Base station energy saving strategy determination method, device and method storage medium

Abstract

The present disclosure provides a method, device and storage medium for determining a base station energy-saving strategy, wherein the method includes: using a training sample set to train multiple machine learning models, using the trained machine learning model to predict and process the verification sample set to generate new training samples; training an energy-saving strategy recommendation model based on the new training samples and the corresponding scoring results, using the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for the target base station, and obtaining a fusion model based on the recommendation degree information and determining the energy-saving strategy. The method, device and storage medium disclosed in the present disclosure integrate the correlation between multiple machine learning models, recommend the optimal energy-saving strategy for different types of base stations, support automatic adjustment of base station parameters, reduce manual participation, effectively reduce the energy consumption of base station equipment without affecting user experience, and reduce the operating cost expenditure of telecommunications operation management.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a method, device and method storage medium for determining a base station energy-saving strategy. Background Art

[0002] The energy consumption of a single 5G base station has increased significantly. At the same time, the complexity of the network and the difficulty of operation and management have caused the operating cost expenditure to increase rapidly. The existing base station energy-saving technology mainly relies on simple models or manually set thresholds to decide whether to switch on or off, and the energy-saving effect is limited. Different base station deployment scenarios (such as indoor shopping malls, outdoor scenic spots, etc.) have different characteristics. In the existing base station energy-saving technology, there are currently a variety of AI-based energy-saving models based on these characteristics, such as energy-saving scene recognition models and business traffic prediction models. However, the existing base station energy-saving technology does not consider the integration of AI models, and in the process of formulating energy-saving strategies, each energy-saving strategy and energy-saving effect are independent of each other, and no correlation analysis is performed, which cannot achieve the best balance between user experience and energy-saving effect. Summary of the invention

[0003] In view of this, a technical problem to be solved by the present disclosure is to provide a method, device and method storage medium for determining a base station energy-saving strategy.

[0004] According to a first aspect of the present disclosure, a method for determining a base station energy-saving strategy is provided, including: obtaining historical data corresponding to multiple base stations, generating a training sample set and a verification sample set based on the historical data; using the training sample set to train multiple preset machine learning models; wherein multiple machine learning models are set for each machine learning model; using the trained machine learning models to perform prediction processing on the verification sample set, and generating new training samples based on the prediction results; obtaining scoring results corresponding to the new training samples, and training a preset energy-saving strategy recommendation model based on the new training samples and the corresponding scoring results; using the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for a target base station; obtaining a fusion model based on the recommendation degree information, and determining an energy-saving strategy for the target base station according to the fusion model; wherein the fusion model includes at least one machine learning model of each machine learning model.

[0005] Optionally, the use of the trained energy-saving strategy recommendation model to determine the recommendation information of each machine learning model for the target base station includes: obtaining prediction data corresponding to the target base station, and generating prediction samples based on the data; using each trained machine learning model to perform prediction processing on the prediction samples respectively, and generating new prediction samples based on the prediction results; using the trained energy-saving strategy recommendation model and determining the recommendation information of each machine learning model based on the new prediction samples respectively.

[0006] Optionally, acquiring the fusion model based on the recommendation degree information includes: taking at least one machine learning model with the highest recommendation degree in each machine learning model as a fusion member model; and establishing the fusion model based on the fusion member models.

[0007] Optionally, the multiple machine learning models include: an energy-saving scenario recognition model and a business traffic prediction model; the algorithms corresponding to the energy-saving scenario recognition model include but are not limited to: a clustering algorithm; the algorithms corresponding to the business traffic prediction model include but are not limited to: an LSTM algorithm and a LightGBM algorithm; the method also includes: establishing an AI energy-saving model knowledge base based on the energy-saving scenario recognition model and the business traffic prediction model.

[0008] Optionally, the fusion model includes: an energy-saving scenario recognition model and a business traffic prediction model; determining the energy-saving strategy for the target base station based on the fusion model includes: obtaining the prediction results of the energy-saving scenario recognition model and the business traffic prediction model in the fusion model; and determining the energy-saving strategy for the target base station based on the prediction results.

[0009] Optionally, energy-saving effect evaluation index information fed back by the target base station for the energy-saving strategy is obtained; and the energy-saving strategy recommendation model and the machine learning model are optimized according to the energy-saving effect evaluation index information.

[0010] Optionally, the obtaining of historical data corresponding to multiple base stations includes: preprocessing the obtained historical data; wherein the historical data includes: one or more of base station load performance KPI data, service quality data, base station working parameter data, and MR data.

[0011] Optionally, the algorithm corresponding to the energy-saving strategy recommendation model includes: a combination algorithm of a GBDT algorithm and a FM algorithm.

