Method and system for automatic machine learning-based prediction of new energy power with cloud-edge collaboration
A method and a system for automatic machine learning-based prediction of new energy power with cloud-edge collaboration are disclosed. The method includes: obtaining, in response to a power prediction demand for a target new energy station, future numerical weather prediction data of the target new energy station in a future period and historical output power of a historical period corresponding to the future period; selecting, based on missing of the future numerical weather prediction data and a data amount of the historical output power, a target power prediction model corresponding to the target new energy station; and adjusting the target power prediction model according to the target working mode, and predicting, by the adjusted target power prediction model, a target output power of the target new energy station in the future period, based on the future numerical weather prediction data and the historical output power.
1. A method for automatic machine learning-based prediction of new energy power with cloud-edge collaboration, which is executed by an edge server, the method comprising:
obtaining, in response to a power prediction demand for a target new energy station, future numerical weather prediction data of the target new energy station in a future period and historical output power of a historical period corresponding to the future period;
selecting, based on missing of the future numerical weather prediction data and a data amount of the historical output power, a target power prediction model corresponding to the target new energy station from a set of power prediction models, each optional power prediction model in the set of power prediction models being trained and delivered by a cloud server based on an automatic machine learning algorithm;
determining, based on the power prediction demand, a target working mode of the target power prediction model, the target working mode comprising a prediction business object, a prediction type, and a prediction time scale; and
adjusting the target power prediction model according to the target working mode, and predicting, by the adjusted target power prediction model, a target output power of the target new energy station in the future period, based on the future numerical weather prediction data of the future period and the historical output power of the historical period corresponding to the future period.
2. The method according to claim 1 , wherein selecting, based on the missing of the future numerical weather prediction data and the data amount of the historical output power, the target power prediction model corresponding to the target new energy station from the set of power prediction models comprises:
determining, when the target new energy station is identified as a non-newly built station or non-expanded station based on the amount data of the historical output power, a data missing level of the future numerical weather prediction data based on the missing of the future numerical weather prediction data; and
selecting, based on the data missing level, the target power prediction model from the optional power prediction models pertaining to the target new energy station in the set of power prediction models.
3. The method according to claim 2 , wherein predicting, by the adjusted target power prediction model, the target output power of the target new energy station in the future period, based on the future numerical weather prediction data of the future period and the historical output power of the historical period corresponding to the future period, comprises:
complementing, if the data missing level is a scattered missing level, missing data in the future numerical weather prediction data, and predicting, by the adjusted target power prediction model, the target output power of the target new energy station in the future period, based on the historical output power and the complemented future numerical weather prediction data; and
predicting, by the adjusted target power prediction model, the target output power of the target new energy station in the future period, based on the historical output power, if the data missing level is a continuous missing level.
4. The method according to claim 1 , wherein selecting, based on the missing of the future numerical weather prediction data and the data amount of the historical output power, the target power prediction model corresponding to the target new energy station from the set of power prediction models, comprises:
determining, when the target new energy station is identified as a newly built station or a newly expanded station based on the data amount of the historical output power, a data missing level of the future numerical weather prediction data based on the missing of the future numerical weather prediction data;
selecting, based on basic information of the target new energy station, a similar station whose similarity to the target new energy station reaches a similarity threshold from other new energy stations, the other new energy stations being stations in a set of stations except the target new energy station; and
selecting, based on the data missing level, the target power prediction model corresponding to the target new energy station from the optional power prediction models pertaining to the similar station in the set of power prediction models.
5. The method according to claim 3 , wherein selecting, based on the basic information of the target new energy station, the similar station whose similarity to the target new energy station reaches the similarity threshold from other new energy stations, comprises:
obtaining the basic information of the target new energy station and the basic information of each of the other new energy stations, the basic information comprising at least one dimension data selected from a group consisting of longitude and latitude data, capacity data, terrain data, landform data and altitude data;
determining, based on the basic information of the target new energy station and the basic information of each of the other new energy stations, a value of similarity between the target new energy station and each of the other new energy stations in each dimension data; and
selecting, based on the value of similarity between the target new energy station and each of the other new energy stations in each dimension data, the similar station whose similarity to the target new energy station reaches the similarity threshold from the other new energy stations.
6. The method according to claim 1 , wherein predicting, by the adjusted target power prediction model, the target output power of the target new energy station in the future period, based on the future numerical weather prediction data of the future period and the historical output power of the historical period corresponding to the future period comprises:
obtaining, if a station type of the target new energy station is a distributed new energy source, a grid position of the target new energy station; and
predicting, by the adjusted target power prediction model, the target output power of the target new energy station in the future period, based on the historical output power and the future numerical weather prediction data corresponding to the grid position.
7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method of claim 1 .
8. One or more non-transitory processor readable storage devices comprising a computer program stored thereon, the computer program, when executed by a processor, causing the processor to implement the steps of the method of claim 1 .
9. A method for automatic machine learning-based prediction of new energy power with cloud-edge collaboration, which is executed by a cloud server, the method comprising:
training, for each new energy station in a set of stations, each initial model based on a training sample set corresponding to the new energy station, and selecting an optional power prediction model corresponding to the new energy station from trained initial models, the training sample set comprising sample weather prediction data and sample output power, and different initial models being constructed based on different algorithms;
constructing, based on the optional power prediction model corresponding to each new energy station, a set of power prediction models, wherein the set of power prediction models is configured for an edge server to:
select a target power prediction model corresponding to a target new energy station, based on missing of future numerical weather prediction data of the target new energy station in a future period and a data amount of historical output power of the target new energy station in a historical period corresponding to the future period; and
adjust the target power prediction model in a target working mode, and predict, by the adjusted target power prediction model, a target output power of the target new energy station in the future period based on the future numerical weather prediction data of the future period and the historical output power of the historical period corresponding to the future period; and
delivering the set of power prediction models to the edge server.
