Solar Model vs. Actual Solar Generation

When I wanted to know if our solar PV system would make any money for us under the Feed In Tariff program through the Ontario Power Authority, i could either listen to sales reps or calculate it. Some sales reps "guaranteed" that we'd make money, others said that it would pay for itself without over promising. I was skeptical of the former guys even though I knew our system would pay under the FIT program with $0.802/kWhr. I used RETScreen and PVWatts to do my modelling. I posted my results in an earlier post. Essentially what I found was that each month saw good generation, higher in the summer than December.
I've plotted my cumulative generation against the RETScreen model of our 4.9 kW system. What I found was the generation was greater than modelled during the autumn of 2010, much lower than the model during the winter, and it seems that March and April are fairly closely modelled.
The plotted data below shows several data sets :

• the blue curve is the results from the datalogger. The PV system was online on 24 August, but I didn't have the datalogger until 01 November so I simply read the meter at the end of each day until the datalogger was installed
• the open brown circles are the Guelph Hydro electricity meter readings
• the RETScreen model is the purple curve with triangles
• our household consumption [load meter] are the red closed circles

The chart is plotting cumulative kWhr. Broad observations of model accuracy should be made on multiple years of data, but we can still see how the model compares. So, what do we see???

• over the autumn our PV system was generating about 2/3rds of our consumption but after the dark and snowy days after 10 December our electricity consumption remained constant or slightly increased, however our panels virtually flat-lined. Our panels cover much of our south-facing roof but it is still not enough to supply our home's consumption. Technically we are not using any of the electricity we are producing - it is metered and sent directly onto the grid. But it would be nice to know that we are contributing what we are using albeit at different times of the day. 
• the panels generated more than the house consumed between 17 March and 27 April. The panels generated 745 kWhrs and the house load was 595 kWhrs [25.2% generation over consumption].
• the Guelph Hydro meter continues to be slightly below the datalogger. I wish they were the same! It could be that the datalogger is recording total generation which is occasionally greater than the inverter rating of 5 kW, whereas the Guelph Hydro meter is recording what is actually going out onto the grid. Or maybe not! Maybe they are measuring the same thing but one of them is wrong! Fortunately the difference is small.
• the RETScreen model seems to be approximating the generation starting around 15 February. Indeed the PV generation may be very slightly higher than the model.
• my financial calculations were based on RETScreen and PVWatts models, after mid-December we are less than my predicted generation [and therefore revenue]. I hope we have a nice bright summer to get close to the model again so we make our forecasted revenue!

Our loan costs from 24 August 2010 to 27 April 2011 was $2,646 and our revenue was $2,354. This is a deficit, however we also went from August to the darkest part of the year and we are just coming back into the sunnier season. I am confident that the low revenues of December and January will be more than offset by this summer's generation. We only need 30 days of over 20 kWhr to make up the $310 deficit, so I am not worried.