Power prediction method and apparatus, and device

1 . A method comprising:

obtaining evaluation metrics of models in a model pool, wherein the evaluation metrics indicate precisions of the models, wherein the evaluation metrics are based on an interval error value used during a power prediction, and wherein the interval error value is based on a summation of deviation values between a predicted power value of the models and an actual power value of the models in a plurality of statistical periods in a electricity price time interval;

selecting, from the model pool and based on the evaluation metrics, a first model for a power prediction of an electrical unit in a data center;

performing the power prediction using the first model to obtain a result;

presenting the result of the power prediction;

determining a charge/discharge policy based on the power prediction and based on electricity prices of a mains power, wherein the charge/discharge policy charges an energy storage system from the mains power when the electricity prices are relatively low and discharges from the energy storage system to the electrical unit when the electricity prices are relatively high; and

implementing the charge/discharge policy.

2 . The method of claim 1 , further comprising predicting future power distribution of the electrical unit using the first model based on a historical power distribution of the electrical unit, wherein the historical power distribution indicates power in at least a first statistical period before a current moment, and wherein the future power distribution indicates power in at least a second statistical period after the current moment.

3 . The method of claim 1 , wherein the evaluation metrics comprise error values, and wherein the method further comprises further selecting the first model based on the error values, wherein the first model has a first error value that satisfies a preset condition.

4 . The method of claim 3 , wherein the preset condition is that the first error value is lower than a preset threshold or that the first error value is lowest among the error values.

5 . The method of claim 1 , wherein selecting the first model comprises:

updating the models in the model pool based on the evaluation metrics to obtain updated models; and

selecting the first model from the model pool based on the evaluation metrics of the updated models.

6 . The method of claim 1 , wherein the evaluation metrics are further based on a single-point error value used during the power prediction.

7 . The method of claim 1 , wherein the model pool comprises two or more models.

8 . The method of claim 1 , wherein the electrical unit is a set of electrical devices comprising at least one electrical device, electrical devices in at least one rack, or electrical devices in at least one data center.

9 . The method of claim 1 , wherein presenting the result of the power prediction comprises presenting, through a graphical user interface, the result of the power prediction.

10 . A device comprising:

a memory configured to store instructions; and

one or more processors configured to execute the instructions to:

obtain evaluation metrics of models in a model pool, wherein the evaluation metrics indicate precisions of the models, wherein the evaluation metrics are based on an interval error value used during a power prediction, and wherein the interval error value is based on a summation of deviation values between a predicted power value of the models and an actual power value of the models in a plurality of statistical periods in a electricity price time interval;

select, from the model pool and based on the evaluation metrics, a first model for a power prediction of an electrical unit in a data center;

perform the power prediction using the first model to obtain a result;

present the result of the power prediction; and

determine a charge/discharge policy based on the power prediction and based on electricity prices of a mains power, wherein the charge/discharge policy charges an energy storage system from the mains power when the electricity prices are relatively low and discharges from the energy storage system to the electrical unit when the electricity prices are relatively high.

11 . The device of claim 10 , wherein the one or more processors are further configured to execute the instructions to predict future power distribution of the electrical unit using the first model based on a historical power distribution of the electrical unit, wherein the historical power distribution indicates power in at least a first statistical period before a current moment, and wherein the future power distribution indicates power in at least a second statistical period after the current moment.

12 . The device of claim 10 , wherein the evaluation metrics comprise error values, and wherein the one or more processors are further configured to execute the instructions to select, the first model based on the error values, wherein the first model has a first error value that satisfies a preset condition.

13 . The device of claim 12 , wherein the preset condition is that the first error value is lower than a preset threshold or that the first error value is lowest among the error values.

14 . The device of claim 10 , wherein the one or more processors are further configured to execute the instructions to:

update the models in the model pool based on the evaluation metrics to obtain updated models; and

select the first model from the model pool based on the evaluation metrics of the updated models.

15 . The device of claim 10 , wherein the evaluation metrics are further based on a single-point error value used during the power prediction.

16 . The device of claim 10 , wherein the model pool comprises two or more models.

17 . The device of claim 10 , wherein the electrical unit is a set of electrical devices comprising at least one electrical device, electrical devices in at least one rack, or electrical devices in at least one data center.

18 . The device of claim 10 , wherein the one or more processors are further configured to execute the instructions to present, through a graphical user interface, the result of the power prediction.

19 . A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores instructions, and wherein the instructions when executed by a processor of a device, cause the device to:

obtain evaluation metrics of models in a model pool, wherein the evaluation metrics indicate precisions of the models, wherein the evaluation metrics are based on an interval error value used during a power prediction, and wherein the interval error value is based on a summation of deviation values between a predicted power value of the models and an actual power value of the models in a plurality of statistical periods in a electricity price time interval;

select, from the model pool and based on the evaluation metrics, a first model for a power prediction of an electrical unit in a data center;

perform the power prediction using the first model to obtain a result;

present the result of the power prediction; and

determine a charge/discharge policy based on the power prediction and based on electricity prices of a mains power, wherein the charge/discharge policy charges an energy storage system from the mains power when the electricity prices are relatively low and discharges from the energy storage system to the electrical unit when the electricity prices are relatively high.

20 . The non-transitory computer-readable storage medium of claim 19 , wherein the evaluation metrics are further based on a single-point error value used during the power prediction.