Base station energy consumption collaborative scheduling method and system based on carbon emission constraint

Abstract

This invention relates to the field of energy management and refined carbon emission control of communication base stations. It discloses a base station energy consumption collaborative scheduling method and system based on carbon emission constraints. The method establishes carbon emission accounting models for various power supply energy sources of the base station and accumulates carbon footprints in real time. It solves for the optimal power supply ratio by using a weighted combination of minimizing carbon emissions and minimizing energy costs as the optimization objective. The prediction is updated and re-solved at preset time intervals. When a base station has a surplus of green electricity, the green electricity quota is transferred to other base stations through a regional collaborative platform, and the carbon quota allocation is updated synchronously. This invention elevates the carbon emission factor from a post-hoc statistical parameter to a real-time scheduling decision signal, achieving refined control of base station-level carbon emissions and dynamic balance of carbon quotas among multiple stations in a region.

Description

Technical Field

[0001] This invention relates to the field of energy management and refined carbon emission control of communication base stations, particularly to the field of multi-energy coordinated scheduling under the goal of carbon neutrality, and especially to a method and system for coordinated scheduling of base station energy consumption based on carbon emission constraints. Background Technology

[0002] With the large-scale commercial deployment of 5G mobile communication networks nationwide, the energy consumption of communication base stations has become increasingly prominent and a core issue of concern for operators. Statistics show that base station electricity costs account for more than 30% of operators' total network operation expenditures, and the annual energy consumption of a single macro base station can reach tens of thousands of kilowatt-hours. Under the policy background of carbon peaking and carbon neutrality goals, telecommunications operators face the dual pressure of ensuring network service quality while controlling total carbon emissions.

[0003] Existing base station energy-saving technologies primarily focus on the radio frequency (RF) side, reducing power consumption through methods such as symbol shutdown, carrier shutdown, and deep sleep. However, RF power consumption accounts for only 40% to 50% of the total energy consumption of a base station; the remaining energy consumption comes from fixed loads such as air conditioning cooling, transmission equipment, and power conversion. This means that carbon emission reduction through power consumption reduction solely at the RF side has a theoretical ceiling. More importantly, existing solutions equate carbon emission reduction with a single-dimensional optimization of power consumption reduction, failing to consider the differences in carbon emission factors among different energy sources on the power supply side. The carbon emission factor of photovoltaic power generation is close to zero, while the carbon emission factor of diesel backup generators can reach 800gCO2 / kWh, a difference of two orders of magnitude. This difference indicates that fine-grained scheduling of energy allocation driven by carbon emission factors on the base station power supply side has carbon emission reduction potential comparable to or even greater than that of RF-side energy saving.

[0004] In the field of energy dispatch optimization, Chinese patent application CN117578537A discloses a microgrid optimization dispatch method based on carbon trading and demand response. This method adopts a tiered carbon trading cost model, with the objective function of minimizing the overall operating cost of the microgrid, and constructs a two-stage robust optimization dispatch model including photovoltaic, energy storage, and gas turbines. However, this scheme is geared towards general microgrid scenarios and does not establish a dedicated constraint system for the time-varying characteristics of communication base station service load and service quality assurance requirements. It also does not involve the separate quantification and real-time cumulative calculation of carbon emission factors of various energy sources at the base station level, and lacks a collaborative mechanism for cross-site allocation of green electricity quotas among multiple base stations within the region.

[0005] In the field of base station energy optimization, existing research has incorporated 5G base stations as flexible resources in the distribution network into the low-carbon planning framework, aiming to minimize both economic costs and carbon emissions through the coordinated optimization of renewable energy allocation and base station deployment. However, such solutions focus on site selection and capacity optimization during the network planning phase, rather than real-time energy dispatch during the operation phase, and cannot cope with the dynamic fluctuations in photovoltaic output and service load over a daily timescale. Furthermore, single-site dispatch schemes do not fully utilize the spatiotemporal complementarity of photovoltaic output from multiple stations within a region. When a station has a photovoltaic surplus while neighboring stations have insufficient photovoltaic power, the lack of a cross-site green electricity transfer mechanism will lead to the abandonment of green electricity and the waste of carbon allowances.

[0006] In summary, existing technologies have the following core shortcomings in the refined management of carbon emissions from base stations: they lack a mechanism to transform carbon emission factors from post-hoc statistical parameters into real-time scheduling decision signals; they cannot quantify, calculate in real time, and dynamically schedule carbon emissions from multiple energy sources at the base station level; and they cannot eliminate the spatiotemporal mismatch of photovoltaic supply and demand at a single station through dynamic balancing of green electricity carbon quotas among multiple stations in a region. Summary of the Invention

[0007] To address the bottleneck in existing technologies where carbon emission control of communication base stations lacks a refined means of using carbon emission factors as real-time scheduling decision signals, resulting in the underutilization of carbon reduction potential on the power supply side of base stations, this invention provides a base station energy consumption collaborative scheduling method and system based on carbon emission constraints. By establishing a real-time accounting model of the carbon footprint of multi-source energy at the base station level and driving energy allocation optimization with carbon emission factor gradients, under the constraint of ensuring the quality of service of communication services, this invention achieves joint optimization of minimizing base station carbon emissions and minimizing energy costs from the perspective of energy carbon metering mechanisms.

