Battery charge to system spike.

All/Paul,
I am modelling 2 wind cases and 1 solar case with a  Generic + battery in a BTM system.     I have set the battery dispatch so it can only charge from the system (charge from grid not enabled) when system>load and discharge only when load >than system (peak shaving button selected).  I have entered a very small sell price in the sell column of the energy charges (buy/sell) all in order to get the economics to calculate a difference in the bill with and without system.  files attache

The individual system cases run fine with logical outputs but when I simulate the Battery plus system, the LCOE calcs are absurdly high and the bill with the system is more than the bill without the system (huge negative savings).  

Strangely, there also appears to be a single spike (on Sept 30th) when the battery sends electricity to the grid  for one hour.  But at that same time period, the system is sending power to the grid (i.e. surplus power over load is being generated by the system) and there is a non zero load in that same time period (in fact no zero load conditions.   I have a sneaking suspicion that the LCOE calcs are only picking up this power spike as the total KWH divisor.

Any suggestions as to how to remedy? 

Thanks Paul,
Currently, existing Diesel gensets, PV farm and  WT's supply the load.  Gensets are dispatched to generate a min of 500 kW for efficiency, Solar comes in as generated and WT are operated to make up the rest of the load.     I wish to model additional new WT's, Solar, and batteries with dispatch strategy of renewables first (and only they charge the battery) and the Gensets making up shortfalls when  WT, PV, (and Battery if  competitive) don't meet load.  I want to minimize the LCOE for new equipment additions (of course) and thus would benefit from having the LCOE figures being sensible.

I had tried the Residential BTM model thinking to model the gensets as grid purchased power.  I have the heat rates and fuel costs for the gensets so I input the $/kWh costs at different loads as the tiered energy buy rates, but was not sure how to get (or if SAM can calculate) the hour by hour grid (genset) cost on that basis as I think it only does the tiering on a daily/ monthly basis?. (but being able to do this would be a nice feature).  The other issue I have, is surplus system power being sold back to the grid.  I set the sell rates very low, which has the benefit that SAM does show a difference in bill with and without system (if sell rate is zero - there is no difference between the bills)  but the total system power generation is now used as the basis for the LCOE calculation when in fact I would want to be curtailing WT generation when system greater than load. Not sure how to do this.

Still, SAM does track the Grid supplied power to system and system power grid,  so in the interim if have been downloading these vectors and the load and calculating in excel a truncated system production and working out the correct costs for grid (genset) supplied power in the System and No system cases, and thus the savings from the system. and whacking that back into the exported cash flow model.   

Any advice you can give about a better way to set up these configurations to reduce the extracting and external manipulation would be much appreciated. 

Additional Comments

The strange things (for me) are:
1.  why the battery discharges to the grid when the battery is set to only discharge when Load is greater than system (why is it not meeting load first?)  and if so, why only once per year for 9kW?

2. Why the savings due from the system are (so) negative?   I find it hard to imagine how taking less power from the grid could be so much more expensive than taking it all from the grid; particularly when there is only a 50% difference between the highest and lowest price tier.