Wind Power UK Probability Distribution by Hour Nov 2008 to Sep 2010

The following two figures show the hourly probability distributions (and min, max, and mean) for Wind Power Production for 23 months from November 2008 to September 2010 (see source of data document in posts over the last couple of days). These statistics were computed from the 5-minute source power data for each hour of the day. Each hour group had approximately 8,388 data points. Showing the data this way demonstrates how the production of data is massively skewed to low power and are not uniformly distributed around the mean (which some may assume is the case).

Rate this:

Like this:

LikeLoading...

Related

This entry was posted on Wednesday, November 3rd, 2010 at 2:19 pm and is filed under Carbon, Climate, Energy, Renewables. You can follow any responses to this entry through the RSS 2.0 feed.
Both comments and pings are currently closed.

4 Responses to Wind Power UK Probability Distribution by Hour Nov 2008 to Sep 2010

Windspeeds are usually modelled as Rayleigh or Weibull distributions, with parameters specific to a given site. Can you calculate the parameters from the overall data and interpret what that means? Presumably you can arrive at the probabilities of less than xx MW at peak demand times?

Yes, given the data I can do the fitting. Quite easy with the tools that I have.

Given the wind speed data, I have the tools to do a best fit distribution, which like the above, result in distributions which are, as you say, Rayleigh or Weibull. The above are “best-fit” which came out as one called “BetaGeneral” which sort of looks like Rayleigh/Weibull. Best is defined by trying a bunch and picked the one with the best fit.

I can show you the fits for each hour if you like … seemed like a lot of graphs to post and with 24 of them they sort of look the same. But there are differences.

What I’m driving to with this work is to get a typical distribution for the power as generated; then model future energy mixes assuming that the basic distributions won’t change. I think that will be a good assumption for wind and will be ok to scale that up as the percentage of power produced by wind increases (which is the current policy/plan). It might be a big guess that the distributions for the base load fuels (coal, gas, and nuclear) will remain the same as they wind down. But for the first pass I’ll assume the pattern is the same.

Basically, we are going to have a large increase of skewed power (to left) from renewables and a decrease of “normally” distributed power that we have. This means a reduction, I believe, in the level of security of supply and will require an investment much higher than planned for renewables just to get the power out that we demand.

My hypothesis is that we can’t realistically do it and when we notice that it won’t work we won’t have investment time/funds left to deal with it. That then runs the risk of us dealing with it by managing demand, e.g. “smart grid” where power is rationed and all that. Just all exploratory now.

I can see the point of wind analysis as it *is* probabilistic. Less so the other generators as they are despatchable. And yes it *should* scale up if the current data was representative (which it aint).

It doesn’t matter so much if the wind fails to blow at times of low demand or if there are enough alternative generators and enough notice (hence the interest in forecasting). The UK’s balancing mechanism has a “gate-closure” time of an hour ahead of each half-hour period AIUI. Some of the other European grid operators work on up to 24 hours ahead. So comparisons between different national operators should be treated with caution.

As I see it the dangers to the UK’s security of supply come only in cold winter peak periods (at teatime) if forecast/expected wind doesn’t appear AND there are not enough alternative generators able to make up the shortfall in time. LOLP = loss of load probability is the jargon for blackouts.

Are you aware of the work of a Finnish lady called Hannele Holtinnen? She analyses this statistically using Monte Carlo simulation techniques, I think, but not for the UK grid, unfortunately. Lots of her stuff is behind subscription paywalls though.

All this is going to get more and more relevant as lots of our baseload coal and nuke stations are retired over the next few years.

Smart grids and perhaps real-time pricing may be good options regardless.

Thanks … realise what I am planning to do is simplistic. Not for a minute would I, as CEO of a power generation company, rely on these simulations! I’m coming at it from the perspective of testing the viability of 80% of the power coming from wind (or other renewables which as yet don’t exist so I’m going to ignore them for the time being). I agree with your assessment of risk to UK security of supply. It’s hard to understand how/why we would hold in backup so much power generation capacity “just in case”. And the quantity of capacity required to produce 80% just, by hunch, seems unrealistic. Like all hypotheses, it can be proved wrong or not proven, I guess.