[0008] The technical solution of this invention is as follows:

[0009] The base station energy consumption collaborative scheduling method based on carbon emission constraints includes the following steps:

[0010] Carbon emission accounting models are established for various power supply energy sources of base stations to calculate the carbon emissions corresponding to the power supply of each type of energy source in real time and accumulate the carbon footprint of the base station. The optimal power supply ratio of each type of energy source is solved with the weighted combination of minimizing carbon emissions and minimizing energy costs as the optimization objective, and with the energy storage state of charge boundary, service load quality assurance requirements and energy output constraints as constraints. Renewable energy output forecast and service load forecast are updated at preset time intervals, and the optimal power supply ratio is solved again based on the updated forecasts. When there is a surplus of renewable energy power generation of a base station, the surplus green electricity quota is transferred to other base stations in the region through the regional coordination platform, and the carbon emission quota allocation of each base station is updated simultaneously.

[0011] This invention also provides a base station energy consumption collaborative scheduling system based on carbon emission constraints, comprising: a carbon emission accounting module, used to establish carbon emission accounting models for various types of power supply energy of the base station, calculate the carbon emissions corresponding to the power supply of various energy sources in real time, and accumulate the carbon footprint of the base station; an energy scheduling optimization module, used to solve for the optimal power supply ratio of various energy sources in the current scheduling cycle with the optimization objective of minimizing carbon emissions and minimizing energy costs, and with the constraints of energy storage state of charge boundary, service load quality assurance requirements, and energy output constraints; a prediction update module, used to update the renewable energy output prediction and service load prediction at preset time intervals, and drive the energy scheduling optimization module to re-solve for the optimal power supply ratio based on the updated predictions; and a regional coordination module, used to transfer the surplus green electricity quota to other base stations in the region through the regional coordination platform when there is a surplus in renewable energy power generation of the base station, and to update the carbon emission quota allocation of each base station synchronously.

[0012] The beneficial effects of this invention include:

[0013] First, this invention establishes a base station-level multi-source energy carbon emission accounting model, assigning differentiated carbon emission factors to four types of energy—mains power, photovoltaics, energy storage, and backup generators—and accumulating their carbon footprints in real time. This transforms the base station's carbon emissions from unobservable, post-hoc statistical parameters into a real-time, perceptible scheduling decision signal. The mechanism lies in the fact that the carbon emission factors of different energy sources differ by two orders of magnitude. By injecting this difference information into the scheduling objective function in real time, energy allocation optimization can be finely controlled in terms of carbon emissions. Existing technologies, which equate carbon reduction with a single path of power consumption and voltage drop, cannot achieve this kind of fine-grained scheduling. Compared to existing microgrid scheduling methods based on tiered carbon trading costs, this invention refines the carbon accounting granularity from the microgrid level to the base station level, and elevates carbon emission information from periodic settlement to real-time accumulation.

[0014] Secondly, this invention establishes a dynamic weighted optimization mechanism between carbon emission constraints and energy costs, and uses a model predictive control framework to achieve rolling solutions, enabling the energy dispatch strategy to adaptively adjust to real-time changes in photovoltaic output and business load. The mechanism involves using the deviation between the cumulative carbon emission value and the carbon quota threshold as an adaptive adjustment signal for the weighting coefficients. When carbon emissions approach the quota limit, the carbon emission weight is automatically increased to trigger priority consumption of green electricity. This adaptive mechanism allows the dispatch system to employ different energy allocation strategies during periods of ample and tight carbon quotas, resulting in carbon emission reduction effects that surpass those of fixed-weight schemes.

[0015] Third, this invention achieves cross-site adjustment of green electricity quotas and dynamic redistribution of carbon quotas among multiple base stations through a regional collaborative platform, thereby resolving the spatiotemporal mismatch between photovoltaic supply and demand at the single-station level at the regional level. The mechanism lies in the natural temporal misalignment between the peak photovoltaic output and the trough of service load at each base station within the region. The carbon quota urgency-based pricing mechanism guides green electricity surpluses from stations with ample carbon quotas to stations with urgent carbon quotas, achieving a Pareto improvement in regional green electricity utilization. This cross-site collaborative effect results in a significantly greater overall regional carbon emission reduction effect than the sum of the effects of independent scheduling at each station, thus generating a nonlinear collaborative gain of 1+1>2. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the base station energy consumption collaborative scheduling method based on carbon emission constraints provided in this embodiment of the invention.

[0017] Figure 2 This is a schematic diagram of the architecture of a base station energy consumption collaborative scheduling system based on carbon emission constraints provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, the base station energy consumption collaborative scheduling method based on carbon emission constraints provided in this embodiment of the invention includes the following steps:

[0020] Step S1: Establish carbon emission accounting models for the various power sources supplying the base station, calculate the carbon emissions corresponding to the power supply of each type of energy source in real time, and accumulate the carbon footprint of the base station.

