How Advances in AI Could Remove the Roadblocks to Renewable Electricity

­In the Netherlands, a solar power boom isbeing undermined by a grid capacity crunch, hindering its progress on climate change targets. Problems connecting to the grid are also holding up the energy transition withsome 930 gigawatts of renewable generating capacity currently held up in grid interconnection queues in the US alone.

Many believe this can only be solved through a huge and highly costly expansion of power gridswith Bloomberg’s NEF estimating we need $21 trillion of investment in 80 million km of new grid infrastructure to achieve net zero by 2050. However, this level of network expansion would impose a huge cost on industry and taxpayers, and could run into opposition from environmental groups and communities in the path of planned infrastructure.It can take up to 25 years to generate a return on such massive investments, and we will not see the first new power lines for 7 - 8 years at best.

What is less widely discussed is the necessity for a more efficient, commercially flexible grid that makes use of its full current-carrying capacity and allows operators to identify the cheapest and most productive renewable generators and accurately predict the amount of clean power that can be integrated into the grid in all locations, at all times. Together, these factors could unlock enormous extra renewable energy generation including distributed energy resources.

The gaps in grid data impeding the energy transition

A severe lack of monitoring capabilities across networks means utilities are failing to take advantage of desirable weather conditions or times of day to draw more from existing renewable power sources or safely carry more current. Utilities are also missing opportunities to ‘double up’ by sharing loads between parallel lines or prevent large-scale power loss and theft. Stopping the waste of renewable energy on our grids is increasingly imperative to delivering on the energy transition.

The waste and inefficiencies on power grids arises from the fact that these networks were not designed for the era of renewable energy. Power grids built around centralised power stations only needed to collect limited data mostly from sub stations and had little need for detailed monitoring of distribution grids or the ability, for example, to predict the impact of the weather on power generation. This means many operators have little to no visibility of the medium and low voltage distribution networks that increasingly draw on distributed power sources from batteries to rooftop solar panels. This prevents utilities drawing on all available clean power sources to reduce reliance on fossil-fuelled ‘peaker plants’ or identifying surplus capacity for interstate and international clean energy trading to accelerate the global transition to clean power.

Many utilities also lack real-time oversight of factors directly influencing renewable generation from windspeed and line temperature to air temperature (Figure 1).

For example, operators cannot predict the periods of most plentiful wind or solar power production without tracking windspeeds and ambient temperatures across their networks.

Network blind spots are also wasting capacity and thus increasing congestion. Many utilities set excessively cautious capacity limits based on crude inaccurate estimates instead of accurately monitoring line temperatures and local weather conditions in real-time. This wastes about 15% - 20% of capacity and reduces the amount of power that can be drawn from distant renewables such as offshore wind, compelling utilities to ‘top up’ from nearby fossil-fueled power plants instead. This needlessly skews the electricity mix towards fossil fuels.

How AI can break through grid barriers

Integrating more far-flung, fluctuating renewable power sources in a way that is commercially flexible and with an awareness of how much power can be handled, requires a shift towards sensitive, adaptive grids. Both in Europe and in the US, the rates for raw electricity, purchased from electricity generators such as PV, and wind, change every 15 minutes. An accurate understanding of the situation in both: grid capabilities and generation projection, and the ability to plan in advance how much energy will be routed along which lines will allow a reduction in electricity prices for the benefit of consumers.

Advanced analytics including AI and Machine Learning (ML) can cost-effectively transform old grids into smart electrical energy supply systems. Multi-sensor grid monitoring, which monitors over 60 electrical, physical and environmental parameters, from voltage, frequency and harmonics to cable sag, temperature and wind speed, produces a very rich database (Figure 2).

Click image to enlarge

Figure 2: EGM’s multi-sensor for overhead lines

This rich database, combined with the predictive power of AI, enables the real-time synchronisation of fluctuating renewable supply and demand harnessing the full capacity of the grid across each network segment. The same innovations can also predict and prevent causes of power loss and enable smarter power flow routes, to further increase the integration of renewable energy.

For example, ML algorithms can use historic data on cable temperatures and weather conditions to predict how much current can be safely carried across networks in specific weathers, daytimes and locations months in advance (Figure 3).

Click image to enlarge

Figure 3: Dynamic Line Rating (DLR) forecast screen based on weather and location of the line

This enables network operators to safely increase capacity and integrate more renewable energy when temperatures are lower such as in the evening or in locations with cooler conditions such as in the mountains. Data on fluctuations in supply and demand could also predict when and where loads could be shared between parallel lines to further boost network capacity.

The cognitive power of AI could help unlock more renewable generating capacity too.  Such powerful analytics systems show the effect of weather variations on renewable generation including distributed energy sources could predict potential spikes in renewable generating capacity in different conditions and locations. Operators could use this to forecast the times of day or year when solar or wind are at their most productive in each place so that utilities always avail of the cheapest and cleanest power sources. This could be matched with data on historic drivers of electricity demand to continuously synchronise renewable supply with demand in all weathers, reducing reliance on fossil fueled flexibility services.

Integrating more renewable energy also demands that we reduce electricity waste such as power loss and theft. New location-based fault detection systems could allow AIs to identify the site and source of power leaks and theft to help protect networks and conserve clean electricity.

Ultimately, this data can improve network designs so that future upgrades or build-outs are based around maximising capacity, conserving power and integrating more renewable energy in every location. Machine Learning systems could suggest the optimal siting of new networks to avail of the cheapest and most plentiful power sources at all times, or new replacement materials that could conduct more electricity. AIs could even predict the optimal network configurations and locations to reduce hazards such as fires and floods and avert power loss, creating more resilient grid designs.

Revamping powergrid strategy: the key to achieving net zero

The energy transition will not only require bigger grids but smarter ones that are tuned in to the requirements of renewable energy. Grids are being rapidly enlarged and diversified to draw on more renewable power sources but we need a parallel expansion of grid sensing, analytical and predictive capabilities. This is particularly essential as it has become clear that traditional grid monitoring systems recording only a few parameters in a few places will not be able to cope with the unique challenges posed by renewable energy.

Transitioning from centralised, stable power sources to intermittent, widely dispersed renewable generators will require networks that can intelligently expand capacity or switch power sources in response to changing conditions. Recent innovations in data analytics and Artificial Intelligence hold the promise of creating agile, adaptable networks that can unlock renewables and even avoid the need for some planned new grid infrastructure.

Electrical Grid Monitoring