220 likes | 302 Views
Corridor Method heuristics to solve rural electrification projects. Joan Triadó Aymerich , Dra. Laia Ferrer Martí, Dr. Alberto García Villoria , Dr. Rafael Pastor Moreno Universitat Politècnica de Catalunya- Barcelona Tech. Institut d’Organització i Control.
E N D
Corridor Method heuristics to solve rural electrification projects Joan Triadó Aymerich, Dra. Laia Ferrer Martí, Dr. Alberto García Villoria, Dr. Rafael Pastor Moreno Universitat Politècnica de Catalunya- Barcelona Tech. Institut d’Organització i Control Ministerio de Ciencia e Innovación de España, proyecto ENE2010-15509, cofinanciado por FEDER. With the support of: 2013 ICSO-HAROSA International Workshop on Simulation-Optimization & Internet Computing BARCELONA 10-12 July 2013
Summary Introduction Technical considerations Objectives MILP model Heuristics based on Corridor method Instances Analysis of results Conclusions
1. Introducción • Currently, more than 1.3 billion people have no access to electricity (IEA 2011) • Especially in rural areas in developing countries • Electrification systems using renewable energies are a suitable solution to provide electricity to communities independently. • They use local resources. • At a lower cost than extending the grid. • Hybrid systems (wind and solar) are one of the possible techniques. • Recently, hybrid systems have been implemented in Cajamarca. 3
Types of micro wind turbines Types of consumption points 1. Introduction • Due to the dispersion between housings, the trend in electrification projects is to use individual systems. • This work considers the design of projects of electricity supply that combinemicro-grids and isolated points generation, considering the resource (wind and sun) at each point. • No consumption is limited in a point to available resource in your housing. • Money can be saved using large equipment. • May be more easily adaptable to increases in consumption. 4 Types of solar panels
2. Technical considerations Wind turbine Consumption point Meter Consumption point Meter Consumption pointMeter Battery Inverter Consumption pointMeter Solar Panel Consumption pointMeter Solar controller • Wind turbine, power generated depends on the type and location. • Solar panel, power generated depends on the type. • Solar controller, controls the charge / discharge of the battery. • Battery,considers capacity and autonomy demand. • Inverter, DC/AC conversion to the nominal distribution voltage. • Meter, measures the energy consumed at the point of microgrid. • Wire, used to distribute energy in the form of radial microgrid. 5
3. Objectives • Designing an autonomous power system with renewable energy (solar and wind). • Determining • the location, number and type of elements forming the system. • the distribution microgrids to use • Considering • the situation and demand of consumption points • the energy resources (solar and wind above all) at each point • At a minimum investment cost. • Developing heuristics based on linear programming to solve the design.
4. MILP model • Variables: xp: Existence of generation at p (binary) Xapa : No. of wind turbines at p of type a xbpb: No. of batteries at p of type b xipi: No. of inverters at p of type i xrspz: No. of solar controllers at p of type z xcpdc: Existence of wire connecting p and d of type c (binary) xccpd: Existence of wire connecting p and d (binary) fepd: Energy flow from p to d fppd: Power flow from p to d vp : Voltage at p xmd : Existence of a meter at d (binary)
4. MILP model • Objective function (investment cost): • Constraints wind turbines wires batteries solar controllers inverters meters solar panels - Demand - Batteries - Voltage drops
5. Heuristics based onCorridor Method • The Corridor method (Sniedovich and Vos, 2006) is based on the idea of using an exact method on restricted portions of the solution space of the given problem. The method is an iterative improvement result of exploring neighborhoods. It is characterized by the way in which the neighborhood is defined. • The neighborhood is generated from the previous values of decision variables, xj, of the model. The constraints that define neighborhoods are: • xj : variable values confined in the neighborhood • j: index of the set of variables which define the neighborhood • RZ+: parameter that determines the size of the neighborhood.
Corridor constraints used in the project 5. Heuristics based onCorridor Method xccpd: existence of wire connecting p and d (binary). RZ+is a parameter that determines the size of the neighborhood. nc: number of wires of the incumbent solution. MX: set of all pairs of points (p, d) of the incumbent solution that are interconnected by a wire, p is the source and d is the destination.
