Optimized Energy Efficiency in a Telecommunication Company: Machine Learning Approach

Abstract

Energy efficiency is the use of technology that requires less energy to perform the same task. It was considered to introduce optimized energy efficiency by using machine learning to reduce power consumption at communication base station (BTS) sites. This process involves reviewing relevant work to identify defects, characterizing and determining the power consumption of the cell site under investigation, developing a SIMULINK model for the cell site under investigation, and identifying the module. It also includes optimizing high power consumption; design a machine learning rule base to monitor the power consumption of the module. Train artificial neural network (ANN) on machine learning rules designed to reduce cell power consumption, thereby improving network performance. The next step is developing an algorithm to implement it, and finally, to design a power consumption model for the network under investigation. The results obtained after a large simulation show that the traditional maximum power consumed at the cell site is 5746 kW, while the power when machine learning is injected into the system is 4733 kW. Integrating machine learning into the system resulted in 4731 kW, an 8.9% performance improvement.

1. Introduction

The trouble of power efficiency is one of the main challenges dealing with wireless cell community vendors around the world today. In Nigeria, some cell network vendors have been affected, resulting in the closing down of websites due to the trouble of energy efficiency. To handle this situation, which offers power consumption reduction in a base station (BS), several processes have been adopted that led to the introduction of green conversation techniques. Power amplifier (PA) improvement has attracted a good deal of attention because it devours the greatest proportion of the strength consumption of BSs. In cellular communications, the energy amplifier in a macro phone BS consumes the most energy, as a great deal as 65% of the complete power bumps off via all BS elements. The trouble is that the strength efficiency of PA is doable solely with interior equipment, but now not with external prerequisites such as not knowing the wide variety of users asking for getting entry to the BS at a time. This changes the electricity effectivity done through the internal equipment. It is normal that excessive bit error fees result in poor conversation overall performance [1]. On the different hand, Akbari et al. [2] certainly emphasizes the need to integrate ultracapacitors into the device for energy efficiency. The author of egalitarianism [3] emphasized the need to redefine wi-fi conversation to enhance its effectiveness. Bazzi et al. [4] reiterated that greater throughput is a core function of multiradio efficiency. Optimization issues seem to be for most or minimum values that a character can take. In the absolute extremes section, we noticed how to resolve sure optimization problems. Here we have located the maximum and minimum values that the function takes in the interval. Machine learning is a synthetic intelligence (AI) software that provides systems with the ability to automatically learn and improve. Machine getting to know teaches computer systems what people take for granted, that is, what they learn from experience. Machine getting to know algorithms use computational methods to "learn" statistics without delay from data, except relying on specific equations as a model. The algorithm adaptively improves overall performance as the wide variety of samples on hand for education increases. The development and modifications that have taken vicinity in the enterprise lately have entered a new phase in parallel with the improvement of computer technology, fuzzy logic, and, ultimately, a completely new subject of synthetic intelligence, the study [5] reveals that artificial talent (AI) is growing because of its manageable to be predictive and sourcing. Renewable strength sources such as wind and solar are very useful and clean sources; their indepleteable houses help in enhancing or smoothing performance in aggregate with different sources, such as biomass, in particular in rural areas. The methodology to achieve the intention of this work is the adherence to the mentioned research goals, which has to do with the tabulation of the gathered data and characterization of the current telecommunication primary based transceiver underneath learn about [6].

1.1 Aim of the work

This paper is about the optimization of energy efficiency through reduction of power consumption in a telecommunication base transceiver station (BTS).

1.2 The study objectives

The high consumption of power by modules of a cell site has impacted on the operations of the telecommunication company. This has necessitated the introduction of optimized energy efficiency through reduction of power consumption in a telecommunication base transceiver station (BTS) site using machine learning. The specific objectives are stated thus to

1. Determine the power consumption of the cell modules to be optimized from the collected data.

2. Perform optimization of the established high power consumed by the modules of the cell to a minimal value.

3. Design a machine learning rules for a reduced power consumption in the cell to enhance its performance.

4. Train the ANN in the designed machine learning rules for reduced power consumption in the cell site.

5. Develop an algorithm that will implement the sequence.

6. Develop a power consumption model for the network under study based on the result obtained when the algorithm is integrated in it.

