Hack from Home 2021, Hosted by DataSwyft
Note
This project concluded with the P2P-Eex Team winning the "Best Pitch" award, validating the uniqueness and commercial potential of such a project.
The P2P Energy Exchange is aiming to revolutionize and democratize energy markets. Our innovative platform integrates individual nodes (households, producers, buildings or, regions) into a decentralized energy market. To deploy our decentralization system, we are offering automated intra-node energy flow optimization. This demo is a proof of concept for this said optimization. We use proprietary algorithms to best allocate energy resources within a given context.
Read more on our website: https://hackathon-from-sfsdf.vercel.app/
In our Proof of Concept (POC) scenario, we examine a hypothetical green development project located in Amsterdam and comprising four homes. This POC is grounded in the needs and developments in the housing sector in Northern Europe and comprises a valid test case. As such, we are simulating a complex of four (4) homes located in Amsterdam and equipped with:
-
Solar Panels: 72 250W panels (
$115m^{2}$ ) @ 15% efficiency (18Kwh max per day) - Battery Pack: 43Kwh w. 70% discharge limit and max 2Kwh daily charging rate.
- Grid Connection: Stable grid connection at the same price as what the solar power can be sold for.
- Solar Irradiance Data: provided via NASA Power API for Prediction of Worldwide Energy Resources.
- Placeholder Solar Panel Data: provided via Greenmatch.co.uk.
- The flexible prosumer: Measuring the willingness to co-create distributed flexibility, Kubli et. al., 2018.
- Time Series and Renewable Energy Forecasting. Time Series Analysis and Applications, Ghofrani et. al., 2018.
The params.json file which is specified as a parameter for the calc_algo_results() function, contains all the necessary
inputs for the model in a user-friendly .json file.
{ "price" : <- Characteristics of the Electricity Price Timeseries
{
"mu" :0.00002, <- Generator mu
"sigma" :0.0003, <- Generator sigma
"start" :0.1437 <- Starting value
},
"demand" : { <- Characteristsics of the Energy Demand Timeseries
"mu" : 0.00006, <- Generator mu
"sigma" : 0.008, <- Generator sigma
"start" : 16 <- Starting value (4Kw * 4 houses = 16)
},
"test_case" : { <- Test case characteristics
"num_solar" : 72, <- Number of solar panels
"battery_capacity" : 42 <- Battery capacity in Kwh
},
"algo" : { <- Algorithm parameters
"random_seed" : 0, <- Seed
"rolling_window" : 15, <- Pricing rolling window for SMA
"critical_limiter" : 0.7, <- Battery Discharge Limit
"battery_start" : 0, <- Battery Starting Value
"defecit_fill_grid" : 0.6, <- Share of Defecit Energy Requested via Grid (60%)
"defecit_fill_battery" : 0.4, <- Share of Defecit Energy Requested via Battery (40%)
"battery_refil_price_multiple" : 1.15, <- Multiple of energy to buy from grid when energy is cheap
"battery_refil_multiple" : 0.2, <- Multiple at which energy can be added to the battery
"daily_battery_charge" : 1 <- Daily 'safety' recharge of battery from grid when below limiter
},
"datapane" : { <- specifications for the datapane report, leave as is
"token" : "placeholder",
"name" : "placeholder"
}
}
The panels.json file which is specified as a parameter for the calc_algo_results() function,
contains the information to use for calculating the energy output of the panels. In our case, we are looking at the
"small panel" with 250W @ 15% ef.
{
"small panel" :
{
"Wattage": 250,
"Area" : 1.6,
"Efficiency": 0.15,
"Price" : 300
},
"medium panel" :
{
"Wattage" : 300,
"Area" : 1.67,
"Efficiency" : 0.16,
"Price" : 320
},
"large panel" :
{
"Wattage" : 350,
"Area" : 1.9,
"Efficiency" : 0.17,
"Price" : 400
}
}
The helpers.py file contains two helpful functions which make data retrival and the creation of random timeseries
for the simulator simple.
The random_noise() function is heavily used in the calc_algo_results() function
to create randomized timeseries based on inputs specified in panels.json.
The function returns an array of the random timeseries generated.
Returns: numpy.array object.
random_noise(mu : float, sigma : float, start_price : float)
Example call:
random_noise(0.00002, 0.0003, 0.1437)
The get_data() function is used in the source code to open the files necessary for the model to run and returns them
as objects (dataframes & dictionaries).
Specifically, it opens params.json, panels.json and, solar_data_AMS.csv, at specified in the parameters of
calc_algo_results.
Returns: array of form [daily_data : pd.Dataframe, solar : dict, params : dict]
get_data(csv_filename, json_filename, params_filename)
Example call:
get_data('solar_data_AMS.csv', 'panels.json', 'params.json')
The algo.py file contains a single function and acts as the core of the simulator.
The calc_algo_results() function goes over the input files as retried from get_data() and uses the specifications
and data to create a simulation of the energy system.
Returns: data_frame : pd.Dataframe
calc_algo_results(solar_data='solar_data_AMS.csv', panel_data='panels.json', setup_file='params.json')
Example call:
calc_algo_results('solar_data_AMS.csv', 'panels.json', 'params.json')
The reporter.py file contains the reporter() function which exports the simulation report in .pdf format
as created using Datapane.
The reporter() function exports the simulation report in .pdf format
as created using Datapane.
Returns: True -- Exports .pdf file.
reporter(data_frame, filename='report.html')
Example call:
reporter(calc_algo_results(a, b, c), 'myreport.html')
This simulation was produced for Hack from Home 2021, organized by Dataswyft, using Python 3.8 on Datalore and Visual Studio Code. This report was generated using Datapane.
Website Development & Pitch Deck by: @biomathcode
Technical Consulting & Code QA by: Taoheed Abdulraheem
Brand Identity Development by: Jan Lewis
Source code & Pitch Deck (+presentation thereof) by: @fivosd