Constraints that deny the possibility of connection between points outside MX. Constraints added to accelerate the computing and to allow the achieving of a local optimal solution 5. Heuristics based onCorridor Method xccmn: existence of wire connecting m and n (binary). X: set of points p of the incumbent solution that are connected with another point. MX: set of all pairs of points (p, d) of the incumbent solution that are interconnected by a wire, p is the source and d is the destination.
Constraints that maintain the basic structure of the microgrids These constraints do not allow the change of direction of the connection, and so the basic structure of microgrids remains intact. On the other hand the generation points may be the same as in the incumbent solution or may change to some new added points. Constraints added to accelerate the computing and to allow the achieving of a local optimal solution 5. Heuristics based onCorridor Method (p,d)MX xccdp: existence of wire connecting d and p (binary) MX: set of all pairs of points (p, d) of the incumbent solution that are interconnected by a wire, p is the source and d is the destination.
Additional constraints that limit the distance of connection (optional). Constraints added to accelerate the computing and to allow the achieving of a local optimal solution Option 1: fixed limit distance Option 2: equivalent limit distance 5. Heuristics based onCorridor Method p, d p, d xcpdc: existence of wire connecting p and d of type c (binary) L[p,d]: distance from p to d. dmax: distance limit allowed for connection dmax_eq: equivalent distance is worth to connect two points
Additional constraints that limit cost (optional). Constraints added to accelerate the computing and to allow the achieving of a local optimal solution 5. Heuristics based onCorridor Method
Algorithm Stage 1 Proposed relaxation: xapa, xsps, xp: not relaxed given that obtained solutions tend to determine fractions of generators avoiding the creation of microgrids. Xccpdnot relaxed, for an excessive number of connections are created and all the points are supplied by a single point of generation. xbpb, xcpdc, xipi, xrspr and xmd: relaxed variables. Stage 2. Taking advantage of the initial solution obtained from stage 1 Addition of new constraints Corridor constraints Constraints that deny the possibility of connection between points outside MX Constraints that maintain the basic structure of the microgrids Local neighborhood searching 5. Heuristics based onCorridor Method
6. Instances • Characteristics of instances • Randomly generated based on real project in El Alumbre (Perú): • Number of consumption points: 10, 20, 30, 40, 50, 60, 70, 80 and 90. • Consumption points concentrations: • Low density; 25% of consumption points concentrated in 20% of area. • High density; 50% of consumption points concentrated in 20% of area. • Energy and power demand by consumption point: • 280Wh/day and 200W respectively. • 420Wh/day and 300W respectively. • Wind potential: • Normal (as wind map). • High (multiplying by 1,5 the wind map values). • For each combination of the 4 parameters (number of points of consumption, consumption points concentration, energy and power demand and wind potential) were generated 5 instances: Total 9 x 2 x 2 x 2 x 5 = 360 instances
7. Analysis of results • Heuristic evaluation comparing with H0 (MILP model resolution for 1 hour). • Characteristics of heuristics implemented CM1, CM2 and CM3: Different heuristics based on Corridor method Trm: Computing limit of time of relaxed model (stage 1) GAPrm: GAP of relaxed model (stage 1) R: number of connection changes allowed between iterations (stage 2) Ti: Computing limit of time for stage 2 GAPi: GAP of stage 2 dmax: maximum distance of connection allowed
7. Analysis of results Results at the end of stage 2
7. Analysis of results Results at the end of stage 2
7. Analysis of results • CM1 and CM3 give better results than H0 • The best results were achieved for CM1: 6.08% of average improvement comparing with H0 for instances of 70 consumption points
8. Conclusions • In this paper we present a heuristics based on MILP design for rural electrification systems. • The MILP heuristic provide the location of the generators and the design of possible microgrid considering wind and solar resource • There has been a computing experience over 360 instances generated based on a real project • The results have been satisfactory for the heuristic studied
Thank you for your attentionCorridor Method heuristics to solve rural electrification projects Joan Triadó Aymerich, Dra. Laia Ferrer Martí, Dr. Alberto García Villoria, Dr. Rafael Pastor Moreno Universitat Politècnica de Catalunya- Barcelona Tech. Institut d’Organització i Control Ministerio de Ciencia e Innovación de España, proyecto ENE2010-15509, cofinanciado por FEDER. With the support of: 2013 ICSO-HAROSA International Workshop on Simulation-Optimization & Internet Computing BARCELONA 10-12 July 2013