[0012] According to a second aspect of the present disclosure, a base station energy-saving strategy determination device is provided, comprising: a first sample processing module, used to obtain historical data corresponding to multiple base stations, and generate a training sample set and a verification sample set based on the historical data; a first model training module, used to use the training sample set to train multiple preset machine learning models; wherein multiple machine learning models are set for each machine learning model; a second sample processing module, used to use the trained machine learning models to perform prediction processing on the verification sample set, and generate new training samples based on the prediction results; a second model training module, used to obtain scoring results corresponding to the new training samples, and train a preset energy-saving strategy recommendation model based on the new training samples and the corresponding scoring results; a recommendation degree determination module, used to use the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for a target base station; a fusion model acquisition module, used to obtain a fusion model based on the recommendation degree information, wherein the fusion model includes at least one machine learning model in each machine learning model; an energy-saving strategy generation module, used to determine the energy-saving strategy for the target base station according to the fusion model.

[0013] Optionally, the recommendation degree determination module is specifically used to obtain prediction data corresponding to the target base station and generate prediction samples based on this data; use the trained machine learning models to perform prediction processing on the prediction samples respectively, and generate new prediction samples based on the prediction results; use the trained energy-saving strategy recommendation model and determine the recommendation degree information of each machine learning model based on the new prediction samples.

[0014] Optionally, the fusion model acquisition module is used to take at least one machine learning model with the highest recommendation degree in each machine learning model as a fusion member model; and establish the fusion model based on the fusion member models.

[0015] Optionally, the multiple machine learning models include: an energy-saving scenario recognition model and a business traffic prediction model; the algorithms corresponding to the energy-saving scenario recognition model include but are not limited to: a clustering algorithm; the algorithms corresponding to the business traffic prediction model include but are not limited to: an LSTM algorithm and a LightGBM algorithm; the device also includes: a model knowledge base establishment module, which is used to establish an AI energy-saving model knowledge base based on the energy-saving scenario recognition model and the business traffic prediction model.

[0016] Optionally, the fusion model includes: an energy-saving scenario recognition model and a business traffic prediction model; the energy-saving strategy generation module is used to obtain the prediction results of the energy-saving scenario recognition model and the business traffic prediction model in the fusion model; and determine the energy-saving strategy for the target base station based on the prediction results.

[0017] Optionally, the energy-saving strategy feedback module is used to obtain the energy-saving effect evaluation index information fed back by the target base station for the energy-saving strategy; and optimize the energy-saving strategy recommendation model and the machine learning model according to the energy-saving effect evaluation index information.

[0018] Optionally, the first sample processing module is used to preprocess the acquired historical data; wherein the historical data includes: one or more of base station load performance KPI data, service quality data, base station working parameter data, and MR data.

[0019] Optionally, the algorithm corresponding to the energy-saving strategy recommendation model includes: a combination algorithm of a GBDT algorithm and a FM algorithm.

[0020] According to a third aspect of the present disclosure, a base station energy saving strategy determination device is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the method as described above based on instructions stored in the memory.

[0021] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the instructions are executed by a processor as described above.

[0022] The base station energy-saving strategy determination method, device and method storage medium disclosed in the present invention use training sample sets to train multiple machine learning models, use the trained machine learning models to predict and process the verification sample sets, and generate new training samples; based on the new training samples and the corresponding scoring results, the energy-saving strategy recommendation model is trained, and the recommended degree information of each machine learning model is determined for the target base station using the trained energy-saving strategy recommendation model, and the fusion model is obtained based on the recommended degree information to determine the energy-saving strategy; the correlation between multiple machine learning models is integrated to recommend the optimal energy-saving strategy for different types of base stations, support automatic adjustment of base station parameters, reduce manual participation, and effectively reduce the energy consumption of base station equipment without affecting user experience, thereby reducing the operating cost expenditure of telecommunications operation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0024] Figure 1 A flowchart of an embodiment of a method for determining a base station energy saving strategy according to the present disclosure;

[0025] Figure 2 A schematic diagram of a process for determining recommendation degree information of a machine learning model in one embodiment of a method for determining a base station energy saving strategy according to the present disclosure;

[0026] Figure 3 A schematic diagram of an application technology architecture of an embodiment of a method for determining a base station energy saving strategy according to the present disclosure;

[0027] Figure 4 This is the schematic diagram of the GBDT+FM algorithm;

[0028] Figure 5 A schematic diagram of a principle of energy-saving strategy recommendation according to an embodiment of a method for determining an energy-saving strategy for a base station disclosed in the present invention;

[0029] Figure 6 A schematic diagram of implementing energy-saving strategy recommendation according to an embodiment of the method for determining an energy-saving strategy for a base station disclosed in the present invention;

[0030] Figure 7 A schematic diagram of modules of an embodiment of a device for determining a base station energy saving strategy according to the present disclosure;

[0031] Figure 8 A schematic diagram of modules of another embodiment of a base station energy saving strategy determination device according to the present disclosure;

[0032] Fig. 9 This is a module diagram of another embodiment of the base station energy saving strategy determination device according to the present disclosure. DETAILED DESCRIPTION

[0033] The present disclosure is described more fully below with reference to the accompanying drawings, in which exemplary embodiments of the present disclosure are described. The technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure. The technical solutions of the present disclosure are described in many aspects below in conjunction with various figures and embodiments.

