Energy-Efficient Base Station Sleep Timing Optimizer

Abstract

The invention is an end-to-end multi-objective optimization platform that minimizes the RAN energy consumption and corresponding Scope 2 carbon emissions of a mobile network operator without violating coverage and QoS obligations. The invention offers an integrated platform where spatio-temporal traffic estimation, carbon intensity estimation, hardware thermal / aging modeling, coverage gap validator, and a multi-level sleep decision engine are combined in an integrated multi-objective MILP optimization solver.

Description

1 TARIFF Energy-Efficient Base Station Sleep Timing Optimizer Technical Area 5 The invention relates to telecommunications and mobile radio access networks (Radio Access Network — RAN). energy management, AI-based grid optimization, and carbon-awareness Energy-sensitive base station sleep timing optimizer used in computing fields. It is related to. State of the Art Mobile network equipment manufacturers offer a variety of products for RAN energy saving. The vast majority of these can be solved with simple threshold-based heuristics (e.g., transported PRB 5%). If it's below that level, turn off certain MIMO layers, especially between 02:00 and 05:00 at night. (carrier muting) works; 3GPP standards include Cell DTX, MIMO antenna muting and carrier 15 It defines technical mechanisms such as shutdown (3GPP TR 38.864, TS 38.300, TS 38.331). In academic literature, reinforcement learning (DRL) based BS on / off and ARIMA are discussed. These include traffic forecasting based on data transfer and renewable energy matching approaches. The main shortcomings of current techniques are as follows: One-dimensional threshold logic: The vast majority of commercial products only respond to the current traffic load with a 20-degree threshold. It considers traffic forecasting, carbon intensity, and equipment thermal / aging costs together. It does not establish a multi-objective optimization process for evaluation. Carbon-blind design: Current products save energy (kWh); however, because it does not take into account the carbon intensity of the electricity grid at that hour (gCO₂eq / kWh), Clearly suboptimal, such as putting people to sleep during windy nighttime hours rather than coal-heavy hours. 25 They can make decisions. Ignoring the cost of hardware aging: Frequent on / off cycles including PA and fan. including the lifespan of electronic components, according to Coffin-Manson style thermal fatigue laws. It shortens the cost accordingly; existing products do not include this cost in the optimization objective. Coverage-unawareness: Closing a BS alone will result in 30% of the energy that neighboring cells will provide. When implemented without a validator that quantifies coverage, coverage gaps and cellular issues occur. This creates edge QoS degradation. Lack of renewable energy integration: such as solar panels + batteries at the BS site. The occupancy rate of micro-production / storage infrastructure is not assessed in conjunction with the decision; “local A strategy of "work with renewables, sleep during fossil fuel-intensive hours" cannot be created. 35 2 Lack of regulatory compliance: Even if existing products generate savings reports, they are not GHG (High-Gross Domestic Product). The protocol conforms to the Scope 2 methodology and the ESRS E1 standard, and is verifiable by a third party. It is not in (auditable) format; this situation is reflected in large enterprise reporting under CSRD. It does not fulfill its obligation. Reactive wake-up: When an unexpected increase in traffic or an emergency occurs, the sleep-induced 5 The wake-up order for BSs is usually determined by randomness or a simple FIFO rule; geographically It is not integrated with priority (e.g., hospital, highway) and public safety alerts (CAP). When the above shortcomings are combined, the operator may either misjudge the BS sleep timing or it may be too much. either be conservative and not achieve significant energy savings, or be aggressive and... risk of coverage issues, reduced equipment lifespan, or being woken up during windy hours. It is exposed to carbon suboptimality, such as being put to sleep during coal-intensive hours. In addition, The savings achieved must be documented with auditable evidence in the CSRD / ESRS E1 report. The inability to access these resources reduces both financial and ESG value creation below the possible level. In conclusion, due to the negative aspects described above and the current solutions being the subject of discussion... Due to its shortcomings, it has become necessary to make improvements in the relevant technical field. 