n step ahead simulation of SARIMA process using R forecast package

I want to simulate an SARIMA process nsteps ahead, with nsim paths. I've already fitted the model, and the 'forecast' package has a 'simulate' function. However as far as I can see whilst it includes an nsim argument it only simulates 1 step ahead? Is this possible with the 'forecast' package?

https://robjhyndman.com/hyndsight/simulating-from-a-specified-seasonal-arima-model/

What I'm trying to achieve is work out the standard error of the sum of auto-correlated residuals (where the SARIMA model is fitted to the residuals of a model which exhibits auto-correlated residuals). So I want to perform monte-carlo simulation nsteps ahead, sum the values per simulation then work out the standard deviation of the sums.

Solution

nsim is the number of observations to simulate ahead. Set it to whatever you like. Here is an example.

library(forecast)
library(ggplot2)
fit <- auto.arima(USAccDeaths)
autoplot(USAccDeaths, series="Data") +
  autolayer(simulate(fit, nsim=36), series="Simulated")

^{Created on 2019-07-30 by the reprex package (v0.3.0)}