[0034] Figure 1 FIG. 1 is a flow chart of an embodiment of a method for determining a base station energy saving strategy according to the present disclosure, as shown in FIG. Figure 1 As shown:

[0035] Step 101: Acquire historical data corresponding to a plurality of base stations, and generate a training sample set and a verification sample set based on the historical data.

[0036] In one embodiment, the base station may be a 4G or 5G base station in different deployment scenarios. Historical data include: one or more of base station load performance KPI (Key Performance Indication) data, service quality data, base station engineering parameter data, and MR data. Service quality data include: bandwidth, packet loss rate, etc. Base station engineering parameter data include parameters such as the location, height, specifications, and frequency of the base station. MR (Measurement Report) refers to a measurement report sent by a user terminal to the network in a call state. MR data includes the user's downlink RSRP value (Reference Signal Received Power) and uplink SINR value (Signal to interference plus noise ratio), etc.

[0037] Historical data can be obtained from a network management system, etc., and preprocessed, including existing data cleaning, feature construction, and data set construction. A variety of existing methods can be used to generate training sample sets and verification sample sets. For example, historical data corresponding to multiple base stations are obtained, the historical data are organized into data with a preset format, and the data are annotated accordingly, a sample set is generated based on the data and the annotations, and the sample set is divided into a training sample set and a verification sample set.

[0038] Step 102: Use the training sample set to train multiple preset machine learning models; wherein, multiple machine learning models are set for each machine learning model.

[0039] In one embodiment, the multiple machine learning models include energy-saving scene recognition models and business flow prediction models, etc. There are multiple machine learning models for the energy-saving scene recognition model, and there are multiple machine learning models for the business flow prediction model. The existing multiple model training method can be used to train all energy-saving scene recognition models and all business flow prediction models using the training sample set.

[0040] Step 103: Use the trained machine learning models to perform prediction processing on the verification sample set, and generate new training samples based on the prediction results.

[0041] In one embodiment, all trained energy-saving scenario recognition models and all service traffic prediction models are used to perform prediction processing on the verification sample set to obtain prediction results, which include the average downlink PRB utilization and the average number of RRC connected users; new training samples are generated based on the prediction results and samples in the verification sample set.

[0042] Step 104, obtain the scoring results corresponding to the new training samples, and train the preset energy-saving strategy recommendation model based on the new training samples and the corresponding scoring results. There can be multiple energy-saving strategy recommendation models, for example, the energy-saving strategy recommendation model corresponds to the algorithm including the combination algorithm of the GBDT algorithm and the FM algorithm.

[0043] In one embodiment, the prediction results generated by the energy-saving scenario recognition model and the business traffic prediction model for the samples in the verification sample set are scored by the user or in other ways, and the scoring can adopt multiple scoring standards. The scoring results are used as the annotation information for the new training samples, and the energy-saving strategy recommendation model can be trained using the new training samples and the scores (annotation information) using the existing multiple model training methods.

[0044] Step 105: Use the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for the target base station.

[0045] Step 106: Obtain a fusion model based on the recommendation degree information, and determine an energy-saving strategy for the target base station according to the fusion model; wherein the fusion model includes at least one machine learning model of each machine learning model.

[0046] Figure 2 FIG. 1 is a flow chart of determining the recommendation degree information of a machine learning model in one embodiment of the method for determining a base station energy saving strategy according to the present disclosure, such as Figure 2 As shown:

[0047] Step 201: Obtain prediction data corresponding to a target base station, and generate prediction samples based on the data.

[0048] In one embodiment, the target base station is a 4G or 5G base station in different deployment scenarios. The prediction data is the same as the historical data, including one or more of base station load performance KPI data, service quality data, base station engineering parameter data, and MR data. The prediction sample is generated based on the preset data; the format and content of the prediction sample are the same as the format and content of the samples in the training sample set and the verification sample set.

[0049] Step 202: Use the trained machine learning models to perform prediction processing on the prediction samples respectively, and generate new prediction samples based on the prediction results.

