Dashboard sharing #94
Replies: 5 comments 19 replies
-
|
Thanks very much to Geoffrey Coan for the original chart..! I love this chart, can't help myself and been tweaking. I am off grid so battery percent is important as once I get beyond solar charging slows down and even stops so to make sense of the solar in I need to see that. I also found that my solar has a very high reporting rate ie 30 times a second... So I've slowed down the chart updates as it was constaintly busy. I also tweaked so that the solar and battery doesn't display past the now line. My use case is unusual so expect most people may not want so sharing in case there are others.
|
Beta Was this translation helpful? Give feedback.
-
|
I am getting an interesting time "shift" using @gcoan's example chart. To fix it, and smooth out the lumpy appearance I add two options to the area series. Today was a brilliantly sunny Winter day, and with the shift of the forecasted values they almost match for much of the day. I did notice that the interpolated "power now"/+30mins/+60mins were off by this same shift, so there might be a code improvement needed there (which I will look into). Before: After: |
Beta Was this translation helpful? Give feedback.
-
|
Just noticed that "monotoneCubic" does not appear to work, despite the doco. It's using "straight"... |
Beta Was this translation helpful? Give feedback.
-
|
I am trying out this patch for the spline values to account for the fifteen minute shift... (edit) And this... |
Beta Was this translation helpful? Give feedback.
-
|
That's an impressive following of the solcast prediction Steve. We're in summer at the moment and at best I get about half a day following the prediction, and then half a day of the solar gen going all over the place. Multiple inverters, multiple roof orientations and so much generation capacity that it pushes the grid voltage above limits all conspire to never give me such a good graph But back to your graph, I hadn't noticed, but I too see the forecast stepping up at 30min segments. The issue you are seeing must be down to the granularity of the data:
Could replace the forecast figure with power now predicted sensor which is interpolated, but that'll only give the historic curve not the forward curve. Or shift by 15 minutes as you've done to improve the "lining up" of the different data series' |
Beta Was this translation helpful? Give feedback.
-
|
The only reason for the dips in production was feed-in being limited to 5kW after battery filled and afternoon shading. Once the battery had filled I turned on the pool pump and chucked some clothes in the dryer (perverted, I know, because great line drying day 😂), but still couldn't use all the generation. A fifteen minute shift makes a lot of sense to me, given the 30-minute averages, so I'm glad you concur. I'll monitor over the next few days, as tomorrow is forecast to be a Winter cracker too. 
|
Beta Was this translation helpful? Give feedback.
-
|
Slight change of plan for the splines code patch. I've instead shifted all the spline arrays by fifteen minutes, achieving the same result, but making more sense in code. I won't bore you with the patch, as it's equivalent to doing what I first thought of. The Apex chart still needs Testing continues. |
Beta Was this translation helpful? Give feedback.
-
|
Here's an alternative chart that I've cooked up. (Edited so many times because issues! This @ 5/09 appears stable and tracks at the moment. 😂🤦🏻) It has a variable x-axis that changes based on sunrise/sunset throughout the year, and also throughout the day, and is based on 'Lovelace Card Templater' from HACS, 
And at 21:09. 
Full auto-span, 23:38. 
Just past midnight.. 
It's not without its idiosyncrasies, flickering when it updates. Not sure why, but I'll work it out. Template sensor required: template:
- trigger:
- platform: time_pattern
minutes: "/5"
sensor:
- name: "Five Minute Update"
unique_id: "five_minute_update"
state: "{{ now().minute }}"
unit_of_measurement: "Minutes"You probably want to exclude this sensor from the long-term storage in recorder:
purge_keep_days: 14
exclude:
entities:
- sensor.five_minute_updateAnd the actual card. type: custom:card-templater
card:
type: custom:apexcharts-card
graph_span_template: |-
{% set sunrise = as_datetime(states('sensor.sun_next_rising')) %}
{% set sunset = as_datetime(states('sensor.sun_next_setting')) %}
{% if sunrise != none and sunset != none %}
{% set compressed = (
(as_local(sunset).hour - as_local(sunrise).hour) + 1 +
(
(
max(now().hour - as_local(sunset).hour, 0)
)
if now().hour > as_local(sunrise).hour
else (as_local(sunrise).hour - now().hour + 2)
) * 2
) %}
{% set compressed = compressed + 1 if now().time() > as_local(sunset).time() else compressed %}
{{ compressed if compressed <= 24 else 24 }}h