Validate and justify the percentage improvement of energy efficiency in the cell site with and without the application of machine learning.

2. A closer look at previous related works

Introduction of renewable strength sources can enhance the diesel generators used in the base stations of all Nigerian carriers, in wind farms, the place the source of electricity is stochastic, the inefficiency of enhancing the percentage efficiency of the mills to raise the production capacity of the industries that fully rely on the generator for their each day manufacturing is addressed with the aid of strength efficiency upgrades of doubly fed induction generator machines, the use of adaptive manipulate approach [7]. In our verbal exchange network, the lengthen in sending data from the transmit point to the get hold of factor is a very massive problem, so the power provides desires to be reliable; the minimization in electricity supply in the United States has induced financial troubles to small-scale industries; one principal hassle attributed to this is inadequate planning mechanism that will forecast the required amount of electricity that will be needed to feed the complete population [8]. During the current period, the renewable energy source has dramatically extended both qualitative and quantitative enhancements with growing pressure to overcome environmental and financial crises; taking awareness of the reality that passing information or transfer of data from one factor to the other has ended up a chronic problem in our communication industry, energy efficiency techniques should be adopted such as laptop; learning this ought to properly improve the robustness of information network, the usage of adaptive modulation technique [9]. Recently, ways to improve energy efficiency have been delivered to improve demand-side management of strength distribution systems. Measures brought in the work of [10] place optimized genetic algorithm (OGA), which was used to improve epileptic electricity provided from the country-wide grid due to instability that has been a problem to strengthen consumers. This instability in strength furnished skilled in energy distribution network ought to be minimized with the aid of introducing optimized genetic algorithm (OGA). Recently, voltage adjustment primarily based on reinforcement gaining knowledge of and distribution evaluation is additionally gaining popularity; renewable energy penetration into the power mix in mild of the developing global demand for strength has elevated the distribution and power satisfaction considerably. Renewable strength systems (RESs) had been hastily developed due to ecological, social, economic, and environmental elements such as the extensively established photovoltaic structures (PVs) and wind turbine systems (WTs) [11]. In learning about [12], Nigeria electricity gadget is confronted with a sequence of technical challenges due to long, radial, weak, and aging transmission network; this paper introduces the idea of electricity conservation and related technologies, as properly as choices to help users achieve the benefits of electricity efficiency improvements. The work includes using computer learning to improve power first-class and decreasing strength consumption as in the case of the telecommunication base transceiver station (BTS).

Research has been carried out on strength effectivity enhancement methods for the motive of reduction in electricity consumption and environmental pollution from unburnt hydrocarbons. Today’s standard wireless get admission to networks consumes more than 50% of the total power consumption of mobile communications networks, which excludes the energy fed on via the cell stations (user terminals) whose more than 50% of electricity consumption is without delay attributed to the base station (BTS) equipment. However, a discount on the electricity consumption of cellular networks is of remarkable importance from within your budget (cost reduction), environmental (decreased CO₂ emissions), and efficiency perspectives. Hence, each reduction in strength consumption and CO₂ emission are key drivers for the future of the ICT industry. In the latest file by way of the International Telecommunications Union (ITU) and Alliance for Telecommunications Industry Solutions (ATIS), a quantity of energy environment-friendly practices and strategies for consideration by way of agencies seeking to gain larger efficiencies within their wi-fi networks have been outlined.

There are lively lookup works on energy consumption, reduction, and efficiency in wireless get entry to networks, but issues touching on to the implementation of desktop learning approach had not been utilized considerably and explicitly addressed. This paper investigates power consumption of base transceivers stations (BTS), schemes that may want to doubtlessly reduce the power consumption have been described, and the management of reusing the conserved energy barring compromising first-class of the carrier (QoS) of the community explored. The research additionally investigates the importance of deploying optimization methods on strength efficiency.