[0050] In one embodiment, all trained energy-saving scenario recognition models and all service flow prediction models are used to perform prediction processing on the prediction samples to obtain prediction results, which include downlink PRB average utilization and average number of RRC connected users, etc., and new prediction samples are generated based on the prediction results and the prediction samples. New prediction samples corresponding to each energy-saving scenario recognition model and each service flow prediction model are generated respectively, and the number of new prediction samples is the same as the number of all energy-saving scenario recognition models and all service flow prediction models.

[0051] Step 202: Use the trained energy-saving strategy recommendation model and determine the recommendation degree information of each machine learning model based on the new prediction sample.

[0052] In one embodiment, multiple prediction samples are sequentially input into a trained energy-saving strategy recommendation model to obtain the recommendation degree for each energy-saving scenario recognition model and each service flow prediction model output by the energy-saving strategy recommendation model. The recommendation degree is used to characterize the score of each energy-saving scenario recognition model and each service flow prediction model for the target base station.

[0053] There are many methods for obtaining a fusion model based on recommendation information. Take at least one machine learning model with the highest recommendation among each machine learning model as a fusion member model, and establish a fusion model based on the fusion member models. For example, obtain the recommendation degree of each energy-saving scene recognition model and each business traffic prediction model, take the machine learning model with the highest recommendation among multiple energy-saving scene recognition models as a fusion member model, take the machine learning model with the highest recommendation among multiple energy-saving scene recognition models as a fusion member model, and establish a fusion model based on the two fusion member models.

[0054] In one embodiment, Figure 3 As shown, the application technology architecture of the present disclosure includes a data source layer, an AI (Artificial Intelligence) energy-saving model layer, an energy-saving strategy recommendation layer, and a device layer. In the data source layer, the O domain, B domain, and M domain data collected are managed and maintained in the data lake. To complete the base station information management in the data lake, it is necessary to manage the collected base station basic data through data processing means, including data storage, update, feature construction, label and evaluation system construction, data set (such as training set, verification set, test set) construction, etc.

[0055] In the AI ​​energy-saving model layer, a variety of AI models are provided for the AI-based energy-saving strategy control module; combined with business logic, relevant parameters are extracted from the data lake and combined with the AI ​​model to facilitate the implementation of the AI-based energy-saving strategy control module. In the AI ​​energy-saving model layer, a variety of machine learning algorithms suitable for energy saving are included, such as clustering algorithms, neural network algorithms, and LightGBM algorithms. Based on the characteristics of base station data combined with scenario factor analysis, different AI energy-saving scenario recognition models are trained using different algorithms. Based on the characteristics of base station business traffic, different AI business traffic prediction models are trained using different algorithms. Through integration into a unified organization, an AI energy-saving model knowledge base is formed.

[0056] In the energy-saving strategy recommendation layer, the combined algorithm of GBDT algorithm and FM algorithm is used to train the energy-saving strategy recommendation model to obtain the fusion model, and the prediction results of each machine learning model in the fusion model are further combined and weighted to generate the final recommendation results based on different deployment scenarios. In the device layer, the communication unit is recommended for optimal energy-saving strategy through the open interface provided by OMC and suppliers.

[0057] In one embodiment, the energy-saving strategy recommendation model is an integrated model, the classifier uses CART, and the integration method is Gradient Boosting. Factorization machine (FM) is a machine learning algorithm based on matrix decomposition, which is easy to achieve feature fusion and can process hundreds of millions of data. However, FM has very limited learning ability and requires a lot of feature engineering to increase the learning ability of the model. GBDT can solve this problem.

[0058] The training sample set is used to train multiple preset machine learning models, and the trained machine learning models are used to predict the verification sample set. The prediction results of multiple machine learning models on the verification sample set include the average downlink PRB utilization and the average number of RRC connected users; based on the verification sample and the corresponding prediction results, as a new training set, the new training sample is recorded as: {(a (1) ,b (1) ),(a (2) ,b (2) ),…,(a (m) ,b (m) ), where a (i) is an n-dimensional data feature vector, which represents the feature vector of the i-th sample in the verification sample set. The features in the feature vector include: base station load performance KPI data, service quality data, base station engineering parameter data, etc. (i) is the output result predicted for the i-th sample based on the trained machine learning model.

[0059] Obtain the scoring results corresponding to the new training samples, that is, the scores of the users on the new training samples and the output results obtained by predicting the new training samples through the machine learning model, to indicate whether the prediction results are consistent with the actual situation. Train the preset energy-saving strategy recommendation model based on the new training samples and the corresponding scoring results.