[0021] In this embodiment, the power supply system of the communication base station consists of four types of energy: mains power, distributed photovoltaic power generation system, energy storage battery system, and diesel backup generator. In actual deployment, the rated load power of a typical macro base station is 3 to 8 kW, of which the power consumption of the radio frequency unit accounts for approximately 40% to 50%, the power consumption of air conditioning heat dissipation accounts for approximately 25% to 30%, and the power consumption of transmission equipment and power conversion accounts for approximately 20% to 25%. The typical capacity of the distributed photovoltaic power generation system installed on or around the base station is 10 to 30 kWp, and the daily power generation is affected by geographical location, season, and weather conditions. In South China, the typical daily power generation in summer is approximately 40 to 100 kWh. The typical configuration capacity of the energy storage battery system is 30 to 100 kWh, using lithium iron phosphate or ternary lithium batteries, with a charge / discharge rate of 0.5C to 1C. The typical capacity of the diesel backup generator is 10 to 30 kW, and it is only activated in the event of a mains power outage or extreme operating conditions. The carbon emission characteristics of each type of energy are fundamentally different. Photovoltaic power generation does not directly emit greenhouse gases during operation, and its carbon emission factor can be considered zero. The carbon emission factor of grid electricity depends on the power grid structure and is typically released by the grid company by region and time period. The carbon emission factor varies significantly across different times of day. During nighttime or midday when wind and solar power generation accounts for a larger share, the grid carbon emission factor can drop below 300 gCO2 / kWh, while during peak electricity consumption periods when fossil fuel output is concentrated, the grid carbon emission factor can rise to over 600 gCO2 / kWh. Diesel standby generators have the highest carbon emission factor among all energy sources, typically around 800 gCO2 / kWh. The carbon emission factor of energy storage batteries is indirect and dynamic; its actual carbon emissions depend on the source of the electricity used for charging.

[0022] Based on the above analysis, this embodiment addresses the scheduling time of the base station. The instantaneous carbon emissions are calculated using the following model: ,in: For base stations during scheduling The carbon emissions generated are scalars, ranging from 0 to positive infinity, with units of gCO2. They are calculated using this formula and represent the total carbon emissions of a base station within a single scheduling time step. The total number of power supply energy types is a positive integer scalar. In this embodiment... These correspond to mains power, photovoltaic power, energy storage, and diesel generators, respectively, and are dimensionless. This is an energy type index, with values ​​ranging from 1 to... Dimensionless, used to traverse various energy sources; For the first Energy-like energy at all times The carbon emission factor is a scalar with a value ranging from 0 to 1000, and the unit is gCO2 / kWh. It is determined by the physical characteristics of the energy type and real-time data released by the power grid, and represents the amount of carbon dioxide emissions generated by consuming 1kWh of this type of energy. For the first Energy-like energy at all times The power supply capacity is a scalar, ranging from 0 to the upper limit of the rated power of this type of energy, and is in kW. It is determined by the solution results of the scheduling optimization module. The time step is the scheduling time step, which is a positive scalar. In this embodiment, it is set to 0.25h, or 15min, with the unit being hours. The selection of this step step balances scheduling accuracy and computational overhead.

[0023] The cumulative carbon footprint of a base station over a statistical period (e.g., a calendar day) is obtained by summing the instantaneous carbon emissions at each scheduling moment. At each scheduling moment... The carbon emission accounting module receives real-time power supply data and carbon emission factor data from various energy sources, calculates the instantaneous carbon emissions at the current moment, and updates the cumulative carbon footprint. Specifically, the carbon emission factor for photovoltaic power generation is fixed at 0, the carbon emission factor for grid power is obtained in real-time from the grid dispatch center, the carbon emission factor for diesel generators is calculated based on the unit's fuel consumption rate, and the equivalent carbon emission factor for energy storage batteries needs to be dynamically determined through carbon traceability methods in subsequent steps.

[0024] Based on the carbon emission accounting model, this embodiment further establishes an energy dispatch priority mechanism driven by carbon emission factor gradients. At the beginning of each scheduling cycle, the system sorts the four types of energy according to the current carbon emission factor of each type of energy, establishing an energy dispatch priority sequence from low carbon to high carbon. Specifically, at a certain scheduling moment... The carbon emission factors of the four types of energy are The system sorts them in ascending order to obtain a priority ranking: ,in: For scheduling time The energy dispatch priority arrangement vector is as follows: A 3D positive integer vector, with values ​​from 1 to... The arrangement is dimensionless and is obtained by sorting the set of carbon emission factors in ascending order. It represents the order in which various energy sources are called at the current moment. The earlier the sort position is, the lower the carbon emission factor should be and the higher the priority should be. For ascending order sorting, return a sequence of indices that sort the elements in the input set from smallest to largest.

[0025] The key innovation of this prioritization mechanism lies in introducing the time-varying characteristics of the grid's real-time carbon emission factor. In traditional schemes, the grid's carbon emission factor is treated as a fixed value, while in this embodiment, the grid's carbon emission factor changes in real-time with the grid's power structure. When the grid is in a period where wind and solar power generation accounts for a high proportion, the grid's carbon emission factor may be lower than the equivalent carbon emission factor of energy storage discharge. In this case, the system automatically prioritizes grid power over energy storage discharge. This dynamic repricing mechanism enables the energy dispatch strategy to accurately track the time-varying waveform of the grid's carbon emission factor, selecting the energy allocation with optimal carbon emissions in each time period.

[0026] Step S2: Taking the weighted combination of minimizing carbon emissions and minimizing energy costs as the optimization objective, and taking the energy storage state-of-charge boundary, service quality assurance requirements of business load, and energy output constraints as constraints, solve for the optimal power supply ratio of various energy sources in the current scheduling cycle.

