How does Arkreen Network verify the trustworthiness of Proof of Green Data?

In order to avoid miners reporting fake power generation data to the network, Arkreen Network uses a set of novel methods for data authenticity judgment. The exact method depends on the category of the miners.

For Energy Generation Category

The judging method is as follows:

Firstly, according to the external weather and photovoltaic oracle data, it is judged whether the power generation data is within the legal range. If the data exceeds the upper limit of the theoretical power generation capacity, the data trustworthiness score will be 0, and the miners will not be able to obtain reward income.

Secondly, according to the average real-time power of the miner, the real-time power data obtained by sampling the data statistics window is normalized, and the real-time power vector is obtained.

Thirdly, find other nearest miners (the witness) within a certain distance around the miner, and calculate the cosine similarity between the target miner and the real-time power vector of each surrounding witness through the following method:

$$\text{cosine similarity} = \frac{M\cdot W}{\left\| M \right\|\left\| W \right\|}=\frac{\sum_{i=1}^{Z}M_iW_i}{\sqrt{\sum_{i=1}^{Z}M_i^{2}}\sqrt{\sum_{i=1}^{Z}W_i^{2}}}$$

Where,

M is the power vector within the statistics window for the target miner.

W is the power vector within the statistics window for each witness.

Z is the sample points, i.e., the total number of the Proof of Green Data reports sent by the miner within the statistics window.

If the cosine similarity to one witness is higher than a threshold (currently 0.5), this target miner gets a vote. The more votes the target miner gets from the surrounding witnesses (up to 6 witnesses), the higher trustworthiness score it gets.

The trustworthiness score T is calculated as:

$$T=sigmoid(votes)$$

The score T will be taken part in the reward weight calculation.

For Energy Consumption Category

For energy consumption category, the data authenticity could be assessed by analyzing the energy curve. The user should power up his appliance (e.g. air conditioner) with the Smart Plug and keep it all the time. The energy consumption curve is collected and compared against typical appliance's curve feature to evaluate the trustworthiness score.

Energy Consumption Curve Assessment