[0060] The energy-saving strategy recommendation model can be an existing GBDT+FM model, which is composed of an existing GBDT (GradientBoosting Decision Tree) feature extraction layer and an existing FM (FactorizationMachine) feature processing layer. The GBDT feature extraction layer is used to extract features, and the FM (FactorizationMachine) feature processing layer is used to predict the recommendation degree based on the extracted features. The recommendation degree is used to represent the user's score on the output result predicted by the trained machine learning model.

[0061] like Figure 4 As shown in the figure, the features extracted by the GBDT+FM model through the GBDT feature extraction layer have the characteristics of high dimension, sparseness, and nonlinearity. The training process of the GBDT+FM model is as follows: Since the CART tree only contains the root node, all new training samples will be divided on this root node. The variance of all samples contained in the root node needs to be calculated, and the formula is:

[0062]

[0063] Among them, δ represents the variance of the root node, m represents the number of samples in the root node, and b (i) Represents the i-th sample label in the root node.

[0064] The next step is to split the CART tree containing only the root node into CART trees containing left and right subtrees until the final division is obtained. The formula is:

[0065]

[0066] Among them, δ1 and δ2 are the sample variances of the left and right subtrees, respectively, and m1 and m2 are the sample numbers of the left and right subtrees, respectively. The left and right subtrees are divided until the sum of the label values ​​of the nodes of the left and right subtrees is When it reaches the maximum, it means that the division result is optimal.

[0067] The GDBT model can be regarded as an additive model of the CART tree, and the formula is:

[0068]

[0069] Solve the optimal model by minimizing the loss function:

[0070]

[0071] Among them, a represents the new training sample, h is the parameter of CART, t represents CART, and w represents the weight of CART. Finally, the leaf nodes of the final GDBT model are used as the input of the FM model to train the final classifier.

[0072] The final GDBT+FM model algorithm can be expressed as the following formula 1-5, where F is the feature output of the GBDT model, U i are the model parameters:

[0073]

[0074] Next, use the new training set to train the GBDT+FM model, after the sigmoid function transformation:

[0075]

[0076] The recommendation degree of the energy-saving strategy is obtained through equations 1 to 6. If the two models with the highest recommendation degree are selected for fusion, the optimal recommendation of the energy-saving strategy is achieved, and the recommended energy-saving strategy model is used to calculate the threshold value for selecting energy-saving technology; wherein, the recommendation degree of each type of model is calculated separately, and the sum of the recommendation degrees of each type of model is 1.

[0077] Based on the recommended effect of the energy-saving strategy, the energy-saving effect evaluation indicators (including user experience data, energy-saving efficiency data, base station load data, etc.) are fed back to the energy-saving strategy recommendation model to perform iterative parameter optimization and form a closed-loop feedback mechanism.

[0078] In one embodiment, Figure 5 As shown, multiple machine learning models include energy-saving scenario recognition models and business traffic prediction models; the algorithms corresponding to the energy-saving scenario recognition models include existing clustering algorithms; the algorithms corresponding to the business traffic prediction models include existing LSTM algorithms and LightGBM algorithms; an AI energy-saving model knowledge base is established based on the energy-saving scenario recognition model and the business traffic prediction model.

[0079] The fusion model includes an energy-saving scenario recognition model and a business traffic prediction model. The prediction results of the energy-saving scenario recognition model and the business traffic prediction model in the fusion model are obtained, and an energy-saving strategy is determined for the target base station based on the prediction results.

[0080] The energy-saving effect evaluation index information of the target base station for the energy-saving strategy feedback can be obtained, and the energy-saving strategy recommendation model and the machine learning model can be optimized according to the energy-saving effect evaluation index information using existing methods. The energy-saving effect evaluation index information includes user experience data, energy-saving efficiency data, and base station load data.

[0081] The base station energy-saving strategy determination method disclosed in the present invention uses the GBDT+FM algorithm to perform correlation analysis on different energy-saving strategy models based on base stations in different deployment scenarios, integrates the business traffic prediction results of base stations in different deployment scenarios based on multiple machine learning models, recommends the optimal energy-saving strategy, and finally applies it to the 5G wireless communication unit system.

[0082] The disclosed method for determining a base station energy-saving strategy considers the closed-loop feedback between energy-saving effects and energy-saving strategy recommendations, achieves a balance between energy-saving and user experience for each base station, reduces overall base station energy consumption without affecting or improving user experience, and ultimately reduces OPEX. The disclosed method for determining a base station energy-saving strategy does not require improvements to terminal equipment, but only requires the use of machine learning methods to formulate an intelligent energy-saving strategy recommendation method on the 5G system, and performs intelligent energy-saving management on the communication unit through the open interface provided by the OMC and the supplier.

