Skip to main content

Prediction

Accurate photovoltaic power prediction, quickly layout future trends.

The function of predicting future data, also known as time series prediction or future trend prediction, is very important in various application scenarios. In the energy industry, it can be used for power prediction, targeting renewable energy sources such as wind power and photovoltaics, predicting future power generation based on weather data and historical output data.

How to Make Predictions?​

1. Click on New Prediction​

2. Edit Prediction Task Name​

You can edit a task name as you like.

3. Select Model​

Support selecting XMO-Time1.0, as well as models you fine-tuned. Note: If you delete a fine-tuned model from the model list, this model will be permanently deleted and cannot participate in predictions.

4. Fill in the necessary station information​

Such as longitude, latitude, installed capacity, etc.

5. Select Prediction Duration​

Prediction duration refers to the time range set during prediction, starting from the current time point, you can choose different time spans for prediction.

  • Short-term prediction (1 day)

    • Predict the situation for the next 1 day starting from today.
    • Supports viewing single-point power.
    • Suitable for situations that require immediate response, such as weather forecasts, electricity demand predictions, etc., to help users make quick adjustments.

  • Medium-term prediction (14 days)

    • Predict the situation for the next 14 days starting from today.
    • Supports viewing single-point power.
    • Suitable for planning in a relatively short period, such as market trend analysis, short-term financial forecasts, etc., providing some forward-looking information.

  • Long-term prediction (30 days, 90 days, 1 year, 3 years):

    • Predict the situation for the next 30 days/90 days/1 year/3 years starting from today.
    • Supports viewing daily average power and daily total electricity.
    • Suitable for strategic planning or long-term investment decision-making, such as annual budget preparation, long-term market trend analysis, etc., providing support for long-term goals.

6. Select Time Interval​

The time interval of the prediction result is the degree of subdivision of the predicted data in the time dimension, that is, the prediction result is published based on how long a unit of time.

  • You can choose 5 minutes, 15 minutes, 1 hour, or 1 day. For example: a time interval of 5 minutes means that the prediction result is published every 5 minutes.
  • If you choose short-term prediction, the time interval supports 5 minutes, 15 minutes, 1 hour, or 1 day. If you choose medium-term and long-term predictions, the time interval can only be 1 day.
  • If you are using a fine-tuned model, following the principle of "only bigger, not smaller," the available time intervals are greater than the time interval of your training file. For example, if you used a file with a 15-minute interval during fine-tuning, then when predicting, you cannot choose 5 minutes, you can only choose 1 hour or 1 day.

7. Select Prediction Time Zone​

  • After selecting the time zone, the time of the prediction result will be displayed based on that time zone.
  • The system's time will be consistent with the browser's time zone.

8. Click on [Start Prediction]/[Re-predict]​

After completing the above information, the system will automatically calculate the consumption of calls for this prediction. Clicking [Start Prediction] or [Re-predict] will be considered as your agreement to the consumption of calls and will be deducted from your account.

In the [Billing] module, you can view the daily and monthly API call situation.

In the official website [Personal Center - Call Monitoring], you can also view the actual usage of different products and the usage of fees.