Base Station Battery SOC Estimation Research And Operation And Maintenance Dispatch System Development

Under the background of the national Internet Plus strategy and the upcoming 5G network,major operators have significantly increased the number of mobile communication base stations.The storage battery as a standby power source for the base station is a key part of the power supply system.If a failure occurs when the battery is in use,the communication service system will be blocked.Because the communication base station has complex site environment,large number and wide distribution location,unattended and other characteristics,and the base station battery is used as a backup power source,many problems arise.Therefore,an efficient and intelligent base station battery operation and scheduling system is needed.Base station battery remaining capacity(SOC)monitoring is the premise of the operation and maintenance of the base station's battery.Therefore,it is of great significance to accurately estimate the battery SOC.This paper uses RBF neural network as an estimation model of battery residual capacity(SOC).Firstly,L1/2 regularization is introduced to select the appropriate hidden layer nodes of the neural network;then,the generalization ability of the neural network is improved by the scaling algorithm based on the fuzzy theory;then the simulated annealing algorithm is integrated into the gradient descent method,thereby improving the training process.It is easy to produce a poor local optimal solution that requires a lot of repeated training.Finally,we will introduce the software and hardware for the development and construction of base station battery operation and maintenance scheduling systems.A cost-optimized maintenance site selection and operation and maintenance scheduling route algorithm is proposed under the condition that the base station carries out the emergency operation and maintenance.The main work and corresponding content of this article are as follows:1.In the training process using RBF neural network,the performance of neural network depends on the network structure,especially the number of hidden layer nodes.When the number of nodes is too small,the data feature cannot be learned through training and accurate results cannot be obtained.When the number of nodes is too large,overfitting occurs due to over-learning of data features,resulting in a large error in the estimation process.Because regularization has the ability of sparse model,this paper selects the number of hidden layer nodes of RBF neural network through L1/2 regularization algorithm to obtain the appropriate number of hidden layer nodes.For a trained neural network,its generalization ability is an important criterion to measure whether the performance of the neural network is excellent.It determines the ability of the neural network to accurately estimate when processing new data.Therefore,the neural network generalization ability is improved by a zooming algorithm based on fuzzy theory.The random selection of initial points of the gradient descent method results in the resulting solution easily falling into the local minimum,The idea of the simulated annealing algorithm is incorporated into the training algorithm to give the ability to jump in the search process,so as to get rid of the local optimal solution that does not meet the requirements and approach the global optimal solution.The above improvement was verified through experimental simulation.2.Build and develop base station battery operation and maintenance scheduling system.For the hardware part,the core control module designed in the system development and the corresponding physical map are described,For the software part,the system module unit function of the operation and maintenance scheduling platform based on Visual Studio is briefly described,and the data transmission function of the platform is tested and verified.Finally,based on battery SOC estimation,under the premise that the base station can be timely operated and maintained,a cost-optimized maintenance site selection,scheduling route design,and personnel dispatching algorithm scheme are proposed.