{% else %}
24h
{% endif %}
span:
start: day
offset_template: |-
{% set sunrise = as_datetime(states('sensor.sun_next_rising')) %}
{% set sunset = as_datetime(states('sensor.sun_next_setting')) %}
{% if sunrise != none and sunset != none %}
+{{
(
as_local(sunrise).hour - max((now().hour - as_local(sunset).hour) + 1, 0)
)
if now().hour > as_local(sunrise).hour else now().hour
}}h
{% else %}
+0h
{% endif %}
header:
title: Solar
show: true
show_states: true
colorize_states: true
apex_config:
chart:
height: 300px
tooltip:
enabled: true
shared: true
followCursor: true
yaxis:
- id: capacity
show: true
opposite: true
decimals: 0
max: 100
min: 0
apex_config:
tickAmount: 10
- id: kWh
show: true
min: 0
apex_config:
tickAmount: 10
- id: header_only
show: false
series:
- entity: sensor.my_home_solar_power
name: Solar power (5 min avg)
type: line
stroke_width: 2
float_precision: 2
color: Orange
yaxis_id: kWh
unit: kW
extend_to: now
show:
legend_value: true
in_header: false
group_by:
func: avg
duration: 5m
- entity: sensor.solcast_pv_forecast_forecast_today
name: Forecast 10-50%
color: LightGrey
opacity: 0.5
stroke_width: 0
type: area
time_delta: +15min
curve: monotoneCubic
extend_to: false
yaxis_id: kWh
show:
legend_value: false
in_legend: true
in_header: false
data_generator: |
return entity.attributes.detailedForecast.map((entry) => {
return [new Date(entry.period_start), entry.pv_estimate];
});
- entity: sensor.powerwall_charge_actual
name: Battery
yaxis_id: capacity
type: line
stroke_width: 1
float_precision: 2
color: DarkGreen
extend_to: now
group_by:
func: avg
duration: 1s
show:
in_legend: true
legend_value: false
in_header: false
- entity: sensor.solcast_pv_forecast_forecast_today
name: Forecast 10%
color: White
opacity: 1
stroke_width: 0
type: area
time_delta: +15min
curve: monotoneCubic
extend_to: false
yaxis_id: kWh
show:
in_legend: false
in_header: false
data_generator: |
return entity.attributes.detailedForecast.map((entry) => {
return [new Date(entry.period_start), entry.pv_estimate10];
});
- entity: sensor.solar_production_today
yaxis_id: header_only
name: Actual today
stroke_width: 2
color: Orange
show:
legend_value: true
in_header: true
in_chart: false
- entity: sensor.solcast_pv_forecast_forecast_today
yaxis_id: header_only
name: Forecast today
color: Grey
float_precision: 1
show:
legend_value: true
in_header: true
in_chart: false
- entity: sensor.solcast_pv_forecast_forecast_today
attribute: estimate10
yaxis_id: header_only
name: Forecast today 10%
color: Grey
float_precision: 1
opacity: 0.3
show:
legend_value: true
in_header: true
in_chart: false
- entity: sensor.solcast_pv_forecast_forecast_remaining_today
yaxis_id: header_only
name: Forecast remaining
color: Grey
show:
legend_value: true
in_header: true
in_chart: false
entities:
- entity: sensor.five_minute_update |
Beta Was this translation helpful? Give feedback.
-
|
I think I need to inset this to modify |
Beta Was this translation helpful? Give feedback.
-
I'm wondering what the need for the 5-minute template sensor is and whether the graph is simpler (and still works) without it? Currently on a mission to try to reduce the size of my HA database, removing/trimming unnecessary sensors as I keep running out of space on my Google drive to keep sufficient copies of the backup. The raw db is 2.7Gb (~900Mb compressed) and a lot of it is due to noisy sensors - when I had that period that HA stopping communicating with the world the backups were ~300Mb! |
Beta Was this translation helpful? Give feedback.
-
|
It irked me, the whole five minute sensor thing. I could not find a realiable way to do it just in the card-templater. Every time I tried it flickered madly eventually crashing the front-end, so if you have a suggestion then I am all ears. (This may be a bug with the card-templater...) The need for the five-minute update? The chart will stop updating if updates are tied to solar actual and battery charge level. Scenario: 'Zero' battery, zero solar generation = no chart update, because values do not change over time. That was how I originally rigged it, and it displeased me. |
Beta Was this translation helpful? Give feedback.
-
|
Ah OK, you want the chart to refresh itself periodically and grow to the right as time advances even if there is no solar generation or battery change, e.g. displaying it on a wall tablet display. Yes you probably need a sensor that changes to cause the refresh. For me I don't (yet) have a tablet, only look periodically via the app and website. You could exclude the 5 minute sensor in the recorder configuration.yaml to stop it filling up the database ... |
Beta Was this translation helpful? Give feedback.
-
|
(Oops. I have actually done that LTS exclusion, but didn't mention it... Now mentioned.) |
Beta Was this translation helpful? Give feedback.
