I would like to know the entire power consumption of the keyboard.
The question here really is:
What do you want to know, exactly?
If you want to give a number for how much power the device itself consumes, then measuring the red wire's current will be sufficient, and accurate. (Your multimeter might still be too slow for the fast changing current consumption of digital devices, but let's assume that's not an issue here, and there's plenty of low-pass filtering by bypass caps on the keyboard's side; still when talking about such minor differences, you simply mustn't neglect the fact that power will change much faster than your multimeter allows you to observe).
The D+/D- wires are the data-carrying lines. These are half-duplex, which means, some times they're driven by the host computer, and some times, they are driven by the device. Let's first consider the computer->device times only, since the power used to transmit in the other direction needs to be taken from the power supply on the V+ wire, so you'd cover this power consumption already. You must make sure that you're not measuring while the device is transmitting, that would count things twice (for that, your measurement device would need to "understand" which device is currently "talking", and that would be a logic/bus analyzer, not a multimeter). More about your measurements below.
But what does "driven" mean, here?
It means that an IC (the "PHY", often integrated into a larger functional unit) changes the voltage on these lines. Since USB is based on voltage signals, not current flowing, in (superficial) theory, there's no energy flow from one side to the other.
In practice, you indeed do need energy to change the voltage on such a line – and that's really the effect of how signal transmission on lines works. The question here really becomes: Do you want to count the power spent on essentially sending a wave into a cable as a power consumption of your device, or not? That power is not consumed within the device, and how much power is sunk depends on how strongly your driver drives the line – there's quite some leeway here, so it wouldn't be fair to "put it on the tab"¹ of the receiving receiving end. On the other hand, the power is spent on communicating with the device.
So, now about your measurements on the D+/D- lines: there's something wrong with them. And I simply think your multimeter is the wrong measurement device for this.
D+/D- are differential lines, so that one should have (theoretically perfectly) one line should simply have the inverse voltage of the other line. Now, since power flow is a directive entity, even when driving, it's always going to be the case that, as current flows in oppsite directions on these two wires, the average current should be zero, especially since USB is designed to avoid the opposite.
So, while power goes into the active driver, you couldn't determine the power flow by looking at average current, squaring that and assuming some resistance. You need to calculate the instantaneous power, calculate the absolute value and average that.
Also, USB works at >1 MHz, and that's far above what any multimeter can measure (it wouldn't be called multimeter but "RF power analyzer", "oscilloscope", "network analyzer"..., depending on what exactly it does). So, whatever you're measuring with your multimeter on these lines is very likely simply not representative on what happens on the lines.
¹ just to keep within Tony M's pub scenario :)