The solarModel class combines seasonal components, bounded
transformations, ARMA-GARCH dynamics, and monthly Gaussian mixture
residual models for solar radiation data.
Value
An R6 object of class solarModel.
Details
The class stores a solarModel_spec object and fitted data summaries. Most
public methods update the object state in place and are intended for the
package fitting workflow.
Public fields
interpolatedLogical values, when TRUE the time-series and the parameters are interpolated.
Active bindings
dataA data frame with fitted data, seasonal quantities, and monthly parameters.
seasonal_dataA data frame containing seasonal and monthly summaries.
monthly_dataA data frame containing monthly parameters.
loglikThe log-likelihood computed on train data.
specA solarModel_spec object that controls the model fitting process.
momentsGet a list containing the conditional moments.
coefficientsGet the model parameters as a named list.
Methods
Method new()
Initialize a solarModel
Returns
Initializes object state during construction.
Method fit()
Initialize and fit a solarModel object given the specification contained in $control.
Returns
Updates the object state. The return value is not intended for use.
Method fit_seasonal_model_Ct()
Initialize and fit a seasonalClearsky model given the specification contained in $control.
Usage
solarModel$fit_seasonal_model_Ct()
Returns
Updates the clear-sky model and fitted data stored in the object.
Method compute_risk_drivers()
Compute the risk drivers and impute the observation that are greater or equal to the clear sky level.
Usage
solarModel$compute_risk_drivers()
Returns
Updates fitted data and outlier information stored in the object.
Fit the parameters of the solarTransform object.
Usage
solarModel$fit_transform(detect_outliers = TRUE)
Arguments
detect_outliersLogical. If TRUE, transformation-bound outliers are detected.
Returns
Updates transformation parameters and transformed data stored in the object.
Method fit_seasonal_mean()
Fit a seasonalModel the transformed variable (Yt) and compute deseasonalized series (Yt_tilde).
Usage
solarModel$fit_seasonal_mean()
Returns
Updates fitted seasonal mean quantities stored in the object.
Method fit_monthly_mean()
Correct the deseasonalized series (Yt_tilde) by subtracting its monthly mean (Yt_tilde_uncond).
Usage
solarModel$fit_monthly_mean()
Returns
Updates monthly mean corrections when enabled.
Method fit_mean_model()
Fit an ARMA model (Yt_tilde) and compute the residuals (eps).
Usage
solarModel$fit_mean_model()
Returns
Updates ARMA fitted values and residuals stored in the object.
Method fit_seasonal_variance()
Fit a seasonalModel on AR squared residuals (eps) and compute deseasonalized residuals eps_tilde.
Usage
solarModel$fit_seasonal_variance()
Returns
Updates seasonal variance quantities stored in the object.
Method fit_monthly_variance()
Correct the standardized series (eps_tilde) by subtracting its monthly mean (sigma_uncond).
Usage
solarModel$fit_monthly_variance()
Returns
Updates monthly variance corrections when enabled.
Method correct_seasonal_variance()
Correct the parameters of the seasonal variance to ensure a unitary variance
Usage
solarModel$correct_seasonal_variance()
Returns
Updates seasonal variance parameters when correction is enabled.
Method fit_variance_model()
Fit a GARCH model on the deseasonalized residuals (eps_tilde).
Compute the standardized (u) and monthly deseasonalized residuals (u_tilde).
Usage
solarModel$fit_variance_model()
Returns
Updates GARCH variance and standardized residuals stored in the object.
Method fit_mixture_model()
Initialize and fit a solarMixture object.
Usage
solarModel$fit_mixture_model()
Returns
Updates the monthly Gaussian mixture model stored in the object specification.
Update the parameters inside object
Usage
solarModel$update(params)
Arguments
paramsList of parameters. See the slot $coefficients for a template.
Returns
Updates model parameters and resets the stored log-likelihood.
Method update_moments()
Update the moments inside object
Usage
solarModel$update_moments()
Returns
Updates the stored conditional moments.
Method update_logLik()
Update the log-likelihood inside object
Usage
solarModel$update_logLik()
Returns
Updates observation-level and total log-likelihood values.
Method update_risk_drivers()
Update the clear sky and risk drivers
Usage
solarModel$update_risk_drivers()
Returns
Updates clear-sky values, risk drivers, and transformed data.
Method update_classification()
Update the classification of the Bernoulli random variable.
Usage
solarModel$update_classification(filter = FALSE)
Arguments
filterLogical. If TRUE, the mixture model is filtered before classification.
Returns
Updates mixture classification columns stored in fitted data.
Filter the time series when new parameters are supplied in the method $update(params).
Usage
solarModel$filter(fit = FALSE)
Arguments
fitLogical. If TRUE, monthly mean and variance corrections are recomputed.
Returns
Updates the stored data summaries. The return value is not intended for use.
Compute the log-likelihood of the model and update the slot $loglik.
Usage
solarModel$logLik(moments, target = "Yt", quasi = FALSE)
Arguments
momentsDataset containing the moments to use for computation.
targetCharacter. Target variable to use, either "Yt" or "GHI".
quasiLogical. If TRUE, compute the Gaussian quasi log-likelihood.
Returns
A numeric vector of observation-level log-likelihood values.
Print method for solarModel class.
Returns
Prints a summary of the object.
Method clone()
The objects of this class are cloneable with this method.
Usage
solarModel$clone(deep = FALSE)
Arguments
deepWhether to make a deep clone.
Examples
spec <- solarModel_spec$new()
spec$specification("Bologna", max_date = "2005-01-10")
#> No outliers!
model <- solarModel$new(spec)
if (interactive()) {
model$fit()
model$loglik
}