Download presentation
Presentation is loading. Please wait.
1
CROSS 2017 Symposium October 2017
pvlib-python et al CROSS 2017 Symposium October 2017 Sunshine Analytics LLC
2
Sunshine Analytics LLC
Introduction Jessica Forbess Sunshine Analytics 2013 – present First Solar EPC – testing DNV GL IE – due diligence Electrical engineer, remote network monitoring for cell networks, leveraged into solar monitoring, leveraged into analytics Sunshine Analytics LLC
3
PV Energy Modeling – What?
Location Sun position / Shading Meteo Irradiance / Temperature / Wind Precipitation (dust / snow soiling) Orientation Tilt / Azimuth Equipment Modules Inverters Other system losses Energy Output Sunshine Analytics LLC
4
PV Energy Modeling – Why?
Developers want a number to sell the project EPCs need a number to test against Debt providers need a conservative number to cover the loan Equity providers want to know what the upside is O&M needs to have a baseline to know if there are operational issues Regulators / Off-takers increasingly care about power quality as well as energy – and forecasting Independent Advisors are called upon to create models for each of these scenarios Sunshine Analytics LLC
5
Consider the timescale – Annual/month
Annual / monthly Overall value, cash flows Metrics: Yield Performance Ratio Tool: PVWatts – NREL (mostly deprecated) Annual and monthly models used to be pretty typical, with insolation for the location calculated on either of those timescales, and then fixed factors applied Sunshine Analytics LLC
6
Consider the timescale – Subhourly
Hourly / subhourly Testing against measured data Troubleshooting Metrics: regressions actual vs expected peer comparisons Tools: PVsyst – the gold standard SAM – NREL’s more detailed modeling tool Helioscope / Aurora – C&I, portfolio based, some automated shading methods Developer models – SPWR and FSLR pvlib – Sandia developed Sunshine Analytics LLC
7
Consider the timescale – Subsecond
Subsecond / subcycle Power quality Tools: GridPV? Sunshine Analytics LLC
8
Current PV Modeling Tools
Free as in Beer and/or speech Sunshine Analytics LLC
9
Sunshine Analytics LLC
SAM SAM – System Advisor Model Slightly less tweakable than PVsyst SDK for inclusion in applications Open sourced soon? Workflow designed to get annual energy models to estimate project value Can extend to project lifetime value Great documentation of modeling methods and calculations Sunshine Analytics LLC
10
Sunshine Analytics LLC
pvlib Sandia internal tool in MATLAB Standardized analyses across the PV group python library ported in 2013 Now maintained on github by Sandia and a assistant prof at U of Arizona user base growing conda and pip packages BSD3 license Sunshine Analytics LLC
11
Sunshine Analytics LLC
pecos Sandia internal tool Monitoring multiple experimental PV arrays Now maintained on github by Sandia pip package modified BSD3 license Designed to monitor and perform quality checks on time series data Sunshine Analytics LLC
12
Sunshine Analytics LLC
pvlib – toolbox Location Sun position / Shading Meteo Irradiance / Temperature / Wind Precipitation (dust / snow soiling) Orientation Tilt / Azimuth Equipment Modules Inverters Other system losses Energy Output module = sandia_modules['Canadian_Solar_CS5P_220M___2009_'] inverter = cec_inverters['SMA_America__SC630CP_US_315V__CEC_2012_'] system = SingleAxisTracker(module_parameters=module, inverter_parameters=inverter, series_modules=15, parallel_modules=300) lat, lon = 45.5, # Portland, OR start = pd.Timestamp.now() end = start + pd.Timedelta(days=7) for fx_class in [GFS, NAM, HRRR, RAP, NDFD]: fx_model = fx_class() fx_data = fx_model.get_processed_data(lat, lon, start, end) irradiance = fx_data[['ghi', 'dni', 'dhi']] weather = fx_data[['wind_speed', 'temp_air']] mc = ModelChain(system, fx_model.location) mc.run_model(fx_data.index, irradiance, weather) mc.ac.plot() Sunshine Analytics LLC
13
Sunshine Analytics LLC
pvlib – Forecasting Forecasting is rare in standard PV modeling tools Unidata’s Siphon library simplifies access to forecast data hosted on the Unidata THREDDS catalog Estimated meteo parameters are provided by an external source, so the weather model is not part of the library Irradiance components may be poor quality Estimating irradiance from cloud cover is part of the pvlib forecast process Experimentation is welcome here Sunshine Analytics LLC
14
Typical process – Actual vs Expected
jupyter notebook config file for system equipment and location measured meteo data and energy run meteo through model, compare to energy on hourly, daily, monthly basis Sunshine Analytics LLC
15
Typical process – Hourly kW comparison
jupyter notebook create typical year energy model (hourly) calculate probabilities that a given hour of a given month would be at a given kW output Use seaborn to visualize Sunshine Analytics LLC
16
pvlib – clear sky detection
Standard library function to detect clear sky periods based on a paper by Matthew Reno Clear sky periods better for certain analysis, peer to peer comparisons Being able to automatically filter for those periods is very powerful Sunshine Analytics LLC
17
Sunshine Analytics LLC
Further development Shading Row to row shading Topographic impacts on row to row External objects Interaction with storage and the grid Optimizing value of PV capacity, battery capacity based on local load and time-of-use rates EV charging / discharging Sunshine Analytics LLC
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.