Tomorrow's load, before tomorrow.
PJJ Forecast predicts your factory's load and your array's solar output up to 24 hours ahead. It learns from roughly six months of your site's own history plus live weather data, and every decision the rest of CESS makes is made against it.
The model is trained on your site, not on a generic factory. That is why the feasibility stage asks for your history first.
A battery can only shave a peak it saved charge for.
Every one of these is a bet on the next few hours. Without a forecast, the controller is guessing.
A demand charge is billed on your single highest interval of the month. By the time that interval is happening, the only thing that can help is charge already sitting in the battery, and whether it is there depends on decisions made hours earlier.
So the useful question is never what is the load now. It is what the load will be this afternoon, how much of it the array will cover, and how much charge that leaves the battery to hold back. A controller that only sees the present cannot answer it.
PJJ Forecast answers it for the next 24 hours, for your load and for your solar, and hands the answer to PJJ Dispatch, which decides what the battery does about it.
A day ahead, drawn before the day happens.
A forecast is a range, not a line. It is narrow for the next hour and wider for this evening, and dispatch plans against the whole range rather than betting on its middle.
In the product, the same picture is the Dashboard's 24-hour chart of total load, grid import, battery and solar, with the peak-tariff window shaded.
Two models, trained on your site and nobody else's.
A factory's load is not weather. It follows the production plan, and the production plan has a shape. The forecast is only as good as its grip on that shape.
Your factory's signature
Shift patterns, batch cycles, which lines run together, what a Monday looks like against a Saturday. Learned from your metered history, so a compressor that starts at the same time every day stops being a surprise.
What your array will make
Expected output hour by hour, including the hours it will make more than the site can use. That surplus is what PJJ Recover is waiting for, so it has to be seen coming.
Live conditions, fed in
Weather data streams into the solar model continuously, so a cloudy afternoon changes tomorrow's plan before it changes tomorrow's bill.
History first. That is why we ask for it before anything else.
The model trains on roughly six months of your site's history. Most of it already exists, in your bills, your meter and your inverters. Nothing needs to be installed to start.
A forecast is a probability, not a promise.
It will be wrong on some days. That is why dispatch uses model-predictive control, which re-plans as the day unfolds rather than committing to the morning's forecast and hoping.One module does not run on its own.
PJJ Forecast is one of four parts of the Cloud Energy Storage System. Each hands something to the next, and the savings come from all four running together.
Find out what CESS would do at your site.
Twelve months of billing data, and we will model your load, size the system and show you the whole financial case, including when the answer is not to build.
Or reach us at solarbuddies@pjjsolutions.tech