CalTestBed Case Study

The California Test Bed Initiative conducted multiple closed-loop microgrid calibration tests using Xendee's OPERATE Model Predictive Controller (MPC) over the course of a year resulting in up to 80% in total cost reduction.

The California Test Bed Initiative—CalTestBed—is a laboratory voucher and commercialization development program for innovators and entrepreneurs working to bring early to mid-stage clean energy concepts to market. It is made possible by a grant from the California Energy Commission. CalTestBed accelerates clean energy innovations toward commercialization by awarding vouchers to test technologies at one of 70+ testbed facilities across Lawrence Berkeley National Laboratory and 9 University of California Campuses.

Multiple calibration tests over the course of the year were performed, leading to a series of successful closed-loop tests using our Xendee OPERATE Model Predictive Controller (MPC). 

Since these tests were conducted for different months, the energy and demand charge tariffs were different for each test in the series. All MPC tests were compared to the existing rule-based approach of solar PV surplus charging and discharging for similar durations as we ran this rule-based approach in parallel to our MPC.

Preliminary tests were performed last August, followed by upgrading the methodology to a patent-pending demand charge reduction algorithm, and performing a final test this year in May.  The comparison of these tests resulted in an 80% reduction of total costs. Specific results for last August indicate that the innovative demand charge reduction methodology increases the performance of MPC by 17% in total energy costs indicating the savings will vary significantly depending on the month.

Additionally, our other pilot program in the Midwest was able to show a 65% reduction in demand costs. Based on our findings, active control by microgrids can lead to massive reductions in demand charges and pricing but depends greatly on the technologies selected. For example, for projects with existing solar panels and batteries, the demand charge reduction might be lower than when when adding additional DER technologies. 

Figure 1. This dashboard showcases how the Model Predictive Controller compares with the rules-based reference case. In this example, the MPC was able to flatten demand charges throughout the day resulting in a maximum demand that was only 30% of the reference case. The chart also shows the forecasted controller signals, and demands, and is updated every five minutes. Click to expand.

The major lessons learned from the testing are as follows:

1. The testing at UCSD revealed that uncertainty introduced by forecasting errors can have a very significant impact on demand charge cost reductions. The challenge associated with demand charges is that they are mostly assessed on the highest monthly peak demand in a 15-min window. Thus, missing the true highest 15-min peak can eliminate all monthly cost savings. Therefore, a new methodology was needed to track the accumulated forecasting errors and ensure high demand charge cost savings. An innovative patent pending methodology has been developed that enables high precision and the aforementioned 80% cost reductions.

2. The communication between DERs and Xendee OPERATE is an integral part of our technology. This can be done in different ways as there are different communication protocols available to integrate the DERs. We used REST/JSON communication protocol to talk with EMS/SCADA for the site, which allowed us to get data from the DERs and send control signals to the DERs via Modbus TCP/IP. Other protocols that can be considered include MQTT.