15 Purpose of the Invention The invention was created by drawing inspiration from existing situations and overcoming the aforementioned drawbacks. It aims to solve the problem. The main purpose of the invention is to determine the RAN energy consumption of the mobile network operator and the corresponding 20 Scope 2 minimizes carbon emissions without violating coverage and QoS obligations. Eden's goal is to provide an end-to-end multi-purpose optimization platform. Another purpose of the invention is to predict spatio-temporal traffic, carbon intensity, and hardware. thermal / aging model, coverage gap validator and multi-level sleep decision The integrated 25 engine is combined into an integrated multi-purpose MILP optimization solver. to provide a platform. Another aim of the invention is to provide both energy-carbon ROI and auditable ESG for the operator. The report generates aggressive RAN sleep scheduling without risking coverage or hardware lifespan. to provide. The structural and characteristic features and all the advantages of the invention are given in the figures below and in these 30 Thanks to the detailed explanation written with references to the figures, it becomes clearer. It will be understood. Figures that will help understand the invention. Figure 1 shows the general architecture of the system that is the subject of the invention. 35 Explanation of Part References 3 1. Multi-Source Telemetry and Environmental Data Collection Module 2. Spatial-Temporal Traffic Prediction Engine 3. Carbon Density Estimation Unit 4. Hardware Thermal and Aging Cost Model 5. Coverage Gap Verifier 5 6. Multi-Objective Optimization Solver 7. Multi-Level Sleep Decision Engine 8. O-RAN A1 / E2 and SBA Policy Adapter 9. Renewable Energy and Battery Coupler 10. Policy and Regulatory Compliance Unit 10 11. Feedback and Closed-Loop Validator 12. Decision Audit and Carbon Accounting Unit 13. Emergency Wake-Up Coordinator Detailed Description of the Invention 15 In this detailed explanation, the preferred configurations of the invention are not merely for better understanding the subject. It is intended to facilitate understanding and will not create any limiting effects. The invention relates to mobile networks. the operator's RAN energy consumption and the corresponding Scope 2 carbon emissions, An end-to-end multi-purpose solution that minimizes coverage and QoS obligations without violating them. It is an optimization platform. 20 The operating principle of the system described in the invention includes the following elements and process steps: It is explained. In the first stage, the Multi-Source Telemetry and Environmental Data Collection Module (1) collects data from each base station. energy efficiency KPIs (DataEC, DataEV, defined in 3GPP TS 28.554) from the station PEE), number of users connected to RRC, PRB usage, cell output power and hardware sensor 25 data (case temperature, fan speed); as well as carbon density from the national grid. data is transmitted via the ENTSO-E Transparency Platform and the EPİAŞ Transparency Platform, air Status forecast from ECMWF / MGM sources, and if available, integrated solar panel / battery on the BS. It receives data via Modbus / SunSpec. All data is geotagged and in real-time. It is written to the flow bus. 30 In the second stage, the Spatial-Spatial Traffic Prediction Engine (2) predicts the next 1–24 for each cell. It estimates hourly traffic. The model is defined on the cell-to-cell adjacency graph G=(V,E). Spatial correlation with a Graph Attention Network (GAT) component, one per cell It learns the temporal correlation with the Temporal Convolutional Network (TCN) component. The output is for each For cell i, the probabilistic value is always in the form (μ_{i,t}, σ_{i,t}) = ST-GNN(x, G, θ) at step t. This is a traffic forecast; where x are telemetry inputs and θ are model parameters. 4 In the third stage, the Carbon Intensity Estimation Unit (3) depends on the production mix in the grid. It predicts carbon intensity (gCO₂eq / kWh) for the next 24–48 hours; the basis it uses The formula is C_t = Σ_f (P_{f,t} · EF_f) / Σ_f P_{f,t}, where P_{f,t} is for each fuel / source f. Power generation at time t and EF_f is the emission factor of this source (IPCC AR6 values). Simultaneously, Renewable Energy and Battery Matcher (9), solar panel 5 in BS site It calculates the production forecast and the battery charge level (SoC). In the fourth stage, the Hardware Thermal and Aging Cost Model (4), each sleep level L ∈ The wake-up cost for {L1, …, L5} is calculated as W_L = k_th · (T_wake,L − T_amb)² + k_aging · Calculations are made in the form (ΔT_L)^β · N_cycles. The first term comes from the thermal RC model. awakening energy, the second term is Coffin-Manson type thermal fatigue-induced aging 10 It represents the cost. At the same time, the Coverage Gap Verifier (5), each BS candidate for closure using radio emission mapping of neighboring cells and cell edge SINR calculation Quantifies