Received: 01 June 2017; Accepted: 13 December 2017; Published: 03 January 2018
This application will allow the user to put in their coordinates and house and roof sizes in order to receive some useful data on solar energy power. The program will give a recommendation for solar panel brand, and whether or not the user should get solar power. In addition, there is second part where the user can visualize recommendations for a specific house and roof size in cities throughout the entire U.S.
Server and code is written in Go, visual aspects written in HTML and Javascript. Deployed using Heroku Cloud PaaS.
Data sourced from US Climate Data, NASA Atmospheric Science Center, NASA, Solar Reviews, timeanddate.com, US Energy Information Administration, and Weatherbase. Equations sourced from Solar Electricity Handbook and Rexel.
Perhaps no technology has generated more hype in recent years than artificial intelligence (Ai). In some industries, it is certainly living up to it. Ai powers the algorithms behind transport apps Uber and Waze, microtargeted online advertising, your commercial flight’s autopilot, and digital assistants like Amazon’s Alexa.
Yet Ai’s impact in energy efficiency and renewables deployment has been less frequently touted. This is unfair. Ai is fast increasing the value of wind and solar generation, energy storage, and other areas through machine learning techniques. One popular method called reinforcement learning (RL) is yielding especially impressive benefits in the energy space.
RL builds algorithms through a dual approach: dataset training to optimize performance and mathematical function approximation to optimize solutions. It excels in large, complex decision spaces where constraints and payoffs can be modeled, but the only way to create and improve analytical solutions in that space is to interact with it. For example, in 2017 Google’s DeepMind RL algorithm (called AlphaZero) needed only basic chess rules, a database of games, and 4 hours of self-play to then dominate the strongest chess program in the world, Stockfish. These results and the relative ease in attaining them amazed a global chess community long accustomed to computer-player superiority.
RL is performing similar feats in power generation. Google DeepMind has turned its attention to making intermittent wind generation more predictable. Using function approximation on weather forecast data and training on historical turbine data, the DeepMind RL algorithm predicts power output 36 hours ahead of actual generation. This allows advance scheduling of electricity deliveries to the grid, changing load profile planning and effectively commoditizing the resource. In 2019, DeepMind AI has boosted the value of 700 MW of wind capacity supplying Google data centers by over 20%.
Ai’s significance is even larger as a system management tool for solar plus storage installations. Ai provides enhanced control, flexibility, and value throughout the project lifetime, diversifying available value streams and balancing out the effect of variable rates. In particular, solar plus storage RL uses historical rate, capacity, and other data alongside creative function approximation to deliver predictive optimization for generation, storage, and purchasing.
One solutions provider, Stem Technologies, estimates that its Athena Ai platform adds 5%-20% project value from demand charge and load shift savings through storage, and another 10%+ from optimized integration into utility demand response, wholesale, and other programs. This can create approximately 50% additional developer revenue, and 1.5-2 times the developer profit. Stem is seeing significant growth: entering 2019, Stem surpassed 100 MWh of deployed Ai-driven energy storage systems, an industry milestone.
Hype can easily lead to disappointment. But Ai’s much-vaunted potential is being realized in the energy sector. RL is one of several Ai techniques creating major efficiencies across the value chain, which will grow as software solutions become more sophisticated and widely adopted. Many Ai-enabled enterprises have yet to connect separate Ai applications into complete end-to-end automated processes, much less into Ai-powered process flows. When that happens, the full potential of Ai will be unleashed.
We are still at the beginning of Ai’s career as a disruptor in the energy space. It is time that Ai (with all its Hollywood connotations) be seen in a new light—as a substantial technology that is fast living up to its efficiency potential.