Battery Not Dispatching fully

I have created a battery only model in SAM. I "tricked" SAM into modelling only the battery by setting losses to 100%. I choose to model a 8,000kWh battery with a c-rate charge and discharge of 0.25. Using the "Peak Shave: 1-day look ahead" I would have expected my model to fully dispatch in the Summer peak, however the most it dispatches in any one hour is 300kwh. Why is the battery not dispatching the full 2,000kWh?

Is there a way I can change the dispatch model to look at rates? I do not have a solar resource with this battery so how do I best optimize utility bill savings?

I also tried a manual dispatch model, setting the battery to discharge fully in super-peak. With this dispatch model the battery Is behaving the way I want it to because it is dispatching its full capacity, and shaving the peak everyday. However, the calculated Utility Bill savings are lower than the previous dispatch model because the battery is having to purchase charging power from the grid and in the second model the battery is being used much more.

With the Peak-shaving model the system is choosing to purchase from the grid instead of using the battery to arbitrage between the offpeak and peak periods. It's perplexing to me though why not utilizing the battery fully results in a higher utility bill savings. I would expect the the manual dispatch model I created, which fully dispatches the battery in peak hours and charges it in offpeak hours to result in a higher utility bill savings.

How can I best optimize my battery only model?

Good evening,

Thanks for the questions, I have examined your project file and think I can explain the behavior you see.

The peak-shaving dispatch model seeks to reduce peak power demand (kW) from the grid as much as possible over 24-hours, but it does not care about reducing energy (kWh). It also considers the peak demand so far in the month and doesn't bother discharging if the daily peak is less than the monthly peak in order to preserve battery life. This, with the assumption that reducing demand charges rather than energy charges will be the most economic use of the battery system. What this really points out is that SAM does not currently perform a cost-based optimization when choosing the dispatch strategy, rather, it uses a set of heuristics focused on reducing grid-power.

I'm not sure what you mean by changing the dispatch model to look at rates, do you mean performing multiple dispatches considering different utility rate structures? If so, you can certainly do this by changing the rate on the "Electricity Rates" page, but we don't have a way for you to do this easily in parametric form currently.

If you want to examine energy charges specifically, it will be better to use the manual dispatch controller as you suggested you tried, where you schedule to discharge the battery at specific times. In your case however, the most expensive energy is only $0.08/kWh, while the most expensive demand is $21/kW, so I don't believe it would be more economic to focus on reducing kWh, and you result in cycling the battery much more.

In general, energy arbitrage between peak and offpeak periods is not a very lucrative endeavor for a battery system due to round-trip efficiency losses and wear-and-tear on the battery, unless the difference in price between these periods is very high (perhaps >$0.20-$.30/kWh). The manual dispatch strategy isn't effective in this case, because you probably don't reduce the peak demand (kW) as much as the automated controller.

So, unfortunately SAM doesn't currently have a way to co-optimize battery usage to reduce demand and energy charges as much as possible, but can give you insight on what to do. If you have very high demand charges, I'd use the Peak Shaving control. If you don't have demand charges, you can use the manual dispatch controller. Also, consider that your battery dispatch will be dependent on the controls avaiable, and will likely not be operating optimally at all times, considering that PV production and electric load are variable quantities.

Hopefully that helped give some insight into what the model is doing. Please let us know if you have additional questions. I'd also encourage you to check out the following reports: www.nrel.gov/docs/fy15osti/64641.pdf , www.nrel.gov/docs/fy16osti/64987.pdf .

Best regards,
Nick

Thank you for your response. I understand now. Operationally, does software exist with an algorithm that can accurately predict the peak demand and dispatch the battery similar to SAM? And do you believe in real applications a peak shave model can "beat" the savings of a battery that discharges over the entire 6-hour peak period?

I have also tried the manual dispatch model forcing the battery to dispatch its full capacity over the 6-hour peak period. Results were quite good. However, I would like to change the Manual Dispatch model to only dispatch according to my given schedule in the Summer months. How can I do this?

Also, I am concerned SAM is not calculating the demand charge correctly for my system. According to Nevada Power's GLS-3 tariff, the demand charge should be the highest KW value for that given month times the demand charge plus the Facilities Charge times the highest KW value over the past rolling 12 months.

SO for instance in the month of August the Demand Charge should be calculated like below. The SAM value is much higher than what the calculated value should be, and the SAM value does not match this customers actual bill from August 2015. Why is the Demand Charge so different?

August Peak KW 5,423
Demand Charge Summer On-Peak $17.83
$96,699

Previous 11 month Peak KW 5,710
Facilities Charge all KW $3.20
$18,273

Total August Demand Charge $114,972
SAM Value $143,006
Difference $28,034

Dear Shannon,

Regarding operational systems predicting peak demand, I unfortunately do not know too much. There are advanced controllers certainly, and there are methods of forecasting demand, I just don't know how it fits together. When we model the dispatch, we are making several assumptions, including 1.) that a controller exists that could be programmed with this kind of dispatch 2.) that the user has access or ability to accurately predict their PV production and electric load. Those are clearly two large assumptions, illustrating that modeled results should be carefully examined and compared to real-world capabilities. I believe that a peak shaving model will almost always be more economic that a battery which discharges over an entire period, simply because the value of demand charges is so much greater than the value of energy arbitrage. Of course, this depends heavily on the utility rate structure you are working with.

To modify the manual dispatch control to only dispatch in the summer months, select the months/hours you'd like to change and then press the keyboard number that corresponds to the period. For example, I've attached a file that changes the dispatch to period 4 for all months except June-September, and programmed period 4 to do nothing but charge the battery from excess PV.


Regarding the demand charges, I believe the issue is in the non-peak periods. In August, it looks like there are three periods as you have it programmed:

- Period 1: $21.03/kW
- Period 2: $6.1/kW
- Period 3: $3.2/kW

And you have no facility demand charge programmed, which should be under "Demand Rates by Month with Optional Tiers" The way SAM calculates the demand charge would then be:

[Peak kW in Period 1]*[$21.03/kW] + [Peak kW in Period 2]*[$6.1/kW] + [Peak KW in Period 3]*[$3.2/kW] + [Peak kW in month]*[Facility demand charge $/kW].

I.e, we don't do a rolling 12-month maximum on the facility demand charge to my knowledge. I believe as outputs we should also consider adding the peak kW in each period/tier to make this easier to reproduce. Does this calculation make sense? I can take a closer look at the load profile and utility rate tariff if needed. Please send me an updated SAM file if you're looking at something newer than what I attached.

Thanks,
Nick