ECN publication
Title:
Development of a Stochastic Weather Simulation Capability for ECN
 
Author(s):
 
Published by: Publication date:
ECN Wind Energy 6-12-2016
 
ECN report number: Document type:
ECN-E--16-064 ECN publication
 
Number of pages: Full text:
16 Download PDF  

Abstract:
This report describes the development of a capability for stochastic weather simulation at ECN, to be made available to the users of ECN's strategic modelling tools (such as ECN Access and ECN Install) in order to run large numbers of independent time-domain simulations for each strategic choice which the users wish to evaluate.

Developing long-term strategy for offshore wind farms involves understanding the uncertainty in environmental conditions. Waves and winds are stochastic processes, meaning that they are fundamentally random and any particular historical or predicted time series can only capture part of the variability inherent in these processes. If only a small number of realisations of these processes are used to evaluate the effectiveness of a strategy, it is likely that bias will be introduced into the results.

The ability to run simulations using all relevant weather conditions offshore - wind speed and direction; wave height, period and direction; sea level; and current speed and direction – is of increasing relevance to the industry. The industry is moving towards a greater understanding of the influence of weather parameters (beyond wind speed and wave height) on the likelihood of successful maintenance operations and therefore the profitability of an offshore wind farm.

Based on examples in the literature, a method has been developed and implemented for simulating multivariate stochastic weather time series, which has been shown to be successful at preserving the relevant statistical properties of existing data sets. This methodology follows the following steps:
1. Fit and remove trends and seasonality for each variable;
2. Transform each variable to a Normal distribution using its cumulative probability density function;
3. Fit a model to identify the autocorrelation function of each variable;
4. Simulate time series from each variable from the model;
5. Transform each variable back, but using the cumulative distribution function conditional on the values of the other variables simulated;
6. Replace trend and seasonality.

This new stochastic weather simulation capability is herein demonstrated to be capable of providing simulations of the five stochastic variables for wind and waves, reproducing the relevant statistics of the original data set, particularly occurrence and weather window probabilities.


Back to List