Power distribution mechanism for off-grid base station operation

Abstract

With the explosive expansion of small cells to sustain mobile traffic growth for ubiquitous access, operation expenditure (OpEx) such as electric charge caused by power consumption over wireless cellular networks has severely increased. In particular, the power consumption of base stations (BSs) has been recognized as a significant factor in inducing environment problems such as greenhouse gas (GHG) emissions. Moreover, products of information and communication technology (ICT) account for 10% of global electricity generation. However, products powered by renewable energy sources (RESs) are free of CO2 emissions. Off-grid BSs (i.e., RES powered BSs) are recently emerging in wireless cellular networks. However, since RESs with intermittent generation are uncontrollable and unpredictable, reliable power distribution for BS operation is not guaranteed. In an effort to overcome the problem of unreliable power supply, we propose a power distribution mechanism for off-grid BS operation. Here, we consider traffic load dependent power consumption models and BSs that have different traffic profiles and their own batteries with different capacities. We assume the BS's traffic load follows a Poisson distribution.

First, we propose a multiple power distribution scheme (M-PDS) where BSs can be powered by other non-home retailers when the home retailer's generation is lacking. Due to the environment of multiple retailers, we differentiate the priority of unit power cost between home and non-home retailers. For M-PDS, we define a multiple power distribution utility function (M-PDU) that uses a method to extract the cellular provider's power cost from a satisfaction function of the power distribution. The M-PDS uses Largrange multipliers and dual decomposition, and divides the master problem into subproblems. Through repeated coordination of parameters between BSs and retailers, the optimal power distribution is then obtained heuristically, since we proved that the M-PDU is a concave function.

Second, to address the situation when insufficient power is supplied to BSs from energy harvesters, we design an energy outage (EO) probability and propose a new PDS to reduce EOs in off-grid BS operation. EOs can define that due to insufficient power supply, mobile users connecting to BSs cannot be guaranteed power demand or QoS. The EO probability can be calculated according to the required power demand through the BS's traffic and the input transmission power obtained by the total power supply. Here, we assume the interval of the BS's traffic arrival follows a uniform distribution. If EOs occur, RESs are difficult to use as the primary energy source. Accordingly, an EO-aware PDS (EO-PDS) is used to extract the EO probability from a satisfaction function for the BS's power demand and supply. The EO-PDS also uses Largrange multipliers and dual decomposition, and divides the master problem into subproblems. Then, through repeated coordination of parameters between BSs and energy harvesters, the optimal power distribution is obtained heuristically, since the EO-PDU is a concave function.

Using a MATLAB tool, we present a performance analysis of the M-PDS and EO-PDS. The existing PDS for the M-PDS in chapter 3 provides a power distribution to reduce $CO_2$ emissions as well as to increase the cellular provider's revenue in a hybrid power supply combining the main grid and RESs. Next, the conventional PDSs for the EO-PDS in chapter 4 are a low load (Low-) PDS that prioritizes the power distribution based on low traffic, and a fair load (Fair-) PDS that equally provides the power distribution to BSs. First, we analyze the performance for single and multiple cellular providers using the M-PDS. If a retailer's generation is insufficient, the M-PDS fixedly provides the power distribution. However, if the retailer's power generation is sufficient, the M-PDS provides the power distribution slightly proportional to the traffic load. If a cellular provider is sufficiently powered by the home retailer, the satisfaction is flat. However, if a cellular provider is insufficiently powered by the home retailer, the satisfaction is slightly proportional to the BS load. Furthermore, according to an increasing power distribution from a retailer with low unit power cost, the satisfaction of the power distribution is high. Next, using the quadratic energy cost model, the M-PDS for multiple cellular providers is compared with the conventional PDS. The M-PDS guarantees better satisfaction with around a minimum of 27.87% (macro BS) and 25.59% (remote radio head, RRH) at low load. Moreover, within power coverage the M-PDS reduces the total power cost for around 30.54% (macro BS) and 24.68% (RRH) at low load. Next, the EO-PDS guarantees satisfactory performance with around a minimum of 0.09% at Low- and Fair-PDSs, and with around a maximum of 17.32% at the Low-PDS. The optimal power distribution is inversely proportional to the traffic load and proportional to the BS's battery condition.