the risk of coverage gap and cell edge QoS degradation. In the fifth stage, the Multi-Objective Optimization Solver (6) optimizes the MILP model to the following objective: It solves under the function: 15 min Σ_{i,t} (α·E_{i,t} + β·C_t·E_{i,t} + γ·Pen_QoS_{i,t} + δ·Pen_Cov_{i,t} + ε·W_{L,i,t}) Here, E_{i,t} is the energy consumption of cell i at time t, C_t·E_{i,t} is the carbon emission, and Pen_QoS is... and Pen_Cov violation penalties, W_{L,i,t} is the wake-up cost. Constraints; traffic per cell. fulfilling the request, not exceeding the coverage load of neighboring cells, emergency call 100% coverage guarantee, BTK coverage obligation, corporate SLAs, battery 20 Discharge limits and wake-up delay time. The solver is LP- for real-time operation. It uses relaxation + rolling instinct. In the sixth stage, the Multi-Level Sleep Decision Engine (7) assigns L1 (symbol level DTX, for each BS, L1 (submilliseconds), L2 (slot level DTX), L3 (carrier shutdown), L4 (deep sleep — RF off, BBU selects the appropriate level between minimum and L5 (cell shutdown). Low traffic short 25 L1 / L2 for windows, L3 / L4 for medium-length predictable low windows, long night L5 is preferred for windows; each choice comes with the wake-up cost generated by DTYM (4). It is balanced. In the seventh stage, the Policy and Regulatory Compliance Unit (10), before the action is implemented BTK coverage regulation, 112 / E112 emergency call accessibility guarantee, corporate customer 30 Verifies compliance with SLAs and CSRD / ESRS E1 obligations; any breaches This prevents the action or reduces it to an alternative level. In the eighth stage, O-RAN A1 / E2 and SBA Policy Adapter (8) transmit the decision to the appropriate commands. It translates and implements: (i) O-RAN sends the policy of module A1, which runs as an rApp, to the Non-RT RIC; (ii) Module E2 service model commands (E2SM-CCC — 35) that run as xApp to Near-RT RIC. Cell Configuration and Control; E2SM-RC — RAN Control); (iii) gNB scheduler, Slot / symbol level switches to DTX mode or carrier is shut down; (iv) on 5GC side Cell closure announcement is made via Namf_Communication, Nnssf is updated; (v) The sleep path of the relevant optical / Eth line is connected to the transport SDN controller via NETCONF / YANG. The system is activated. In the ninth stage, Feedback and Closed Loop Validator (11), post-implementation 5 It measures actual energy consumption, carbon emissions, coverage, and QoS metrics; estimated The difference Δ between savings (Ê) and realized savings (E_obs) is given by r = α·Δ − β·penalty It converts this into a reward-punishment system and feeds it back into the weights (α, β, γ, δ, ε) of the CAOQ. In the tenth stage, the Emergency Wake-Up Coordinator (13) dealt with unexpected traffic. explosion, neighboring cell malfunction, high volume of emergency calls or public 10 originating from AFAD / CAP It is activated when a security alert is detected; dormant BSs are given geographical priority (hospital, It sorts and instantly wakes up the area (highway, crowded area, critical infrastructure) using an algorithm. Wake up During this process, the warm-up period is managed by DTYM, and the coverage risk is monitored by KBD. In the eleventh and continuous phase, the Decision Audit and Carbon Accounting Unit (12) monitors every hour kWh saved, gCO₂eq avoided, renewable kWh used and grid 15 for the segment Carbon intensity is measured using GHG Protocol Scope 2 location-based and market-based methods. It accounts accordingly; CSRD / ESRS E1 (specifically with E1-5 energy consumption and mix). It produces reports in a third-party verifiable format under the headings of E1-6 GHG emissions. The report serves both as a data layer for the annual sustainability report and as a performance indicator. In the pricing model based on a percentage of savings, proof for supplier payment is 20. provides. The core innovation of the invention is; (a) spatio-temporal traffic estimation, dynamic grid carbon density and hardware thermal / aging cost in a single multi-purpose MILP (b) co-modeling in optimization, from symbol level to cell closing (c) 25 Determination of the cost of waking up each of the five levels of sleep depth. Coverage gap verifier and emergency call guarantee action for regulatory compliance (d) GHG Scope 2 location-based and market-based accounting fully compliant with CSRD / ESRS E1, verifiable by third parties. (e) ensuring the public with the geographic priority algorithm of the emergency wake-up coordinator Integrated operation with safety alerts and (f) O-RAN A1 / E2 + 5GC SBA + transport SDN 30 It is the simultaneous implementation of a multi-interface coordinated application. These six aspects This combination provides a novel technique that cannot be derived from a known combination of known techniques.