[0027] In this embodiment, the energy dispatch optimization module constructs and solves a multi-objective optimization problem within each dispatch cycle. The optimization objective function combines carbon emissions and energy costs into a single objective function using weighted coefficients. Constraints cover the upper and lower bounds of the energy storage state of charge (SOC), the output range of various energy sources, power balance, and the quality of service (QoS) requirements of communication service loads. Specifically, the SOC constraint requires that the energy storage battery's SOC at any given time be no less than 10% and no more than 90% to protect battery life and reserve capacity for sudden load increases. Energy output constraints require that photovoltaic power generation not exceed the inverter's rated power, mains power supply not exceed the transformer capacity, and diesel generator output not exceed the rated power and not less than the minimum technical output. Power balance constraints require that the total power supplied by the four energy sources at each dispatch time equal the sum of the base station load power and the energy storage charging and discharging power. These constraints collectively constitute the feasible region of energy dispatch, within which the optimization algorithm searches for the optimal solution of the carbon emissions and energy cost weighted objective function. The calculation of energy cost in the objective function is based on the time-of-use pricing mechanism. The grid electricity cost is calculated by multiplying the peak-valley time-of-use electricity price by the grid electricity consumption. The marginal cost of photovoltaic power generation and energy storage discharge is calculated by the depreciated and amortized cost per kilowatt-hour. The cost of diesel generators is calculated by multiplying the diesel unit price by the fuel consumption rate.

[0028] In the time series alignment analysis of photovoltaic (PV) power output forecasting and base station load forecasting, a key issue is the mismatch between peak PV power output (typically around noon) and low base station load (typically from early morning to late morning). This mismatch results in limited absorption capacity of base stations during peak PV power generation periods, requiring a large amount of PV power to be charged into energy storage or abandoned. This leads to the dynamic accounting problem of the equivalent carbon emission factor of energy storage discharge: the electricity stored in energy storage batteries may come from charging at different times and from different sources, and its carbon emission attributes should be determined through carbon traceability methods.

[0029] The dynamic accounting model for the energy storage equivalent carbon emission factor established in this embodiment is as follows:

[0030] ,in: For energy storage systems at all times The equivalent carbon emission factor during discharge is a scalar with a value ranging from 0 to 1000 and a unit of gCO2 / kWh. It is calculated by this formula and characterizes the amount of carbon emissions indirectly corresponding to each kWh of electricity released in energy storage. Deadline Energy storage from the first The cumulative amount of electricity charged by this type of energy source is a scalar value, ranging from 0 to the rated capacity of the energy storage, and is expressed in kWh. It is determined by the charging records of the energy storage management system. This is an index for the energy type of the charging source, with values ​​ranging from 1 to... Dimensionless; For the first Energy-like substances during their charging time The carbon emission factor is a scalar value, ranging from 0 to 1000, with units of gCO2 / kWh, and uses the carbon emission factor value at the charging time. The round-trip efficiency of the energy storage system, i.e., the charge-discharge round-trip efficiency, is a scalar quantity with a value ranging from 0.85 to 0.95. It is dimensionless and is determined by the technical parameters of the energy storage battery. For typical lithium iron phosphate batteries, the value is about 0.92. This factor reflects the amplification effect of energy loss on the equivalent carbon emission factor during the energy storage charging and discharging process.

[0031] The physical meaning of the above equivalent carbon emission factor is that if all the electricity in energy storage comes from photovoltaic charging ( If the energy in the storage is entirely from mains charging, then the equivalent carbon emission factor for energy storage discharge is zero (ignoring indirect carbon emissions from losses); if all the electricity in the storage comes from mains charging, then the equivalent carbon emission factor is equal to the mains carbon factor at the charging time divided by the cycle efficiency, since... The equivalent carbon factor will be higher than the original grid carbon factor, reflecting the carbon efficiency loss caused by charging and discharging losses. By introducing this dynamic equivalent carbon emission factor, the ranking position of energy storage discharge in the carbon factor gradient priority sequence is no longer fixed, but dynamically fluctuates with the changes in the composition of charging sources, enabling scheduling decisions to more accurately reflect the true carbon emission attributes of energy storage discharge.

[0032] Step S3: Update the renewable energy output forecast and business load forecast according to the preset time interval, and re-solve the optimal power supply ratio based on the updated forecast.

[0033] In this embodiment, the scheduling system employs a model predictive control framework to achieve rolling optimization. Within each rolling optimization time window, the system constructs and solves a finite-time optimal control problem. Specifically, at the current time... The system is for the future For each scheduling time step in the prediction time domain, construct the objective function for the following multi-objective optimization problem:

[0034] The objective function is a dynamic weighted sum of carbon emissions and energy costs. In this objective function, the weighting coefficient for carbon emissions is not fixed but adaptively adjusted based on the deviation between the current cumulative carbon emissions and the carbon quota threshold. This embodiment establishes the following adaptive weighting adjustment mechanism:

[0035] ,in: For a moment The weighting coefficient of carbon emissions in the objective function is a positive scalar, and its value ranges from 1 to 2. The positive infinity, dimensionless, is calculated by this formula and represents the importance of optimizing carbon emissions relative to optimizing energy costs at the current moment. The baseline value for carbon emission weights is a positive scalar with a range of 0.3 to 0.7. It is dimensionless and is set by the operator during system initialization based on the intensity of carbon emission reduction targets. A typical value is 0.5, representing an initial state in which carbon emissions are equally important as energy costs. This is a sensitivity parameter for adaptive weight adjustment. It is a positive scalar with a value range of 1 to 5 and is dimensionless. This parameter controls the steepness of the weight increase with the cumulative carbon ratio. If the value is too small, the adaptive effect will be not obvious and the weight change will be too gentle. If the value is too large, the weight change will be too drastic and may cause frequent switching of scheduling strategies. In actual deployment, a value of 2 to 3 is recommended. Deadline The cumulative carbon emissions of base stations during the current statistical period are positive scalars, ranging from 0 to... The unit is kgCO2, which is updated in real time by the carbon emission accounting module. This is the carbon emission quota threshold for base stations in the current statistical period. It is a positive scalar with the unit being kgCO2, and is determined by the operator after being allocated to each base station according to the carbon neutrality target.