We know that base transceiver station (BTS) is a transceiver that acts as an interface between the mobile stations (MS) to the network. A BTS will have between 1 and 16 transceivers (TRX) depending on the geography and demand for the provider of an area. Each TRX represents one ARFCN (absolute radio-frequency channel number). However, relying on geography, carrier demand, and operator’s network method and architecture, a BTS can also host up to two, three, or six sectors, or a cell might also be serviced using various BTSs with redundant sector coverage. Each area is protected by a quarter antenna, which is a directional antenna. Figure 1 indicates the typical macro BTS we found today. A range of remarkable documentations of hooked-up research techniques and philosophy have been mentioned extensively. Unfortunately, little comparison and integration throughout studies exist. In this article, a frequent appreciation of computer mastering and sensible agents’ research and its implementation was undertaken. This paper does not extensively discuss electricity effectivity technologies but is looking to utilize modern-day mathematical strategies to successfully minimize strength consumption to retailer price in the industries. A dialogue on the framework protecting the literature on AI and ML research is restricted to energy effectivity techniques. Rather, it attempts to supply a beginning factor for integrating understanding throughout research in this area and suggests paths for future research. It explores research in certain novel disciplines: environmental pollution, medicine, maintenance, manufacturing, etc. Further lookup is wished to lengthen the current boundary of know-how in computer learning and optimization approaches. Utilizing machine gaining knowledge of disciplines into the current AI frameworks should through greater light maximize the beneficial properties of this strategy. This paper provides precious thoughts and perspectives for undergoing research on AI and ML. The closing aim was once to comprehend a reduction in electricity consumption using power efficiency means. The work offers a basis for future implementation of intelligent agents and machine learning strategies to achieve power savings that would finally translate to costs savings.

Figure 1.

Designed machine learning fuzzy inference system that will monitor the power consumed on the modules and minimize it if high.

3. Materials and methods

To get the desired results and achieve our purpose, there is a need to follow the stated objectives sequentially and observe the procedure.

Characterize and determine the power consumption of the modules of the cell site under study; to characterize the cell site and determine its power consumption, the type of base station (BS), configuration model, transceiver, and power models were inspected. The cell site or base station or base transceiver station (BTS) is a microcell managed by IHS Towers West Africa Limited, with site No. IHS-EN-T4670—2G/3G/4G networks (Indoor/Outdoor Site), housing MTN Nig Ltd and Airtel Nig Ltd base station equipment at Mount Street by Idaw River Layout Awkunanaw, Enugu. The site control (hop) about 30 other MTN/Airtel base stations (Terminal and Fiber sites) within its coverage area, it handles transmission (TX) and reception (RX) of voice, data, and streaming services.

A period of 27 days was used to monitor and carry out the measurements. The days include the morning period (peak), afternoon (off-peak), and evening/night (main peak). The readings are shown in sub-section 3.2.

3.1 Obtaining the data from the company under study and determining the power consumption of the cell modules to be optimized

The method used for data collection was a time series method of measurement for 27 days at the cell site under study.

In each day, measurements were taken at a period of 2 h, with an interval of every 15 min for 8 times. In the end, an average for the eight intervals was taken for each day for all equipment.

For instance, on Day 1, the 2G BTS Airtel with the current of 25 Amps as the average of eight intervals for every 15 min in the 2 h has the following current readings of 25.4 A; 24.6 A; 25.8 A; 24.3 A; 25.6 A; 24.4 A; 24.7 A; and 25.4 A for the intervals. The average is

$$\begin{array}{rcl}\mathit{\text{Average}}& =& \frac{25.4+24.6+25.8+24.3+25.6+24.4+24.7+25.4}{8}\\ & =& 25.03\text{Amps}\cong 25\text{Amps}\end{array}$$

An example of the 2G BTS Airtel measurement process is shown in Table C1 in Appendix C. The measurement was first performed at the BTS equipment booth, which is the backbone of the cell site that houses the transmitter and receiver modules. BTS is also connected to equipment on the tower for broadcast and hopping activities with other cell sites linked to the tower via RF and microwave antennas. At the end of the measurements performed, a 27-day summary was calculated.

3.2 To determine the module to go on sleep mode and its power requirement

$$\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{V}}_{\text{Aver}}\times {\mathrm{I}}_{\mathrm{T}}\left(\text{Watt}\right)$$E1

where P_Con is the power consumed in (Watt); V_Aver is the average voltage calculated from each day measurement (Volt); I_T is the average total current consumed by the equipment in the cell site (Amps).

