Power Grid Investment Module (PowerGIM)

Created on Tue Aug 16 13:21:21 2016

@author: Martin Kristiansen, Harald Svendsen

class powergama.powergim.SipModel(M_const=1000, CO2price=False)

Power Grid Investment Module - stochastic investment problem

Methods

computeAreaCostBranch(model, c, include_om=False)

Investment cost of single branch

corresponds to firstStageCost in abstract model

computeAreaCostGen(model, c)

compute capital costs for new generator capacity

computeAreaEmissions(model, c, phase, cost=False)

compute total emissions from a load/country

computeAreaPrice(model, area, t, phase)

cumpute the approximate area price based on max marginal cost

computeAreaRES(model, j, phase, shareof)

compute renewable share of demand or total generation capacity

computeAreaWelfare(model, c, t, phase)

compute social welfare for a given area and time step

Returns: Welfare, ProducerSurplus, ConsumerSurplus,

CongestionRent, IMport, eXport

computeCostBranch(model, b, include_om=False)

Investment cost of single branch

corresponds to firstStageCost in abstract model

computeCostGenerator(model, g, include_om=False)

Investment cost of generator

computeCostNode(model, n, include_om=False)

Investment cost of single node

corresponds to cost in abstract model

computeDemand(model, c, t)

compute demand at specified load ant time

computeGenerationCost(model, g, phase)

compute NPV cost of generation (+ curtailment and CO2 emissions)

This corresponds to secondStageCost in abstract model

createConcreteModel(dict_data)

Create Concrete Pyomo model for PowerGIM

Parameters:

dict_data : dictionary

dictionary containing the model data. This can be created with the createModelData(...) method

Returns:

Concrete pyomo model

createModelData(grid_data, datafile, maxNewBranchNum, maxNewBranchCap)

Create model data in dictionary format

Parameters:

grid_data : powergama.GridData object

contains grid model

datafile : string

name of XML file containing additional parameters

maxNewBranchNum : int

upper limit on parallel branches to consider (e.g. 10)

maxNewBranchCap : float (MW)

upper limit on new capacity to consider (e.g. 10000)

Returns:

dictionary with pyomo data (in pyomo format)

createScenarioTreeModel(num_scenarios)

Generate model instance with data. Alternative to .dat files

Parameters:

num_scenarios : int

number of scenarios. Each with the same probability

Returns:

PySP scenario tree model

This method may be called by “pysp_scenario_tree_model_callback()” in

the model input file instead of using input .dat files

extractResultingGridData(grid_data, model=None, file_ph=None, stage=1, scenario=None)

Extract resulting optimal grid layout from simulation results

Parameters:

grid_data : powergama.GridData

grid data class

model : Pyomo model

concrete instance of optimisation model containing det. results

file_ph : string

CSV file containing results from stochastic solution

stage : int

Which stage to extract data for (1 or 2). 1: only stage one investments included (default) 2: both stage one and stage two investments included

scenario : int

which stage 2 scenario to get data for (only relevant when stage=2)

Use either model or file_ph parameter

Returns:

GridData object reflecting optimal solution

saveDeterministicResults(model, excel_file)

export results to excel file

Parameters:

model : Pyomo model

concrete instance of optimisation model

excel_file : string

name of Excel file to create

writeStochasticProblem(path, dict_data)

create input files for solving stochastic problem

Parameters:

path : string

Where to put generated files

dict_data : dictionary

Pyomo data model in dictionary format. Output from createModelData method

Returns:

string that can be written to .dat file (reference model data)

powergama.powergim.annuityfactor(rate, years)

Net present value factor for fixed payments per year at fixed rate

powergama.powergim.sampleProfileData(data, samplesize, sampling_method)

Sample data from full-year time series

Parameters:

X : matrix

data matrix to sample from

samplesize : int

size of sample

sampling_method : str

‘kmeans’, ‘lhs’, ‘uniform’, (‘mmatching’, ‘meanshift’)

Returns:

reduced data matrix according to sample size and method

powergama.powergim.sample_kmeans(X, samplesize)

K-means sampling

Parameters:

X : matrix

data matrix to sample from

samplesize : int

size of sample

This method relies on sklearn.cluster.KMeans

powergama.powergim.sample_latinhypercube(X, samplesize)

Latin hypercube sampling

Parameters:

X : matrix

data matrix to sample from

samplesize : int

size of sample

This method relies on pyDOE.lhs(n, [samples, criterion, iterations])

powergama.powergim.sample_meanshift(X, samplesize)

M matching sampling

Parameters:

X : matrix

data matrix to sample from

samplesize : int

size of sample

This method relies on sklearn.cluster.MeanShift

It is a centroid-based algorithm, which works by updating candidates

for centroids to be the mean of the points within a given region.

These candidates are then filtered in a post-processing stage to

eliminate near-duplicates to form the final set of centroids.

powergama.powergim.sample_mmatching(X, samplesize)

The idea is to make e.g. 10000 randomsample-sets of size=samplesize from the originial datasat X.

Choose the sampleset with the lowest objective: MINIMIZE [(meanSample - meanX)^2 + (stdvSample - stdvX)^2...]

in terms of stitistical measures