jade LEAP
34 Negotiation Battery MGCC Start Negotiation. The MGCC orders the system to start a new cycle. This can be done in variable steps (5min- 1hour) PV House A House B House E House C House D
35 Battery Negotiation MGCC PV agent Announces Production PV House A House B House E House C House D
36 Negotiation Battery MGCC Battery agent announces Production & SOC. The estimation of the available energy can be done using different methods (level of SOC, frequency, etc ) PV House A House B House E House C House D
37 Negotiation Battery MGCC Agents Start Negotiating. The simple algorithm suggests that agents should consume equally. Therefore the one with the higher consumption has the lower priority. PV House A House B House E House C House D
38 One significant part of the agent communication and decision process is the auction algorithms. The auction algorithm is a tool that allows the agents to decide which one of them will acquire a certain object or a good. In this experiment the simple algorithm called English Auctions is implemented
39 In the English Auction the auctioneer seeks to find the market price of a good by initially proposing a price below that of the supposed market value and then gradually raising the price. Each time the price is announced, the auctioneer waits to see if any buyers will signal their willingness to pay the proposed price. As soon as one buyer indicates that it will accept the price, the auctioneer issues a new call for bids with an incremented price. The auction continues until no buyers are prepared to pay the proposed price, at which point the auction ends. If the last price that was accepted by a buyer exceeds the auctioneer's (privately known) reservation price, the good is sold to that buyer for the agreed price. If the last accepted price is less than the reservation price, the good is not sold
40 The shedding procedures start later In this case the frequency if almost 52Hz. This is an indication that the batteries are full and the PV inverters via the droop curves limit their production.
41 One critical part of any implementation MAS implementation is the ontology. In Kythnos CIM (IEC 61970) based ontology was tested.
42 Highlights Control Implementation of Distributed Control The houses decide their actions Software Java Based Application FIPA Compliant Architecture CIM Based ontology Hardware Embedded Controller Communication & Control Capabilities
43 The Kythnos was the first test site where the MAS system was implemented A Controller with an Embedded processor has been used to host the agents. New techniques have been tested such as: negotiation algorithms, wireless communication, CIM based ontology etc.. The architecture is too complex for small systems but offers great scalability.
44 Non-technical lessons learned MAS for energy optimization provides a technical limitation and protection of the system to prevent over-use. This helps to maintain the good relationships between the neighbours. Importance of involving or at least explaining to users negotiation process to equally share the available energy - development of demonstration software The technical and economical aspects of system operation are evaluated positively: the system works quite reliably, users pay regularly, the maintenance and repairs of the system are well organized.
45 It works!!!
46 Purely Decentralized Economic Dispatch
47 Algorithmic Description Information flow 1 p1 p2 2 3 p3 4 p4 p5 p6 5 6
48 Gossip Protocols The fundamental concept behind gossip algorithms is that every participating node interacts with one or more nodes and exchanges a piece of information Information is anticipated to be disseminated rapidly like a virus (another name for gossip protocols is epidemic algorithms) Appealing characteristics: fault-tolerant & scalable Fully capable of distributed calculation of separable functions
49 Practical Applications Power grid state estimation by distributed calculation of electric quantities Distributed optimization of grid operation (congestion management, losses reduction, voltage deviations) gaining access only to locally available information The wide deployment of smart meters with enhanced capabilities (bidirectional communication, information storage and processing) creates the infrastructure for such applications
51 SmartHouse/SmartGrid Project The goal was to demonstrate how ICT-enabled collaborative aggregations of Smart Houses can achieve maximum energy efficiency Customer-interactive in-house technology for energy management Demand side: real-time information and dynamic tariffs Customer as prosumer: generation within the house can be integrated into the system Interaction with the Smart Grid Distributed control in a decentralized energy world Intelligent agent-based control Web services at the device level and at higher system levels Prof. Nikos Hatziargyriou, Cardiff, September 2013
52 SH/SG Market Actors Producer Transmission system operator Distribution system operator Supplier / retailer Final customer Transactions within the electricity market Prof. Nikos Hatziargyriou, Cardiff, September 2013
53 SmartHouse/SmartGrids (SH/SG) Business Cases Energy Usage Monitoring and Feedback Real-time Portfolio Imbalance Reduction Offering (secondary) Reserve Capacity to the TSO Distribution System Congestion Management Distribution Grid Cell Islanding in case of Higher- System Instability Black-Start Support from Smart Houses Integration of Forecasting Techniques Prof. Nikos Hatziargyriou, Cardiff, September 2013
54 MAS Coordination Approach CO2 Market Ancillary Services Bilateral Contract Spot Market Wholesale Market ESCO DSO Network Tools DNO DMS Economic Management MGCC Meltemi MGCC Network Management MGCC DG House DG House DG House Substation 1 Substation 2 Substation N Prof. Nikos Hatziargyriou, Cardiff, September 2013
55 Field Test Diagram Diesel PV PC Monitor & Control Monitor Jade Platform Equipment Equipment Load Controller Load Controller Load Controller LAN DSO (MGCC) Retailer, ESCo Prof. Nikos Hatziargyriou, Cardiff, September 2013
56 Actual Network Architecture Prof. Nikos Hatziargyriou, Cardiff, September 2013
57 Meltemi Testbed Prof. Nikos Hatziargyriou, Cardiff, September 2013
58 Pictures from Meltemi Test Site House 1 House 3 House 4 House 6 House 5 Main House 7 House 8 PV Generator House 2 House 10 House 9 House 11
59 Recordings PV / Total Consumption Γενικός τύπος ActivePower -Parking PV Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Electricity consumption in "Meltemi" Γενικός τύπος Γενικός τύπος Power (MW) Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός Γενικός Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος Γενικός τύπος τύπος τύπος Hour
60 The Operation of the Load Controller Prof. Nikos Hatziargyriou, Cardiff, September 2013
61 Recordings from Consumers The consumer left the water heater on during the whole morning
62 Forecasting Results The technical integration of forecasting technologies is feasible The forecasting errors are very high both for RES and load due to the small size of the settlement It is realistic to have forecast at the substation level Expected consumption per device during a summer day Average total consumption per house
63 Congestion Management Solution Approach Key question: How much energy can be used per house? Energy consumption per day/per house (actual measurements) Simple algorithm MAGIC algorithm Knowledge of the consumer behaviour
64 Congestion Management Time to reduce transformer overload This is feasible using the forecasting modules The system knows that an incident is imminent to happen. 15 min warning is sufficient Energy shed during the congestion The system sheds energy from controllable appliances not the whole household or group of houses ~25% of energy from controllable appliances can be shifted