• Day 1 at 13:30–15:30 h on 3rd September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=109.8\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.5\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=109.8\times 52.5$

  4. $=5764.50\text{Watts}$

• Day 2 at 11:00–13:00 h on 4th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=102.4\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=50.7\text{Volts}$

  3. $\text{Power consumed}=\text{PCon}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=102.4\times 50.7$

  4. $=5191.68\text{Watts}$

• Day 3 at 15:00–17:00 h on 5th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=110.8\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=110.8\times 52$

  4. $=5761.60\text{Watts}$

• Day 4 at 10:00–12:00 h on 6th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=94.9\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.8\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=94.9\times 52.8$

  4. $=5010.72\text{Watts}$

• Day 5 at 14:00–16:00 h on 9th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=105.6\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=51.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=105.6\times 51.3$

  4. $=5417.28\text{Watts}$

• Day 6 at 10:30–12:30 h on 11th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=98.5\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=98.5\times 52.3$

  4. $=5151.55\text{Watts}$

• Day 7 at 16:00–18:00 h on 12th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=107.9\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.9\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=107.9\times 52.9$

  4. $=5707.91\text{Watts}$

• Day 8 at 12:00–14:00 h on 15th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=97.5\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=97.5\times 52.7$

  4. $=5138.25\text{Watts}$

• Day 9 at 15:30–17:30 h on 17th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=108.3\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=108.3\times 52.4$

  4. $=5674.92\text{Watts}$

• Day 10 at 14:30–16:30 h on 19th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=106.9\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.2\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=106.9\times 53.2$

  4. $=5687.08\text{Watts}$

• Day 11 at 13:00–15:00 h on 20th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=104.7\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=104.7\times 52.7$

  4. $=5517.69\text{Watts}$

• Day 12 at 19:00–21:00 h on 23rd September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=160.8\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=160.8\times 53.7$

  4. $=8634.96\text{Watts}$

• Day 13 at 18:00–20:00 h on 26th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=151.1\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=151.1\times 53.7$

  4. $=8114.07\text{Watts}$

• Day 14 at 20:00–22:00 h on 27th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=144.3\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=141.3\times 52.4$

  4. $=7404.12\text{Watts}$

• Day 15 at 18:30–20:30 h on 28th September 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=151.7\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=151.7\times 53.3$

  4. $=8085.61\text{Watts}$

• Day 16 at 19:30–21:30 h on 2nd October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=153\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=153\times 53.4$

  4. $=8170.20\text{Watts}$

• Day 17 at 06:30–08:30 h on 3rd October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=131.4\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.8\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=131.4\times 52.8$

  4. $=6937.92\text{Watts}$

• Day 18 at 07:00–09:00 h on 6th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=125\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=125\times 53.3$

  4. $=6662.50\text{Watts}$

• Day 19 at 08:00–10:00 h on 7th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=121.2\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=121.2\times 52.4$

  4. $=6350.88\text{Watts}$

  5. $=5138.25\text{Watts}$

• Day 9 at 15:30–17:30 h on 17th October 2019

  1. $\text{Total current}=I_T=108.3\text{Amps}$

  2. $\text{Average voltage}=V_{\text{Aver}}=52.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}=I_T\times V_{\text{Aver}}=108.3\times 52.4$

  4. $=5674.92\text{Watts}$

• Day 10 at 14:30–16:30 h on 19th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=106.9\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.2\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=106.9\times 53.2$

  4. $=5687.08\text{Watts}$

• Day 11 at 13:00–15:00 h on 20th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=104.7\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=104.7\times 52.7$

  4. $=5517.69\text{Watts}$

• Day 12 at 19:00–21:00 h on 23rd October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=160.8\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=160.8\times 53.7$

  4. $=8634.96\text{Watts}$

• Day 13 at 18:00–20:00 h on 26th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=151.1\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=151.1\times 53.7$

  4. $=8114.07\text{Watts}$

• Day 14 at 20:00–22:00 h on 27th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=144.3\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=141.3\times 52.4$

  4. $=7404.12\text{Watts}$

• Day 15 at 18:30–20:30 h on 28th October 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=151.7\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=151.7\times 53.3$