[0036] The physical meaning of the aforementioned adaptive weight adjustment mechanism is that, at the initial stage of the statistical period, the cumulative carbon emissions... Much smaller than the quota threshold The exponential term is close to The carbon emission weight is approximately equal to the benchmark value. At this point, the scheduling strategy balances carbon emissions with energy costs. As the statistical period progresses, carbon emissions gradually accumulate. As the ratio increases, the index term increases, and the carbon emission weight increases. The scheduling strategy automatically shifts towards prioritizing the consumption of green electricity and reducing carbon emissions. When the cumulative carbon emissions approach the quota threshold, the carbon emission weight will increase sharply, and the system will prioritize the use of zero-carbon or low-carbon energy sources with almost no cost. This mechanism enables self-organized optimization of the time allocation of carbon quotas, allowing the appropriate use of high-carbon energy sources to reduce electricity costs during periods of ample carbon quotas, while strongly suppressing carbon emissions during periods of tight carbon quotas to ensure compliance with limits.

[0037] The rolling optimization execution flow of the model predictive control framework is as follows: at each scheduling time step The prediction update module acquires the latest meteorological observation data (including irradiance, temperature, cloud cover, and wind speed) and real-time base station service load data (including user connections, data throughput, and channel utilization), and uses the trained time series prediction model to generate future forecasts. The photovoltaic power output forecast sequence and the service load forecast sequence are calculated for each time step. In this embodiment, the forecast time domain... The value is 12, indicating a prediction of the photovoltaic power output and business load trends over the next 3 hours. Scheduling time step. The time step is set to 15 minutes, taking into account both the typical fluctuation period of photovoltaic power output and the real-time requirements of scheduling calculations. The energy scheduling optimization module constructs a finite-time optimization problem based on the updated predictions and solves it to obtain the future... The system calculates the optimal energy allocation sequence step by step. It executes only the first decision in this sequence, then shifts the time window forward one step to reconstruct and solve the optimization problem with a new initial state. Through this rolling solution method, the scheduling strategy can continuously incorporate the latest forecast information and correct previous forecast errors, achieving adaptive tracking of photovoltaic power output fluctuations and changes in business load. Compared to open-loop methods that solve the entire day's scheduling plan in one go, rolling optimization has inherent robustness. Even if the photovoltaic power output forecast deviates significantly in a certain period, the re-solution in the next time step will automatically incorporate actual power output data for correction, preventing the deviation from accumulating in subsequent periods.

[0038] In terms of renewable energy output forecasting, this embodiment employs a time-series forecasting model trained on historical meteorological data and real-time meteorological observation data, with a forecast time granularity at the minute level. Specifically, the input features of the photovoltaic output forecasting model include forecasted solar irradiance, forecasted ambient temperature, forecasted wind speed, forecasted cloud cover, and actual photovoltaic output time-series data for the previous 24 hours. The forecasting model is built based on deep learning architectures such as Long Short-Term Memory Networks or Gated Recurrent Units, and is put into use after training with no less than 180 calendar days of historical data. Regarding model updates, the system incrementally updates the model every 7 calendar days using the latest actual data to adapt to long-term trends such as seasonal changes and equipment aging. The input features of the business load forecasting model include time period numbers (reflecting intraday periodicity), weekday numbers (reflecting weekly periodicity), holiday / holiday markings, and actual load data for the same period in the previous 7 calendar days. The forecasting model outputs predicted values ​​and their confidence intervals, which are obtained statistically from the results of multiple random discard inferences. Confidence intervals are used to construct robust constraint boundaries for the optimization objective, ensuring that the scheduling strategy can still meet service load guarantee requirements within the prediction error range. Specifically, the service load in the power balance constraint uses the upper bound of the confidence interval instead of the point prediction value, reserving a safety margin for prediction errors.

[0039] Regarding service quality assurance for workload, this embodiment establishes the following constraints: at any scheduling time The total power supply of all types of energy sources shall not be less than the sum of the current service load power of the base station and the preset safety margin, that is... ,in For base stations at time The service load power, To ensure a safety margin of power, and in addition, the power outage time during energy switching will not exceed a preset service interruption tolerance threshold. These constraints ensure that energy dispatching, while pursuing optimization of carbon emissions and costs, will not compromise the quality of communication services.

[0040] Step S4: When there is a surplus in renewable energy power generation at a base station, the surplus green electricity quota is transferred to other base stations in the region through the regional coordination platform, and the carbon emission quota allocation of each base station is updated simultaneously.

