Tetris with Electricity

• Scheduling loads with an energy management algorithm is like playing a game of Tetris where you have a surface to fill with blocks of various dimensions.
• These blocks have a time dimension and a power dimension. Together these dimensions add up to a quantity of energy.
• The solar resource is your board. You have an idea of what the board is going to look like based on weather prediction but you don't have perfect information - you only have an estimate.
• You have an idea of how accurate your estimate is based on what the estimate actually is. Let's say it's 8 a.m. and your estimate for noon is for a clearness index of .8 (clear sky). Let's assume the error of this prediction is +/- 10%. Alternately, let's say that your clearness index estimate for noon is .5 (cloudy). Your error bounds with a cloudy sky prediction will be more like +/- 30%.
• Example: If you estimate having 4 kW to work with in a clear sky situation you plan for 3.6 to 4.4 kW. If you estimate having 2 kW to work with in cloudy weather situation you set your bounds at 1.4 to 2.6 kW.
• The reliability of your weather prediction plugs into your energy management decisions. If you have loads with a range of flexibilities you want to assign those flexibilities to counter the unreliability of your weather prediction where possible.
• Back to Tetris. The most flexible block you could possibly have would have a 1 x 1 dimension. Blocks of 1 x 2, 1 x 3, 2 x 2 and so on are still flexible but less so. Blocks of 5 x 5 and 10 x 20 are flexible in terms of scheduling but you still have to fill them into the resource frame which is unpredictable.
• Your Tetris board fills up unpredictably with unscheduled loads. People turn lights on, cook dinner and watch TV. Your EMS system will make a fuzzy prediction of these unscheduled loads. It will assign error factors to the prediction of unscheduled load just as it assigns error factors to the prediction of weather. Your EMS learns you.
• An EMS system will want to scatter flexible load blocks such that the 10 x 20 blocks are matched up to the most reliable predictions, the 5 x 5 blocks are used in the medium error zones and the 1 x 1 blocks are used in the high error zones.