  4. $=8085.61\text{Watts}$

• Day 16 at 19:30–21:30 h on 2nd November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=153\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=153\times 53.4$

  4. $=8170.20\text{Watts}$

• Day 17 at 06:30–08:30 h on 3rd November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=131.4\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.8\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=131.4\times 52.8$

  4. $=6937.92\text{Watts}$

• Day 18 at 07:00–09:00 h on 6th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=125\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=125\times 53.3$

  4. $=6662.50\text{Watts}$

• Day 19 at 08:00–10:00 h on 7th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=121.2\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=121.2\times 52.4$

  4. $=6350.88\text{Watts}$

• Day 20 at 07:30–07:30 h on 9th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=124\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.6\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=124\times 53.6$

  4. $=6646.40\text{Watts}$

• Day 21 at 06:45–08:45 h on 16th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=122.3\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.2\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=122.3\times 52.2$

  4. $=6384.06\text{Watts}$

• Day 22 at 06:00–08:00 h on 17th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=127.8\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.4\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=127.8\times 53.4$

  4. $=6824.52\text{Watts}$

• Day 23 at 07:45–09:45 h on 20th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=120.7\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=120.7\times 52.7$

  4. $=6360.89\text{Watts}$

• Day 24 at 08:15–10:15 h on 22nd November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=117.5\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=52.7\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=117.5\times 52.7$

  4. $=6192.25\text{Watts}$

• Day 25 at 08:15–10:15 h on 26th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=128.8\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.3\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=128.8\times 53.3$

  4. $=6865.04\text{Watts}$

• Day 26 at 08:30–10:30 h on 27th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=122.1\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.1\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=122.1\times 53.1$

  4. $=6483.51\text{Watts}$

• Day 27 at 06:15–08:15 h on 28th November 2019

  1. $\text{Total current}={\mathrm{I}}_{\mathrm{T}}=125.4\text{Amps}$

  2. $\text{Average voltage}={\mathrm{V}}_{\text{Aver}}=53.1\text{Volts}$

  3. $\text{Power consumed}={\mathrm{P}}_{\mathrm{Con}}={\mathrm{I}}_{\mathrm{T}}\times {\mathrm{V}}_{\text{Aver}}=128.8\times 53.3$

  4. $=6658.74\text{Watts}$

  5. $\text{Total}=170098.78\text{Watts}=170.09878\mathrm{kW}=170.1\mathrm{kW}\left(\text{approximation}\right)$

  6. $\text{Number of hours for}27\text{days}=27\times 2=54\mathrm{h}$

  7. $\mathrm{KWH}=170.1\times 54=9185.4\mathrm{kWh}$

  8. $\#60=1\mathrm{kWh}$

  9. $9185.4\mathrm{kWh}=\#9185.4\times 60=\#551124$

3.3 Optimization of the established high power consumed by the modules of the cell site to a minimum value

Days	Time (h)	Power consumption (Watts)
1	13:30	5764.50
2	11:00	5191.68
3	15:00	5761.60
4	10:00	5010.72
5	14:00	5417.28
6	10:30	5151.55
7	16:00	5707.91
8	12:00	5138.25
9	15:30	5674.92
10	14:30	5687.08

Table 1.

Power consumed from the characterized cell site under study.

Minimize

P = X + 13.3Y

SUBJECT TO

X + 13.3Y ≤ 5764.50

7X + 16Y≤ 5707.91

where P is the minimum power consumed by the cell site; X is the day the power is consumed in the cell site; Y is the hour the power is consumed in the cell site.

>> % OPTIMIZING ENERGY EFFICIENCY THROUGH REDUCTION OF POWER CONSUMPTION IN A TELECOMMUNICATION BASE TRANCEIVER STATION (BTS) SITE USING MACHINE LEARNING

%

% Minimize P = X + 13.3Y

%    Subject to

%   X + 13.3Y ≤ 5764.50

%  10X + 14. 3Y≤ 5687.08

%

%  Where

% P is the minimum power consumed by the cell site

%X is the day the power is consumed in the cell site

%Yis the hour the power is consumed in the cell site

  f=[-1;-13.3];

  A=[1 13.3;7 16];

  b=[5764.50;5707.91];

Aeq=[0 0];

beq=[0];

  LB=[0 0];

  UB=[inf inf];

  [X,FVAL,EXITFLAG]=linprog(f,A,b,Aeq,beq,LB,UB)

Optimization terminated.