[0041] In this embodiment, regional collaborative scheduling is a key mechanism for solving the problem of spatiotemporal mismatch between photovoltaic supply and demand at a single station. In actual operation scenarios, due to differences in geographical location, roof orientation, and surrounding obstructions, the peak time and peak power of photovoltaic power generation vary among base stations within a region. For example, south-facing photovoltaic systems have the highest output at noon, while east- or west-facing systems reach their peak output in the morning and afternoon, respectively. Meanwhile, the service load of each base station is affected by factors such as population density, commercial activities, and traffic flow in its area, exhibiting differentiated temporal distribution characteristics. Base stations in commercial areas have higher loads during weekdays, while base stations in residential areas have higher loads in the evening. This inherent temporal and spatial difference in photovoltaic output and service load provides an objective basis for inter-station green energy allocation within the region. When a base station's photovoltaic power generation exceeds its own absorption capacity at a certain time and its energy storage is fully charged, the surplus green energy, if not utilized, will lead to curtailment and wasted carbon allowances. Through the regional collaborative platform, surplus sites can virtually allocate excess green energy quotas to neighboring sites within the region that have tight carbon allowances, realizing the inter-station transfer of green energy value. Virtual allocation refers to a system where each base station remains physically independent of the power grid. The transfer of green electricity quotas is completed through the accounting system of a regional collaborative platform. Specifically, surplus base stations record their green electricity quotas in the carbon emission accounting account of the receiving base station, which then deducts an equivalent amount from its carbon emission calculation based on its grid electricity consumption. This virtual allocation mechanism avoids the additional construction costs of physical transmission lines and achieves the redistribution of green electricity value within the region through software-level carbon quota accounting management.

[0042] This embodiment establishes a carbon quota urgency index as the basis for allocating cross-site green energy transfer. For the first [unit / region] within the region... The carbon quota urgency of each base station is defined as follows:

[0043] ,

[0044] in: For the first Each base station at time The carbon quota urgency index is a scalar with a value range of 0 to 1 and is dimensionless. It is calculated by this formula. The closer the value is to 1, the more urgent the carbon quota of the base station is, that is, the higher the cumulative carbon emission ratio and the less time left. For the first Base station cutoff time The cumulative carbon emissions are positive scalars, in kgCO2 units, and are updated in real time by the carbon emissions accounting module. For the first The carbon emission quota threshold for each base station in the current statistical period is a positive scalar, in kgCO2, and is allocated by the operator. This represents the total duration of the current statistical period. It is a positive scalar, with a value of 24 when the statistical period is based on natural days. The unit is hours. The cumulative time within the statistical period at the current moment is a scalar, with a value ranging from 0 to 1. The unit is h.

[0045] The design logic of the aforementioned urgency indicator is that it comprehensively considers two dimensions: the cumulative progress of carbon emissions and the remaining time within the statistical period. If a base station has already consumed most of its carbon allowance in the first half of the statistical period ( (Close to 1), and there is still a lot of time remaining ( If the urgency index is close to 1, it indicates that the base station urgently needs to control subsequent carbon emissions by receiving green electricity quotas. Conversely, if a base station has a low carbon quota utilization rate and the statistical period is about to end, the urgency index is low, and the base station has the capacity to allocate green electricity externally.

[0046] Based on the carbon quota urgency index, the regional collaborative platform establishes a green electricity allocation mechanism as follows: During each regional collaborative scheduling cycle, the platform collects the carbon quota urgency and green electricity surplus / deficit information of all base stations in the region; the base stations are sorted from high to low urgency, and the quota of green electricity surplus sites is preferentially allocated to the sites with the highest urgency.

[0047] The adjustment price adopts a tiered pricing mechanism based on the urgency of carbon allowances. (Receiving site) To receive The price paid for a kWh green electricity credit is:

[0048] ,

[0049] in: For the first The price paid by each base station to receive green electricity transfer is a positive scalar in yuan, calculated by this formula, and represents the transaction cost of cross-site green electricity transfer. The benchmark unit price for green electricity adjustment is a positive standard quantity, ranging from 0.1 to 0.5 yuan / kWh, and is set by the regional coordination platform based on the local green electricity market price. The urgency premium coefficient is a positive scalar with a value range of 0.5 to 2.0 and is dimensionless. This parameter controls the influence of urgency on the adjustment price. If the value is too small, the impact of the urgency difference on the price will not be enough to incentivize green electricity to flow from low-urgency sites to high-urgency sites. If the value is too large, the adjustment cost of high-urgency sites will be too high, which may inhibit the willingness to adjust. A value of 1.0 is recommended. For the first The carbon quota urgency index for each base station is defined in the aforementioned formula; For the first The green electricity quota received by each base station during the current adjustment cycle is a positive scalar quantity in kWh, determined by the allocation algorithm of the regional collaborative platform.

[0050] The economic implications of this tiered pricing mechanism are that stations with higher carbon quota urgency pay a higher premium to acquire green electricity, reflecting their stronger demand for carbon emission reduction. Stations with ample carbon quotas gain revenue by selling green electricity, while stations with tight carbon quotas obtain carbon emission reduction benefits by paying a premium. This mechanism creates an inter-station green electricity trading market, guiding the optimal allocation of green electricity within the region through price signals.

[0051] After the green electricity allocation is completed, the regional collaborative platform synchronously updates the carbon emission quota allocation for each base station. Sites receiving green electricity receive an equivalent amount of carbon emission reduction credits, and the portion of their cumulative carbon emissions from grid electricity is replaced by green electricity. Sites transferring green electricity transfer their green electricity credits from their carbon quotas to the recipients, and their carbon emission accounting remains unchanged.

[0052] At a higher time scale, the scheduling system in this embodiment adopts a two-stage collaborative framework of day-ahead planning and intraday rolling. In the day-ahead stage, based on the next day's weather forecast and historical service load patterns, a daily energy scheduling plan and regional green energy allocation plan are generated for each base station the previous evening. In the intraday stage, during the next day's operation, the day-ahead plan is rolled over and corrected under a model predictive control framework based on real-time photovoltaic output and service load deviations. When the actual cumulative carbon emissions deviate from the day-ahead plan by more than a preset deviation threshold, the system triggers the regional collaborative platform to reallocate green energy allocation quotas.