Claims

6 REQUESTS 1. It is an energy-sensitive base station sleep timing optimizer, and its features include: Telemetry data from each base station, electricity grid carbon density. measurements, weather forecasts, and, if applicable, solar data integrated onto the base station. Multi-source telemetry and environmental data simultaneously collecting panel / battery controller data 5 collection module (1), Graph neural network that predicts the traffic density per cell in 1–24 hours Spatial-temporal traffic prediction engine (2), which is a Neural Network based model, Instantaneous and forward-looking carbon footprint, varying depending on the electricity generation mix in the grid. Carbon intensity estimation unit (3), which estimates the intensity (gCO₂eq / kWh), 10 Warm-up for base station power amplifier (PA), band processing unit (BBU), and chassis fan. representing the cooling dynamics with a first-order thermal RC model; each wake-up- the thermal stress created by the sleep cycle and the cost of electronic component aging Equipment thermal and aging costs quantified using the Coffin-Manson empirical law. model (4), 15 If one or more base stations are put into different sleep levels, neighboring Estimation of the coverage that the cells will provide and the cell edge signal-to-noise ratio (SINR). Coverage gap verifier (5), energy consumption, carbon emissions, coverage risk, and QoS latency targets together 20 that evaluates mixed integer linear programming (MILP) based solvers purposeful optimization solver (6), Five sleep levels compliant with 3GPP TR 38.864 (NR Network Energy Savings). Multilevel sleep decision engine (7), To implement the given sleep decision, rApp and Near- on O-RAN Alliance Non-RT RIC. RT RIC is spoken as an xApp, following the A1 policy and E2 service model (E2SM-CCC, 25 O-RAN A1 / E2 and SBA policy adapter (8) which generates E2SM-RC) commands, If the base station site has integrated solar panels and batteries; production estimate, battery charge level. Base station consumption profile with level of discharge (SoC), deep discharge limit and lifecycle cost curve. pairing and implementing a "active during fossil fuel-intensive periods, dormant during active periods while working with local renewables" strategy. Renewable energy and battery matching prioritizer (9), 30 BTK coverage obligations, emergency call (112 / E112) accessibility guarantee, corporate customer SLAs and EU CSRD / ESRS E1 (Scope 2 emissions reporting) obligations Policy and regulatory compliance unit that defines the rule and applies it to all modules (10), Energy saved (kWh), emissions prevented (gCO₂eq), and emissions generated after each sleep decision. Monitoring QoS deviations and customer complaint signals in real time; measuring decision effectiveness. 7 And feedback and a closed loop that feeds this back into the next round of optimization. verifier (11), Greenhouse Gas (GHG) Protocol in accordance with Scope 2 methodology for each time zone kWh saved, gCO₂eq avoided, renewable kWh used, and carbon emissions. Decision audit and carbon accounting unit (12), accounting for density values, 5 unexpected traffic surge, neighboring cell failure, emergency call overload, or public safety. In Common Alerting Protocol (CAP) alert situations, dormant base stations The quickest way to wake someone up; the order of wake-up is determined by geographical priority (hospital, highway, crowd). Emergency wake-up coordinator (13) who determines with the algorithm (field) It includes. 10

Images