-
|
This is a little OT because I can't really share it - none of this is really portable to HA, but I first started using Solcast back in the day when I thought python was a snake. I still store all my Powerwall data at six second resolution in an SQL Server database, and, up until recently (because I've changed to a tariff where I no longer need to forecast consumption or insolation), was using historical load data, Solcast forecasts, and my own logic to forecast how much to charge, and when. When I first started forecasting, I discovered that I'd get it wrong, and in trying to work out why, I discovered that forecasting too early was problematic, because (surprise surprise) forecasts change. Nonetheless, once the code is written and running, it keeps running, and as a result, I have a database going back years with the earliest forecast for each half hour, and the latest. Here's what it shows for today in Melbourne:
As you can see, expected insolation has fallen off a cliff for today. And here's what the dashboard shows for yesterday, today, and the next four days.
I eventually refined my algorithm to work out how late it could make its decision based on the remaining time window until peak, the rate that I could charge my battery sufficiently to ride through peak (and, ideally off peak until solar generation was going to exceed load the next day, but failing that, at least peak), and not start charging from the grid until as late as possible, with some margin for changing forecasts, especially if forecast insolation was dropping. It got to the point that I never paid for peak power, and never over-charged and spilled FIT either, and the algorithm was, in the end, incredibly simple - total load to breakeven point (insolation exceeds load) less remaining insolation = shortfall. That's how much I needed to charge. Charge rate was known, and with a few tweaks, the algorithm would push up the battery reserve level to try to make the shortfall = 0 by the beginning of peak. AI and machine learning be damned. Now my charge cost is $0.00 for two hours a day, which is enough to add 10kWh every day - should I need to. No need for forecasts until that tariff is no longer available. However, with the way the market is in Victoria, that's unlikely to go away any time soon. |
Beta Was this translation helpful? Give feedback.
-
|
Hi all, So, was wondering if anyone could help me out with tomorrows forecast? Just today I added my house load to the OG graph to test out. (I know it was a terrible solar day today here in Sydney) |
Beta Was this translation helpful? Give feedback.
-
|
@autoSteve, yes you're absolutely right. i will let it go :) @gcoan, thank you for the code. i ended up following your lead and got rid of the double forecasts. Now my graphs page looks like this. and don't mind the others as im testing and not sure if figures are right |
Beta Was this translation helpful? Give feedback.
-
Ah yes, sorry, I have today and tomorrow 10% forecast helper sensors, just to make it easier to display on dashboards and graph the history (which Steve will hate me doing). I'd forgotten that I was using that in the chart Your graphs look fine, nice idea to show load power alongside solar forecast. Very much the Australian summer, I'm forecasting 2.1kWh total generation today and much the same for the next few days. Gloomy winter weather here in the UK |
Beta Was this translation helpful? Give feedback.
-
|
I was super jealous of you guys in the Northern five months ago, Geoff. Now feeling super smug and wearing shorts with the pool at 29°C. Switching to three free power hours between 11AM and 2PM (thanks @BJReplay) mean that I have been paid for power use since the start of November, not billed. Could have been October too, but I took a while to get my head around the whole VPP thing, where I need to do the scheduling. The dramatic battery drop is me dumping ~13kWh of stored energy that I won't need to the grid in short order at 'super' peak payback rate. (It is annoying that doing so seems to curtail the micro-inverters. Ah, well. What evs. Enphase and Tesla kind-of play nicely together. Mostly.) 
|
Beta Was this translation helpful? Give feedback.
-
|
I have finally got around to less dramatic dumping (half the storage), thanks, @autoSteve, though I can probably go harder, only third day of trying. Yesterday was a combination of a sunny day, putting 11kWh into the EV from solar (and a little bit of battery / grid in the morning), stopping when excess solar !> charging load, then dumping when excess solar !> household load until battery SoC <= 70%. Right now, Solar > household load and minimum SoC was 33, so I probably could have dumped until 50%.
|
Beta Was this translation helpful? Give feedback.
-
|
Further enhancement to my chart above, the x-axis labels by default are formatted as DD MMM, e.g. 14 Dec I can't remember the date but the day of the week is more useful to me, so I changed the x-axis format: Can use any format string, e.g. ddd dd MMM will display Mon 14 Dec but the labels are centred on the start of the day, so adding more text results in them shifting further to the left. Despite looking at the manual I couldn't find a way to centre-align the labels under the data, so shortened the label instead |
Beta Was this translation helpful? Give feedback.
-
Hi all, I thought it would be helpful to have a dedicated thread for sharing UI's.
I've got a simple table I really like showing worst (P10) and estimated forecast.
Use this code to create a template
Then you need to install Flex Table Card from HACS and here is the YAML
Matt
Beta Was this translation helpful? Give feedback.
All reactions