X =

 0.0000

 356.7444

FVAL =

-4.7447e+003

EXITFLAG = 1

>>

3.4 Designing a machine learning rule base that will monitor the power consumed on the modules and minimize it if high

3.5 Training ANN in the designed machine learning rules for reduced power consumption in the cell site, thereby enhancing its network performance

Figures 3 and 4 will be implemented in the machine learning to enhance its proper functioning to minimize the power consumption in the cell site to save costs.

3.6 Developing an algorithm that will implement 4, 5, and 6

1. Identify the much power consumed by the module of cell site.

2. Optimize the identified much power consumed by the module of the cell site to a minimal.

3. Apply designed machine learning rule base that will monitor the power consumed on the modules and minimize it if high.

4. Apply the trained ANN in 3 to retain minimal power consumption in the module of the cell site.

5. Does the power consumption at the module of the cell site minimized after the application of 4?

6. No, go to 4.

7. Yes, go to 9.

8. Minimized power consumption by the module of the cell site.

9. Stop.

10. End.

11. To develop a power consumption model for the network under study based on results obtained.

3.7 Developing power consumption model for the network under study based on results obtained

The power consumption model is shown in Figure 5 reflecting all the simulations obtained as depicted in Figures 6–9.

4. Results and discussion

The results obtained using machine learning to minimize power consumption are presented and discussed. In Figure 1, we have two inputs of power consumed at the cell site and congestion. It also has an output of results (Table 1).

Figure 2 is a designed machine learning rule base that will monitor the power consumed on the modules and minimize it if high. It monitors the power consumed by the modulus of the cell site and minimizes it when detected high. A comprehensive analysis of the rules is as shown in Table 2 where details of designed machine learning rule base that will monitor the power consumed on the modules and minimize it if high are tabulated. In Figure 3, ANN was trained 10 times in the machine learning five rules to give 50 neurons that look like human brain 10 × 5 = 50. These neurons mimic human intelligence and do what it is instructed to do.

Figure 2.

Designed machine learning rule base that will monitor the power consumed on the modules and minimize it if high.

1	If power consumed at the cell site is high reduce	And congestion at the cell site is high reduce	Then, result is bad
2	If power consumed at the cell site is partially high reduce	And congestion at the cell site is partially high reduce	Then, result is bad
3	If power consumed at the cell site is low maintain	And congestion at the cell site is low maintain	Then, result is good
4	If power consumed at the cell site is high reduce	And congestion at the cell site is partially high reduce	Then, result is bad
5	If power consumed at the cell site is partially high reduce	And congestion at the cell site is high reduce	Then, result is bad

Table 2.

Details of designed machine learning rule base that will monitor the power consumed on the modules and minimize it if high.

Figure 3.

Trained ANN in the designed machine learning rules for reduced power consumption in the cell site thereby enhancing its network performance.

Figure 4 is incorporated in the machine learning to enhance its efficacy in terms of reducing the power consumed in the cell site, thereby enhancing the financial status of the site. Figure 5 is the developed power consumption model for the network under study based on results obtained. The results obtained after simulation are as shown in Figures 6–9. Figure 6 shows the comparison between conventional and machine learning power consumed in cell site in day 1 (Table 3). In Figure 6, the highest conventional power consumed by the cell site is 5764 kW while that when machine learning is inculcated in the system is 4733 kW (Table 4). With these results, it signifies that the percentage improvement in the reduction of power consumed in the cell site when machine learning is incorporated in the system in day 1 is 17.9%.

Figure 4.

Model that resulted in the training.

Figure 5.

Developed power consumption model for the network under study based on results obtained. The results obtained after simulation are as shown in Figures 6–9.

Figure 6.

Comparing conventional and machine learning power consumed in cell site in day.

Figure 7.

Comparing conventional and machine learning power consumed in cell site in day 3.

Figure 8.

Comparing conventional and machine learning power consumed in cell site in day 5.