I’m not aware of Hannele Holtinen or her work. Interesting she using Monte Carlo. That’s exactly what I am doing (I was avoiding using that techy term!). I’ll research her work. I’ll check it out.

There is a lot of work out there that I’m sure I’m not aware of. I need my staff back.

Windspeeds are usually modelled as Rayleigh or Weibull distributions, with parameters specific to a given site. Can you calculate the parameters from the overall data and interpret what that means? Presumably you can arrive at the probabilities of less than xx MW at peak demand times?

Yes, given the data I can do the fitting. Quite easy with the tools that I have.

Given the wind speed data, I have the tools to do a best fit distribution, which like the above, result in distributions which are, as you say, Rayleigh or Weibull. The above are “best-fit” which came out as one called “BetaGeneral” which sort of looks like Rayleigh/Weibull. Best is defined by trying a bunch and picked the one with the best fit.

I can show you the fits for each hour if you like … seemed like a lot of graphs to post and with 24 of them they sort of look the same. But there are differences.

What I’m driving to with this work is to get a typical distribution for the power as generated; then model future energy mixes assuming that the basic distributions won’t change. I think that will be a good assumption for wind and will be ok to scale that up as the percentage of power produced by wind increases (which is the current policy/plan). It might be a big guess that the distributions for the base load fuels (coal, gas, and nuclear) will remain the same as they wind down. But for the first pass I’ll assume the pattern is the same.

Basically, we are going to have a large increase of skewed power (to left) from renewables and a decrease of “normally” distributed power that we have. This means a reduction, I believe, in the level of security of supply and will require an investment much higher than planned for renewables just to get the power out that we demand.

My hypothesis is that we can’t realistically do it and when we notice that it won’t work we won’t have investment time/funds left to deal with it. That then runs the risk of us dealing with it by managing demand, e.g. “smart grid” where power is rationed and all that. Just all exploratory now.

I can see the point of wind analysis as it *is* probabilistic. Less so the other generators as they are despatchable. And yes it *should* scale up if the current data was representative (which it aint).

It doesn’t matter so much if the wind fails to blow at times of low demand or if there are enough alternative generators and enough notice (hence the interest in forecasting). The UK’s balancing mechanism has a “gate-closure” time of an hour ahead of each half-hour period AIUI. Some of the other European grid operators work on up to 24 hours ahead. So comparisons between different national operators should be treated with caution.

As I see it the dangers to the UK’s security of supply come only in cold winter peak periods (at teatime) if forecast/expected wind doesn’t appear AND there are not enough alternative generators able to make up the shortfall in time. LOLP = loss of load probability is the jargon for blackouts.

Are you aware of the work of a Finnish lady called Hannele Holtinnen? She analyses this statistically using Monte Carlo simulation techniques, I think, but not for the UK grid, unfortunately. Lots of her stuff is behind subscription paywalls though.

All this is going to get more and more relevant as lots of our baseload coal and nuke stations are retired over the next few years.

Smart grids and perhaps real-time pricing may be good options regardless.

Thanks … realise what I am planning to do is simplistic. Not for a minute would I, as CEO of a power generation company, rely on these simulations! I’m coming at it from the perspective of testing the viability of 80% of the power coming from wind (or other renewables which as yet don’t exist so I’m going to ignore them for the time being). I agree with your assessment of risk to UK security of supply. It’s hard to understand how/why we would hold in backup so much power generation capacity “just in case”. And the quantity of capacity required to produce 80% just, by hunch, seems unrealistic. Like all hypotheses, it can be proved wrong or not proven, I guess.

I’m not aware of Hannele Holtinen or her work. Interesting she using Monte Carlo. That’s exactly what I am doing (I was avoiding using that techy term!). I’ll research her work. I’ll check it out.

There is a lot of work out there that I’m sure I’m not aware of. I need my staff back.