# State space representation of autoregressive model with long lag

I am working with hourly data and using an autoregressive model:

$$y_t = \phi_1y_{t-1} + \phi_2y_{t-2} + \phi_{24}y_{t-24}+\varepsilon_t$$

Expressing this model using the standard state-space formulation (Rob Hyndman slides) would need a 24-dimensional state-vector and the parameter restrictions $$\phi_3...\phi_{23}=0$$.

Is it possible to express this model more efficiently using fewer states?