New systems have to be installed to support this kind of forecasting. Forecasting the output is critical, as it determines when to fire up the large fossil plants to support days when the wind is either going to blow too strongly or not at all, or as in the case of photovoltaic, when the cloud is present and blocks the sun.
So a key problem then is how to optimise production scheduling for conventional power plants. Ideally you would take a plant offline for maintenance shutdown, when certainty of renewable resources is high.
There is an ancillary problem also being discovered by distribution companies. Not only are their systems in danger of occasionally collapsing, but more prevalent is the added wear and tear on components, which is increasing maintenance costs and reducing grid reliability.
One way to address this problem is to use sophisticated forecasting techniques and measure the load across all the components of the grid. In this way the maintenance tasks can be prioritised. This however is a significant computational effort that usually takes weeks to produce a prediction.
This problem is being solved by high performance in memory computing, which can render load forecast results in minutes rather than days.
This then allows the condition in the grid to be monitored in as close to real time as possible and remediation can be taken far earlier for critical situations likely to lead to collapse.
Companies like Alliander of the Netherlands, and Hydro One of Canada are in the early phases of adopting these technologies to minimise disruption and reduce maintenance costs.
This takes us one step closer to realising the benefits of an energy revolution.
Phillip Vaughan is Director for Utilities, SAP Asia Pacific Japan
Sign up for CIO Asia eNewsletters.