10. The method according to claim 9 , wherein training each initial model based on the training sample set corresponding to the new energy station comprises:
performing data mining processing on original weather data in original data corresponding to the new energy station to obtain the sample weather prediction data in the training sample set;
performing multi-stage noise reduction processing on original output power in the original data to obtain the sample output power in the training sample set; and
training each initial model based on the training sample set.
11. The method according to claim 10 , wherein performing the data mining processing on the original weather data in the original data corresponding to the new energy station to obtain the sample weather prediction data in the training sample set comprises:
performing data cleaning on the original weather data in the original data based on a slicing technology;
performing correlation analysis on the original output power and the cleaned original weather data to determine a principal component feature set in the original weather data;
validating validity of the principal component feature set to obtain a valid feature set; and
using the valid feature set as the sample weather prediction data in the training sample set.
12. The method according to claim 9 , wherein training each initial model based on the training sample set corresponding to the new energy station comprises:
obtaining, for any initial model, a hyperparameter initial space and an initial step length corresponding to the initial model;
adjusting the initial step length based on the number of iterations;
searching for a hyperparameter optimization space within the hyperparameter initial space, based on the adjusted step length;
obtaining, based on a hyperparameter initial value and a Bayesian optimization algorithm, at least one optimized hyperparameter to be validated within the hyperparameter optimization space; and
performing, based on the training sample set corresponding to the new energy station and each optimized hyperparameter to be validated, k-fold cross validation on the initial model to complete the training of the initial model.
13. The method according to claim 9 , wherein selecting the optional power prediction model corresponding to the new energy station from trained initial models comprising:
inputting a validation sample set corresponding to the new energy station into each trained initial model, to obtain a validated output power output by each trained initial model;
determining, based on an actual output power corresponding to a validation sample in the validation sample set and the validated output power output by each trained initial model, a prediction accuracy of each trained initial model; and
selecting, based on the prediction accuracy, the optional power prediction model corresponding to the new energy station from trained initial models.
14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method of claim 9 .
15. One or more non-transitory processor readable storage devices comprising a computer program stored thereon, the computer program, when executed by a processor, causing the processor to implement the steps of the method of claim 9 .
16. A device for automatic machine learning-based prediction of new energy power with cloud-edge collaboration, the device comprising:
a response module configured to obtain, in respond to a power prediction demand for a target new energy station, future numerical weather prediction data of the target new energy station in a future period and historical output power of a historical period corresponding to the future period;
a selection module configured to select, based on missing of the future numerical weather prediction data and a data amount of the historical output power, a target power prediction model corresponding to the target new energy station from a set of power prediction models, each optional power prediction model in the set of power prediction models being trained and delivered by a cloud server based on an automatic machine learning algorithm;
an adjustment module configured to determine, based on the power prediction demand, a target working mode of the target power prediction model, the target working mode comprising a prediction business object, a prediction type, and a prediction time scale; and
a prediction module configured to:
adjust the target power prediction model according to the target working mode; and
predict, by the adjusted target power prediction model, a target output power of the target new energy station in the future period, based on the future numerical weather prediction data and the historical output power.
17. A device for automatic machine learning-based prediction of new energy power with cloud-edge collaboration, the device comprising:
a training module configured to:
train, for each new energy station in a set of stations, each initial model based on a training sample set corresponding to the new energy station, and
select an optional power prediction model corresponding to the new energy station from trained initial models, the training sample set comprising sample weather prediction data and sample output power, and different initial models being constructed based on different algorithms;
a constructing module configured to construct, based on the optional power prediction model corresponding to each new energy station, a set of power prediction models, wherein the set of power prediction models is configured for an edge server to:
select a target power prediction model corresponding to the target new energy station, based on missing of future numerical weather prediction data of the target new energy station in the future period and a data amount of the historical output power of the target new energy station in the historical period corresponding to the future period; and
adjust the target power prediction model in a target working mode, and predict, by the adjusted target power prediction model, a target output power of the target new energy station in the future period based on the future numerical weather prediction data of the future period and the historical output power of the historical period corresponding to the future period; and
a delivery module configured to deliver the set of power prediction models to the edge server.
18. A system for automatic machine learning-based prediction of new energy power with cloud-edge collaboration, the system comprising:
a cloud server configured to:
train, for each new energy station in a set of stations, each initial model based on a training sample set corresponding to the new energy station;
select an optional power prediction model corresponding to the new energy station from trained initial models, the training sample set comprising sample weather prediction data and sample output power, and different initial models being constructed based on different algorithms;
construct, based on the optional power prediction model corresponding to each new energy station, a set of power prediction models; and
deliver the set of power prediction models to the edge server; and
an edge server configured to:
receive the set of power prediction models;
obtain, in response to a power prediction demand for a target new energy station, future numerical weather prediction data of the target new energy station in a future period and historical output power of a historical period corresponding to the future period,
select, based on missing of the future numerical weather prediction data and a data amount of the historical output power, a target power prediction model corresponding to the target new energy station from a set of power prediction models, each optional power prediction model in the set of power prediction models being trained and delivered by the cloud server based on an automatic machine learning algorithm;
determine, based on the power prediction demand, a target working mode of the target power prediction model, the target working mode comprising a prediction business object, a prediction type, and a prediction time scale; and
adjust the target power prediction model according to the target working mode, and predict, by the adjusted target power prediction model, a target output power of the target new energy station in the future period, based on the future numerical weather prediction data and the historical output power.