[0053] The determination mechanism for carbon emission deviation triggering redistribution in this embodiment is as follows:

[0054] ,

[0055] in: For the first Base station cutoff time The actual cumulative carbon emissions, as defined above; The first in the recent plan Base station cutoff time The planned carbon emission values ​​are positive scalar values, expressed in kgCO2, and are determined by the day-ahead scheduling plan. The carbon emission deviation trigger threshold is a positive scalar with a value range of 0.05 to 0.20. It is dimensionless and controls the sensitivity of triggering redistribution. If the value is too small, it will trigger frequently, increasing communication and computing overhead. If the value is too large, the deviation will accumulate too much and miss the best adjustment opportunity. The recommended value is 0.10, that is, triggering when the deviation exceeds 10%.

[0056] When the aforementioned deviation conditions are met, the regional collaborative platform initiates an emergency green energy allocation procedure, re-collecting the latest carbon quota urgency and green energy surplus information for each base station, and re-allocating the allocation according to urgency. This mechanism ensures that deviations between the day-ahead plan and the day-ahead actual emissions can be captured and corrected in a timely manner, preventing carbon emission deviations from accumulating continuously within the statistical period.

[0057] In addition, this embodiment also includes a carbon emission intensity ranking and energy-saving retrofit recommendation function. After each preset statistical period, the system calculates the carbon emission intensity index based on the cumulative carbon emissions and energy cost data of each base station, and ranks all base stations in the region. For base stations whose carbon emission intensity exceeds a preset threshold, the system automatically analyzes their energy mix structure, identifies high carbon emission periods and the proportion of high carbon emission energy, and generates an energy-saving retrofit report containing energy mix adjustment recommendations and equipment modification recommendations for operators to refer to in their decision-making.

[0058] The structure of the base station energy consumption collaborative scheduling system based on carbon emission constraints is described below.

[0059] like Figure 2 As shown, the base station energy consumption collaborative scheduling system based on carbon emission constraints provided in this embodiment of the invention includes a carbon emission accounting module, an energy scheduling optimization module, a prediction and update module, and a regional collaboration module.

[0060] The carbon emission accounting module is deployed in the energy management controller of each base station, acquiring real-time data on the power supply and carbon emission factors of various energy sources through a data acquisition interface. Specifically, the module reads the mains power supply and consumption data from the smart meters in the base station power cabinet via an RS485 communication interface, the photovoltaic power generation and cumulative power generation from the photovoltaic inverter via the Modbus protocol, the energy storage charging and discharging power, state of charge, and charging source marking information from the energy storage battery management system via the CAN bus, and the power generation and fuel consumption rate from the diesel generator controller via an analog signal interface. The data acquisition frequency is once per minute, ensuring that the time resolution of carbon emission accounting meets the scheduling decision requirements. Following the carbon emission accounting model described in step S1, the module calculates the instantaneous carbon emissions for the four energy sources: mains power, photovoltaic power generation, energy storage discharge, and standby generators, and establishes an energy dispatch priority sequence according to the carbon emission factor gradient. The carbon emission accounting module is also responsible for calculating the equivalent carbon emission factor of the energy storage system and dynamically updating the order of energy storage discharge in the priority sequence by tracing the source composition of the stored electricity. In terms of data storage, the carbon emission accounting module writes minute-by-minute power supply data, carbon emission factor data, and carbon emission data for various energy sources into a local time-series database, and reports aggregated data to the regional collaborative platform every 15 minutes. The output of the carbon emission accounting module includes the real-time carbon footprint of base stations, energy dispatch priority sequence, and carbon emission quota usage progress.

[0061] The energy dispatch optimization module receives carbon footprint data and priority sequences from the carbon emission accounting module. Combining this with photovoltaic power output forecasts and service load forecasts provided by the forecast update module, it constructs and solves for the optimal energy mix for the current dispatch cycle according to the optimization model described in step S2. This module employs a carbon quota deviation adaptive weight adjustment mechanism, dynamically adjusting the carbon emission weights based on the deviation between the current cumulative carbon emission value and the quota threshold. The output of the energy dispatch optimization module is the power supply command for various energy sources, which is sent to the base station power management system for execution.

[0062] The prediction update module, following the rolling update mechanism described in step S3, acquires the latest meteorological observation data and base station service load data at preset time intervals to update the photovoltaic output prediction and service load prediction. The prediction model is trained based on historical meteorological data and real-time meteorological observation data, outputting the predicted value and its confidence interval. The prediction update module pushes the updated prediction data to the energy dispatch optimization module, driving it to re-solve the optimal power supply ratio under the model prediction control framework, realizing the rolling absorption of prediction information by the dispatch strategy.

[0063] The regional coordination module is deployed in a regional-level virtual power plant platform. Following the regional coordination mechanism described in step S4, it collects information on the carbon quota urgency and green electricity surplus / deficit of each base station within the region, and executes tiered pricing based on carbon quota urgency and green electricity allocation. The regional coordination module is also responsible for executing the day-ahead and intraday coordinated scheduling. During the day-ahead phase, it generates daily energy scheduling plans for each base station and regional green electricity allocation plans. During the intraday phase, it monitors carbon emission deviations and triggers a reallocation of green electricity when the deviation exceeds a threshold. The output of the regional coordination module includes green electricity allocation instructions and carbon quota update notifications for each base station.