[0083] In one embodiment, Figure 5 As shown in the figure, data is collected from the OMC interface, and historical data corresponding to multiple base stations are obtained by parsing data packets, including base station load performance KPI data, service quality data, basic engineering parameter data, and MR data. The collected historical data is preprocessed, including data cleaning, such as data de-duplication, removal of unnecessary fields and features with high similarity, etc. The training set and validation set are split according to the business logic in a ratio of 7:3.

[0084] Integrate existing machine learning models (AI models) to form an AI model knowledge base. Machine learning models include energy-saving scene recognition models based on multiple clustering algorithms (Kmeans, SVM), and base station traffic prediction models based on algorithms such as LSTM and LightGBM.

[0085] Obtain the prediction data corresponding to the target base station, and generate prediction samples based on this data; the format and specific data of the prediction samples are the same as those of the training set and the validation set. Use the trained energy-saving scenario recognition models and business traffic prediction models to perform prediction processing on the prediction samples respectively, and generate new prediction samples based on the prediction results.

[0086] For example, the trained energy-saving scenario recognition models and service flow prediction models are used to obtain the prediction results of PRB and RRC for the prediction samples, respectively, as new prediction samples, recorded as: (1) ,b (1) ),(a (2) ,b (2) ),…,(a (m) ,b (m) ).

[0087] In one embodiment, formula (1-6) is applied to target base stations located in shopping malls and schools to calculate the recommendation degrees of two energy-saving scene recognition models and two business traffic prediction models, and the results of the fusion model are shown in Table 1 below.

[0088]

[0089]

[0090] Table 1-Recommendation calculation and model recommendation results

[0091] As shown in Table 1, the energy-saving strategy recommendation model is applied to make the optimal energy-saving strategy recommendation. For base stations with shopping mall scenes, the fusion model of energy-saving scene recognition model 2 and business traffic prediction model 1 is recommended. For base stations with scenic scenes, the fusion model of energy-saving scene recognition model 1 and business traffic prediction model 1 is recommended. The prediction results of the energy-saving scene recognition model and the business traffic prediction model in the fusion model are used to set and determine the energy-saving strategy. The energy-saving strategy is based on multiple threshold values ​​and the prediction results of the energy-saving scene recognition model and the business traffic prediction model to determine the symbol, channel, etc. shutdown strategy for the target base station, such as symbol shutdown, channel shutdown, etc.

[0092] The energy-saving strategy recommendation effect is used to recommend energy-saving strategies for base stations, and user experience data, energy saving data and base station load data are fed back to the energy-saving strategy recommendation model to perform iterative parameter optimization and form a closed-loop feedback mechanism.

[0093] In one embodiment, Figure 7 As shown, the present disclosure provides a base station energy-saving strategy determination device 70, including: a first sample processing module 71, a first model training model 72, a second sample processing module 73, a second model training module 74, a recommendation degree determination module 75, a fusion model acquisition module 76 and an energy-saving strategy generation module 77.

[0094] The first sample processing module 71 obtains historical data corresponding to multiple base stations, and generates a training sample set and a verification sample set based on the historical data. The first model training module 72 uses the training sample set to train multiple preset machine learning models; wherein, multiple machine learning models are set for each machine learning model. The second sample processing module 73 uses the trained machine learning models to perform prediction processing on the verification sample set, and generates new training samples based on the prediction results. The second model training module 74 obtains the scoring results corresponding to the new training samples, and trains the preset energy-saving strategy recommendation model based on the new training samples and the corresponding scoring results.

[0095] The recommendation degree determination module 75 uses the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for the target base station. The fusion model acquisition module 76 acquires the fusion model based on the recommendation degree information, wherein the fusion model includes at least one machine learning model in each machine learning model. The energy-saving strategy generation module 77 determines the energy-saving strategy for the target base station according to the fusion model.

[0096] In one embodiment, the recommendation degree determination module 75 obtains prediction data corresponding to the target base station and generates prediction samples based on the data. The recommendation degree determination module 75 uses the trained machine learning models to perform prediction processing on the prediction samples respectively and generates new prediction samples based on the prediction results. The recommendation degree determination module 75 uses the trained energy-saving strategy recommendation model and determines the recommendation degree information of each machine learning model based on the new prediction samples.

[0097] The fusion model acquisition module 76 uses at least one machine learning model with the highest recommendation degree in each machine learning model as a fusion member model, and establishes a fusion model based on the fusion member model. The first sample processing module 71 preprocesses the acquired historical data; wherein the historical data includes: one or more of base station load performance KPI data, service quality data, base station working parameter data, and MR data. The algorithm corresponding to the energy-saving strategy recommendation model includes a combination algorithm of the GBDT algorithm and the FM algorithm, etc.

