As we’ve explored in recent weeks, energy storage can deliver a host of benefits to renewable energy asset owners and grid operators. Successful operation of energy storage systems to meet differing demands, sometimes throughout a single day, relies on more than just the cells that power them. System operators call on a range of smart software solutions to anticipate and deliver across various markets and ensure their energy storage assets continue chasing the optimum operational strategy.

In Great Britain, for example, energy storage connected across the grid network can utilise a very diverse portfolio of revenue streams including the Balancing Mechanism, ancillary services like Dynamic Containment, wholesale trading, renewable energy peak shaving, and back-up power delivery if on a commercial site. While stacking these revenue streams, where feasible, remains the best way to extract maximum value from assets, each places unique requirements on the technology, cycling it differently and ultimately contributing to system degradation.

Artificial intelligence-based systems are widely being used to counter the technical and market risks that can stand in the way of energy storage investments. AI technology can anticipate system loads by drawing on weather forecasting, renewable generation data and market need through electricity price signals, to determine the most optimal use of an energy storage system in terms of both revenue and system operations. This can be used to facilitate smart software driven bidding into several competitive markets or, in the case of some applications in isolated regions, form whole grid systems by integrating and optimising generation from any source.

The advantages of adopting AI can be seen in markets around the world, from New York’s Value of Distributed Energy Resources scheme and Canada’s Summerside Electric Utility living lab strategy, to the UK’s Quest project, with the latter seeking to release 2 GW of renewable power capacity and save network operators £250m (~$337m).

Machine learning can also be used to deliver predictive maintenance at energy storage sites and track rates of degradation associated with their activities, which are often tied to performance warranty agreements. Such agreements can place demanding requirements on operators depending on how the system is used, with the number of times a battery is cycled and how it is maintained dictating how many years of use it can offer. System providers can require hardware to be kept at constant temperatures and states of charge, with data to be provided as evidence in case of a claim against the warranty. Some energy storage vendors demand system data is logged every 15 minutes – only smart systems can track and store the required metrics.

Such solutions will continue to play a critical role – operating and maintaining energy storage systems is a far more complex proposition than seen with other technologies that underpin the energy transition.