Figure 9.

Comparing conventional and machine learning power consumed in cell site in day 7.

Time (s)	Conventional power consumed in cell site in DAY 1 (kW)	Machine learning power consumed in cell site in DAY1 (kW)
0	0	0
1	3800	3000
2	5000	4100
3	5300	4500
4	5764	4733
10	5764	4733

Table 3.

Comparison between conventional and machine learning power consumed in cell site in day 1.

Time (s)	Conventional power consumed in cell site in DAY3 (kW)	Machine learning power consumed in cell site in DAY3 (kW)
0	0	0
1	3700	3000
2	5000	4100
3	5500	4500
4	5191	4731
10	5191	4731

Table 4.

Comparison between conventional and machine learning power consumed in cell site in day.

Figure 7 shows that the highest conventional power consumed in the cell site in day 3 is 5191 kW while that when machine learning is integrated in the system is 4731 kW. This clearly showed that the percentage improvement in power consumption reduction in the cell site when machine learning technique is imbibed in the system in day 3 is 8.9%.

In Figure 8, it is obvious that the highest conventional power consumed in the cell site is 5417 kW. On the other hand, when machine learning is integrated in the system, it reduced drastically to 4448 kW, which is 17.9% power consumed by the cell site reduction (Table 5). Figure 9 symbolizes that the highest conventional power consumed by the cell site is 5708 kW while that when machine learning is introduced into the system is 4687 kW, which is 17.9% better that the conventional approach as regards power consumption reduction in the cell (Table 6).

Time (s)	Conventional power consumed in cell site in DAY5 (kW)	Machine learning power consumed in cell site in DAY5 (kW)
0	0	0
1	3200	2700
2	4700	3800
3	5200	4900
4	5417	4448
10	5417	4448

Table 5.

Comparison between conventional and machine learning power consumed in cell site in day 5.

Time (s)	Conventional power consumed in cell site in DAY7 (kW)	Machine learning power consumed in cell site in DAY7 (kW)
0	0	0
1	3400	3000
2	5000	4000
3	5500	4500
4	5708	4687
10	5708	4687

Table 6.

Comparison between conventional and machine learning power consumption in day 7.

5. Conclusion

Sustainable high power consumption at cell sites has reduced the financial status of some of these cell sites. This ugly situation of high cell site power consumption is mitigated by introducing optimized energy efficiency by using machine learning to reduce power consumption at communication base station (BTS) sites. To achieve this enthusiastically, in this process, the related work is checked to find out its shortcomings, and the power consumption of the inspected cell site module is characterized, determined, and inspected. SIMULINK model is developed, specified, and optimized. Machine learning rule base that minimizes the high power consumption of the cell site module monitors the power consumed by the module and minimizes it at high power. Design and train ANNs with designed machine learning rules to reduce power consumption improve its network performance at base stations. Next, we will develop an algorithm that implements it. Finally, based on the results obtained when the algorithm was integrated into the network, we developed a power consumption model for the network under investigation and improved energy efficiency at the cell site with and without machine learning, validated and justified the rate. The results of extensive simulation show that the conventional maximum power consumption of the cell site is 5746 kW, while the maximum machine learning power consumption of the system is 4733 kW. From these results, the improvement rate of power consumption reduction of cell sites by integrating machine learning into the system is 17.9% on the first day, while the maximum power consumption of conventional cell sites is 3 days of machine learning. You can see that it is 5191 kW by eye. Learning is integrated into the system and is 4731 kW. From these results, it can be seen that the improvement rate of the power consumption reduction of the cell site when the machine learning technology is incorporated into the system on the third day is 8.9%, and the conventional maximum power consumed by the cell site is 5417 kW. I understand, on the other hand, when machine learning is integrated into the system, the reduction of cell sites significantly reduces the power consumption to 4448 kW, which is 17.9% of the power consumption, and the conventional maximum power consumption of cell sites is 5708 kW. The learning built into the system is 4687 kW, which is 17.9 superior to the traditional approach in terms of cell site power savings.

Acknowledgments

I appreciate the effort of my son Junior N. Bassey for the sustained interest in assisting to read the manuscript.

Conflict of interest

There is no conflict of interest.