[0098] In one embodiment, Figure 8 As shown, the base station energy-saving strategy determination device 70 also includes a model knowledge base establishment module 78 and an energy-saving strategy feedback module 79. The various machine learning models include energy-saving scene recognition models and business flow prediction models, etc.; the algorithms corresponding to the energy-saving scene recognition models include clustering algorithms, etc., and the algorithms corresponding to the business flow prediction models include LSTM algorithms, LightGBM algorithms, etc. The model knowledge base establishment module 78 establishes an AI energy-saving model knowledge base based on the energy-saving scene recognition model and the business flow prediction model.

[0099] The fusion model includes an energy-saving scenario recognition model and a service flow prediction model. The energy-saving strategy generation module 77 obtains the prediction results of the energy-saving scenario recognition model and the service flow prediction model in the fusion model, and determines the energy-saving strategy for the target base station based on the prediction results. The energy-saving strategy feedback module 79 obtains the energy-saving effect evaluation index information of the target base station for the energy-saving strategy feedback, and optimizes the energy-saving strategy recommendation model and the machine learning model according to the energy-saving effect evaluation index information.

[0100] In one embodiment, Fig. 9 FIG. 1 is a module diagram of another embodiment of a base station energy saving strategy determination device according to the present disclosure. Fig. 9As shown, the device may include a memory 91, a processor 92, a communication interface 93 and a bus 94. The memory 91 is used to store instructions, the processor 92 is coupled to the memory 91, and the processor 92 is configured to execute the above-mentioned base station energy saving strategy determination method based on the instructions stored in the memory 91.

[0101] The memory 91 may be a high-speed RAM memory, a non-volatile memory, etc., or a memory array. The memory 91 may also be divided into blocks, and the blocks may be combined into virtual volumes according to certain rules. The processor 92 may be a central processing unit CPU, or an application specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the base station energy saving strategy determination method disclosed in the present invention.

[0102] In one embodiment, the present disclosure provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implements the method for determining a base station energy-saving strategy in any of the above embodiments.

[0103] The base station energy-saving strategy determination method, device and method storage medium in the above-mentioned embodiments use training sample sets to train multiple machine learning models, use the trained machine learning models to predict and process the verification sample sets, and generate new training samples; based on the new training samples and the corresponding scoring results, the energy-saving strategy recommendation model is trained, and the recommended degree information of each machine learning model is determined for the target base station using the trained energy-saving strategy recommendation model, and the fusion model is obtained based on the recommended degree information and the energy-saving strategy is determined; the association between multiple machine learning models is integrated to recommend the optimal energy-saving strategy for different types of base stations, and the automatic adjustment of base station parameters is supported, which reduces manual participation, effectively reduces the energy consumption of base station equipment without affecting the user experience, reduces the operating cost expenditure of telecommunications operation management, and achieves greater economic benefits; and, data information is obtained by gateway equipment, etc., and the energy-saving strategy only needs to be sent to the device through the OMC interface or the open interface provided by the supplier, and there is no need to directly control the device, so that the solution is promoted without difficulty.

[0104] The method and system of the present disclosure may be implemented in many ways. For example, the method and system of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0105] The description of the present disclosure is given for the purpose of illustration and description, and is not intended to be exhaustive or to limit the present disclosure to the disclosed form. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present disclosure, and to enable those of ordinary skill in the art to understand the present disclosure and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. A method for determining a base station energy saving strategy, comprising: Acquire historical data corresponding to a plurality of base stations, and generate a training sample set and a verification sample set based on the historical data; Using the training sample set to train a plurality of preset machine learning models; wherein a plurality of machine learning models are provided for each machine learning model; Use the trained machine learning models to perform prediction processing on the verification sample set, and generate new training samples based on the prediction results; Obtaining a scoring result corresponding to the new training sample, and training a preset energy-saving strategy recommendation model based on the new training sample and the corresponding scoring result; Use the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for the target base station; A fusion model is obtained based on the recommendation degree information, and an energy-saving strategy is determined for the target base station according to the fusion model; wherein the fusion model includes at least one machine learning model of each machine learning model.

2. The method according to claim 1, wherein the step of using the trained energy-saving strategy recommendation model to determine the recommendation degree information of each machine learning model for the target base station comprises: Obtain prediction data corresponding to the target base station, and generate prediction samples based on the data; Use the trained machine learning models to perform prediction processing on the prediction samples respectively, and generate new prediction samples based on the prediction results; The trained energy-saving strategy recommendation model is used and the recommendation degree information of each machine learning model is determined based on the new prediction sample.

3. The method according to claim 1, wherein obtaining a fusion model based on the recommendation degree information comprises: At least one machine learning model with the highest recommendation degree among each machine learning model is used as a fusion member model; The fusion model is established based on the fusion member models.

