Gridded modelling of GB wind generation and electricity demand to 2035:

Disaggregation of National Grid annual scenarios

 

Ed Sharp:

ed.sharp@ucl.ac.uk | www.esenergyvis.wordpress.com | @steadier_eddy

 

Context and Framework

Modelling context: Scenarios

• National Grid FES
• Counteracts some uncertainty
• National annual resolution
• Many changes modelled
• Focus is wind and heat pumps
• Good supporting data
• Other demands and supplies wrapped up in techno economic modelling
• Fit for purpose
•  Lack of disaggregated modelling
• Therefore uncertainty remains ……

• Scenarios are designed to counteract some of the uncertainty in projecting change to the energy system.
• Due to this uncertainty, however, it is unrealistic to perform this scenario modelling at a lower resolution than national, annual
• Therefore uncertainty remains on the implications of these scenarios, particularly when aspects of supply and demand which fluctuate at small resolutions such as those which are driven by weather are introduced/increased.
• The view of my research is that if scenarios are predetermined it is possible to reduce this uncertainty through disaggregated modelling.
• Also very importantly that demand should be modelled, to allow the consideration of new demand, not just growth of existing demands.

Why model on a grid?

Research hypothesis

“National annual resolution scenario modelling can be complemented through spatiotemporally disaggregated modelling which captures the inherent variability of wind generation and weather driven electricity demand; furthermore, disaggregation of scenarios can be achieved using existing methods and data.”

        Gridded Approach

• Disaggregation can be carried out using any spatial units, though practically the choice is between census geographies (at any resolution) and grid.
• Here the grid is used for a number of reasons
  • Data availability (especially weather – NCEP CFSR)
  • Flexibility in programming, arrays vs. databases
  • Homogeneity
  • Time: Census geographies change over time
  • Space: especially over offshore areas
• The model grid
  • 0.5 degree lat/lon grid
  • >200 onshore grid squares

Why model on a grid?

Weather Data

Weather data: Evaluating the gridded approach

• Grid point values – don’t necessarily represent the whole grid
  • Fundamental to gridded approach
• Therefore compared to in situ measurements of wind speed
• Overall a high correlation and low RMSE found, with some exceptions…
  • No land use or coastal bias found
  • But the highest elevation sites were not well represented by CFSR
• Analysis performed at 10m
  • Unfortunately not the height of turbines

Sharp, E. Dodds, P. Barrett, M. and Spataru, C. Evaluating the accuracy of CFSR reanalysis hourly wind speed forecasts for the UK, using in situ measurements and geographical information. Renewable Energy, 77, 527-538. 2015. https://goo.gl/gJrcR1

Estimating Hourly Wind Generation

Wind generation: estimating hourly generation using CFSR

• Archetype curves used in modelling
• Simple height correction
• Manufacturer curves don’t represent the operation of a turbine in reality​

Factors not incorporated in model

• Wake Effects
• Downtime
• Lag in turbine operation
• Energy Density over blades
• Air density
• Decline in performance with age
• Wind speed changes < 1 hour in frequency Location specific turbine curves
• Operating efficiency
• Atmospheric stability, surface roughness and orography in height correction

Wind generation: evaluating the approach

• Approach yielded reasonable results
• Against hourly generation R2 = 0.66,
• RMSE =  0.22 GW
• Monthly ROC data R2 = 0.89
• RMSE = 0.08 GW
• Room for improvement, but national generation OK

Wind generation: estimating hourly generation from different capacities

•  Site specific approach would have yielded more accurate results
• The gridded approach however facilitates disaggregation of scenarios, without the need to select exact farm locations
• How can National Grid’s scenarios be redistributed to the model grid?

Wind generation: estimating hourly generation from different capacities

 Step 1: Eliminate unsuitable areas

• Constraints on development identified from the literature
• Particularly resource estimation studies
• Spatial data collated
• Tested against operational capacity
• Small number of farms within these zones
• 31% of GB removed from consideration
• In reality there are further restrictions e.g. from social acceptability
• However very low likelihood areas are removed
• This analysis removes sites above 600 m.

Wind generation: estimating hourly generation from different capacities

 Step 2: Identify suitable land uses

• Using locations of existing wind farms and land use mapping
• Clearly four commonly used land types
• These take precedence in subsequent allocation

 

 Step 3: Identify high wind sites

• Characterising wind speed from CFSR by grid square in long term >30 years shows all GB grid squares suitable

Wind generation: estimating hourly generation from different capacities

 Step 4: Allocate capacity to grid

•  Fill available land
•  Prioritise high wind, land uses, spatial diversity
•  Generous assumptions made on turbine spacing (using archetypes)
•  Offshore use development zones
•  Ample space for all scenarios

Analysis - Wind Generation

Wind generation: annual changes in generation and wind speed

• Annual variation of 20% - with static capacities at end of scenarios
• Driven by different wind profiles
• Hourly mean wind, onshore and offshore found to be highly correlated with generation – useful for prediction.

