计算 20 分钟平均值和最佳 20 分钟平均值

Question

I'm trying to learn Python, with a project where I'm reading data from a bike power meter. Right now I'm just calculating the average power from start to finish, by adding each power reading to a total sum variable, and dividing it with the number of readings.

I'd like to calculate the average power for 20 minutes, and if possible, keep calculating the 20 minute average each 30 seconds after the first 20 minutes, comparing to the previous value and storing it if it's higher than the last, so in the end I could have the higher 20 minute average power that I held during a one hour workout, no matter where it happened in that hour, for example.

Data from the power meter is event based, as far as I can tell it's not a regular intervals, but definitely once a second or faster. This is the base of my code so far:

def average_power(power, count):
    global PM1_sumwatt
    global PM1_avgwatt
    PM1_sumwatt = PM1_sumwatt + power
    PM1_avgwatt = PM1_sumwatt / count
    PM1_avgLog = open(PM1_avgFile, 'w')
    PM1_avgLog.write("<div id=\"pwr\"> %d W</div>" % (PM1_avgwatt))
    PM1_avgLog.close()

def power_data(eventCount, pedalDiff, pedalPowerRatio, cadence, accumPower, instantPower):
     global PM1_avgcount
     if WCORRECT1: calibpower = instantPower+(CF1w)
     else: calibpower = instantPower*(CF1x)
     power_meter.update(int(calibpower))
     PM1_avgcount = PM1_avgcount + 1
     average_power(int(calibpower), PM1_avgcount)


power  = BicyclePower(readnode, network, callbacks = {'onDevicePaired': device_found,
                                              'onPowerData': power_data})


# Starting PM readings
power.open(ChannelID(PM1, 11, 0))

Not quite sure how to tackle this! Any help or pointer is much appreciated!

Answer 1

You can use pandas dataframe to store the power output for each instance.

Considering that you receive a value each 30 second, you can store them all in data frame.

Then calculate a 40 data point moving average using rolling function in python.

Take the max value you get after the rolling function, this would be your final result.

refer this for doc : https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.rolling.html

Answer 2

if you are reading data in real time, I assume you are reading the data in a while loop:

sum = 0
number_of_readings = 0
while True:  # forever
    new_value = input()  # here I read from the console input
    sum += new_value
    number_of_readings += 1
    average = sum/number_of_readings 
    print(average)

Here I type a number in the console and press enter to simulate your bike power meter.

>>> 1
1.0
>>> 3
2.0
>>> 4
2.6666666666666665
>>> 2
2.5

---

Now, if you wants to make a moving average, you must store the readings that you wants to average. This is because you want to remove them later, when they will be too old. A list is perfect for that:

Here is a solution averaging the last n readings:

n = 2
Last_n_readings = []
while True:  # forever
    # add a new reading
    new_value = int(input())  # here I read from the console input
    Last_n_readings.append(new_value)
    # discard an old one (except when we do not have enough readings yet)
    if len(Last_n_readings) > n:
        Last_n_readings.pop(0)
    # print the average of all the readings still in the list
    print(sum(Last_n_readings) / len(Last_n_readings))

Which gives:

>>> 1
1.0
>>> 3
2.0
>>> 4
3.5
>>> 2
3.0

Note that lists are not very efficient when removing elements near the start, so there are more effective ways to do this (circular buffers), but I try to keep it simple ;)

You could use this by guessing how many readings/seconds you have and choose n so you average over 20 minutes.
If you want to truly average all the result which are less than 20 minutes ago, you need to not only record the readings but also the times when you red them, so you can remove the old readings wen they get more than 20 minutes old. If this is what you need, tell me and I will expand my answer with an example.