4. The method of claim 1, wherein: The multiple machine learning models include: an energy-saving scene recognition model and a business flow prediction model; the algorithm corresponding to the energy-saving scene recognition model includes: a clustering algorithm; the algorithm corresponding to the business flow prediction model includes: an LSTM algorithm and a LightGBM algorithm; the method also includes: An AI energy-saving model knowledge base is established based on the energy-saving scenario recognition model and the business traffic prediction model.

5. The method of claim 4, wherein: The fusion model includes: an energy-saving scenario recognition model and a service flow prediction model; and determining an energy-saving strategy for a target base station according to the fusion model includes: Obtaining prediction results of the energy-saving scenario recognition model and the service flow prediction model in the fusion model; An energy saving strategy is determined for the target base station based on the prediction result.

6. The method of claim 1, further comprising: Obtaining energy-saving effect evaluation index information fed back by the target base station for the energy-saving strategy; The energy-saving strategy recommendation model and the machine learning model are optimized according to the energy-saving effect evaluation index information.

7. The method of claim 1, further comprising: Preprocessing the acquired historical data; The historical data includes one or more of base station load performance KPI data, service quality data, base station engineering parameter data, and MR data.

8. The method according to any one of claims 1 to 7, wherein: The algorithms corresponding to the energy-saving strategy recommendation model include: a combination algorithm of the GBDT algorithm and the FM algorithm.

9. A base station energy saving strategy determination device, comprising: A first sample processing module, configured to obtain historical data corresponding to a plurality of base stations, and generate a training sample set and a verification sample set based on the historical data; A first model training model is used to train a plurality of preset machine learning models using the training sample set; wherein a plurality of machine learning models are provided for each machine learning model; A second sample processing module is used to use the trained machine learning models to perform prediction processing on the verification sample set and generate new training samples based on the prediction results; A second model training module is used to obtain a scoring result corresponding to the new training sample, and train a preset energy-saving strategy recommendation model based on the new training sample and the corresponding scoring result; A recommendation degree determination module, used to determine the recommendation degree information of each machine learning model for the target base station using the trained energy-saving strategy recommendation model; A fusion model acquisition module is used to acquire a fusion model based on the recommendation degree information, wherein the fusion model includes at least one machine learning model in each machine learning model The energy-saving strategy generation module is used to determine the energy-saving strategy for the target base station according to the fusion model.

10. The device of claim 9, wherein: The recommendation degree determination module is specifically used to obtain prediction data corresponding to the target base station, and generate prediction samples based on the data; use the trained machine learning models to perform prediction processing on the prediction samples respectively, and generate new prediction samples based on the prediction results; The trained energy-saving strategy recommendation model is used and the recommendation degree information of each machine learning model is determined based on the new prediction sample.

11. The device of claim 9, wherein: The fusion model acquisition module is used to take at least one machine learning model with the highest recommendation degree in each machine learning model as a fusion member model; and establish the fusion model based on the fusion member model.

12. The device of claim 9, wherein: The multiple machine learning models include: an energy-saving scene recognition model and a business flow prediction model; the algorithm corresponding to the energy-saving scene recognition model includes: a clustering algorithm; the algorithm corresponding to the business flow prediction model includes: an LSTM algorithm and a LightGBM algorithm; the device also includes: A model knowledge base establishment module is used to establish an AI energy-saving model knowledge base based on the energy-saving scenario recognition model and the business traffic prediction model.

13. The device of claim 12, wherein: The fusion model includes: an energy-saving scenario recognition model and a service flow prediction model; The energy-saving strategy generation module is used to obtain the prediction results of the energy-saving scenario recognition model and the service flow prediction model in the fusion model; and determine the energy-saving strategy for the target base station based on the prediction results.

14. The apparatus of claim 9, further comprising: An energy-saving strategy feedback module is used to obtain the energy-saving effect evaluation index information fed back by the target base station for the energy-saving strategy; and optimize the energy-saving strategy recommendation model and the machine learning model according to the energy-saving effect evaluation index information.

15. The device of claim 9, wherein: The first sample processing module is used to pre-process the acquired historical data; wherein the historical data includes: one or more of base station load performance KPI data, service quality data, base station working parameter data, and MR data.

16. The device according to any one of claims 9 to 15, wherein: The algorithms corresponding to the energy-saving strategy recommendation model include: a combination algorithm of the GBDT algorithm and the FM algorithm.

17. A base station energy saving strategy determination device, comprising: Memory; and a processor coupled to the memory, wherein the processor is configured to execute the method according to any one of claims 1 to 8 based on instructions stored in the memory.

18. A computer-readable storage medium storing computer instructions, wherein the instructions are executed by a processor to execute the method according to any one of claims 1 to 8.

Images