The percentage difference between SpWind and NG at different capacities.

Histograms of onshore and offshore mean national wind speeds, 2007 - 2010.

Wind generation: geographical diversity

Correlation between generating grid squares for each scenario, 2010 - 2035.

See  Sinden (2007) for original method

Method –

Estimating hourly electricity demand

Electricity demand: overview of method

Top Down

Bottom Up

• Weather independent, building independent, e.g. cooking
• Advantages: simplicity, available data
• Disadvantages: difficult to model changes in the way energy is used

• Dependent on weather and building fabric
• Advantages: model changes in the way energy is used
• Disadvantages: more complex data needed

Electricity demand: people in SpDEAM

• Domestic demands must be disaggregated to the grid so that wasted heat can be included in the heat demand equation
  • Gridded population data used (Gridded Population of the World)
  • Assumed that people use the same amount of energy
    • Could be changed

• Adapted to % of pop in each grid square, this remains static in scenario modelling, absolute population changes
• The model structure does however provide the ability to model migration  etc.
• Heat from people calculated from pop database
• Assuming even distribution across building types (see buildings in next slide)

 

Electricity demand: buildings in SpDEAM

• Dwellings heated to desired internal temperature (only heating) from external T (CFSR)
• Heat loss dependent on dwelling type and floor area
  • Calculated using census data
  • Scenarios increase housing stock and change heat loss coefficients (including the effect of demolitions)
• Solar gains assuming glazed area using CFSR data
• Incidental gains calculated from redistributed top down demands
  • Evenly distributed amongst dwelling types (could be altered)

Electricity demand: heating technologies in SpDEAM

• Heat demand Assigned to heating technologies based on census central heating data (E and W only, Scotland needed assumptions)
  • Uniformly distributed across dwelling types
  • These are adapted in scenario modelling to represent the introduction of heat pumps
  • Relationship between COP and external T derived from field trials

Electricity demand: calibration and evaluation of SpDEAM

• Calibrated from 2000 – 2010
• Evaluated 2010 – 2014
• Remarkably accurate given the challenge of modelling the entire GB electricity system with limited data and necessary simplification and grouping of demands

All electricity demand - hourly

All gas demand - daily

Analysis - residual electricity demand

Analysis - Variability

Residual Electricity Demand: Annual variability

• On an annual basis, the reduction in residual demand is approximately proportional to the increase in wind capacity
• Final years of the scenario period demonstrate that there may be lengthy periods of colder weather with reduced wind speeds
• Accentuated by shifts in the way in which electricity is consumed, as demonstrated by the steeper increase in residual demand under the GG scenario

Annual changes in residual demand

Hourly variability in wind generation, electricity demand and residual demand

Increased variability in both scenarios, higher capacity factors throughout the later in years under Gone Green on the left, especially in the colder months.

Predictable variability under both scenarios for all years.  Little evidence of the impact of heat pumps on the temporal patterns of electricity demand.

The Gone Green scenario experiences greater variability as a result of more wind capacity, particularly offshore.

Wind Generation

Electricity Demand

Residual Demand

Hourly variability in residual demand

Publications:

• Thesis available at: https://goo.gl/ETYWiO   
• Blog: www.esenergyvis.wordpress.com 
• Sharp, E. Dodds, P. Barrett, M. and Spataru, C. Evaluating the accuracy of CFSR reanalysis hourly wind speed forecasts for the UK, using in situ measurements and geographical information. Renewable Energy, 77, 527-538. 2015. https://goo.gl/gJrcR1
• Sharp, E. Spatiotemporal Disaggregation of GB Scenarios Depicting Increased Wind Capacity and Electrified Heat Demand in Dwellings. 38th IAEE International Conference, Antalya, Turkey, 2015.
• Sharp, E. Spataru, C. Barrett, M. and Dodds, P. Incorporating building specific heat loss and associated energy demand into electricity demand models for Great Britain. Building Simulation and Optimisation UCL, London, UK, 2014.

Contact etc.:

Email: ed.sharp@ucl.ac.uk

Linkedin: ed.sharp.09

Web: www.bartlett.ucl.ac.uk/energy

Twitter: @ucl_energy | @steadier_eddy