[0064] To verify the technical effectiveness of the method of this invention, a field test was conducted in a city in South China, selecting an area consisting of 15 macro base stations. Each base station was equipped with a 20kWp distributed photovoltaic power generation system and a 50kWh lithium iron phosphate energy storage battery. The grid power connection adopted peak-valley time-of-use pricing, with a peak-hour price of 1.05 yuan / kWh, a normal-hour price of 0.65 yuan / kWh, and a valley-hour price of 0.35 yuan / kWh. Each station was equipped with a 10kW diesel backup generator. The test area was located near 23 degrees north latitude, with an average annual sunshine duration of approximately 1600 hours. The test period coincided with the summer months when photovoltaic output was strongest. The test period was 30 consecutive calendar days, covering various weather conditions including sunny, cloudy, and rainy days. The comparison schemes are set as follows: Scheme A is the method of this invention (carbon emission constraint cooperative scheduling), Scheme B is a traditional radio frequency energy-saving scheme (symbol shutdown loading wave shutdown but no power supply side carbon scheduling), and Scheme C is a fixed-weight multi-objective optimization scheme (carbon emission weight is fixed and there is no regional coordination). The three schemes are operated simultaneously under the same base station hardware conditions and external environment. The technical effectiveness of this invention is evaluated by comparing the three core indicators of carbon emissions, green electricity utilization rate, and energy cost of each scheme.

[0065] The measured results show that, in terms of carbon emissions, compared to Option B, the average daily carbon emissions per station under Option A decreased from 85.3 kg CO2 to 65.6 kg CO2, a reduction of 23.1%. Compared to Option C, the average daily carbon emissions per station under Option A decreased from 71.8 kg CO2 to 65.6 kg CO2, an additional reduction of 8.7%. Regarding green electricity utilization, the regional green electricity utilization rate increased from 65.2% under Option B and 78.1% under Option C to 87.6% under Option A, while the green electricity curtailment rate decreased from 34.8% under Option B to 12.4% under Option A. In terms of energy costs, the average daily electricity cost per station decreased from 126.5 yuan under Option B and 115.2 yuan under Option C to 108.7 yuan under Option A, reductions of 14.1% and 5.6%, respectively. Further analysis of the carbon emission reduction sources reveals that regional coordinated allocation contributed 5.3% to the additional carbon emission reduction from Scheme A compared to Scheme C, while the adaptive weighting mechanism contributed 3.4%. The tiered pricing based on carbon quota urgency reduced the standard deviation of carbon emissions from 12.7 kg CO2 in Scheme C to 4.3 kg CO2 in Scheme A, a reduction of 66.1%, indicating a significantly more balanced distribution of carbon quotas within the region. Under extreme weather conditions (three consecutive days of rain leading to severe photovoltaic power shortages), Scheme A, through regional coordinated allocation, obtained green electricity quotas from neighboring sites with relatively sufficient photovoltaic power, keeping carbon emissions within the quota threshold. In contrast, Scheme C exceeded the quota by 12% under this scenario.

[0066] The above measured data verify the technical effectiveness of this invention in terms of refined carbon emission control and regional collaborative optimization: the refined scheduling driven by the carbon emission factor gradient enables optimal control of the carbon content of each kilowatt-hour, the adaptive weighting mechanism enables the self-organized allocation of carbon quotas in the time dimension, and the regional collaborative adjustment enables the optimal cross-station allocation of green electricity in the spatial dimension. The synergistic effect of the three factors produces a nonlinear carbon emission reduction gain that exceeds the sum of the independent effects of each mechanism.

[0067] The core mechanism of this invention is as follows: differentiated quantification of carbon emission factors → energy priority ranking driven by carbon factor gradient → dynamic tracking of energy storage equivalent carbon factors → adaptive weight adjustment of carbon quota deviation → regional carbon quota urgency gradient pricing → day-ahead and intraday deviation triggering correction. In this chain, underlying innovation (the inversion of carbon emission factors from post-hoc statistical parameters into real-time scheduling decision signals) drives algorithmic innovation (adaptive weighting, urgency gradient pricing, and deviation-triggered redistribution), forming a causal and complementary relationship. The higher the precision of carbon emission accounting, the more accurate the driven scheduling decisions, and the more significant the resulting carbon reduction effect. This mechanism means that the technical effect of this invention is not a linear superposition of its various technical features, but rather exhibits nonlinear synergistic gain characteristics.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A base station energy consumption collaborative scheduling method based on carbon emission constraints, characterized in that, Includes the following steps: Carbon emission accounting models are established for the various power sources of base stations, and the carbon emissions corresponding to the power supply of each type of energy source are calculated in real time and the carbon footprint of the base station is accumulated. With the optimization objective of minimizing carbon emissions and minimizing energy costs, and with constraints such as the energy storage state-of-charge boundary, service quality assurance requirements for business loads, and energy output constraints, the optimal power supply ratio of various energy sources in the current scheduling cycle is solved. The renewable energy output forecast and the business load forecast are updated at preset time intervals, and the optimal power supply ratio is re-solved based on the updated forecasts. When a base station has a surplus in renewable energy generation, the surplus green electricity quota is transferred to other base stations in the region through the regional coordination platform, and the carbon emission quota allocation of each base station is updated simultaneously.

2. The method according to claim 1, characterized in that, The carbon emission accounting model assigns differentiated carbon emission factors to four types of energy: grid power, photovoltaic power generation, energy storage discharge, and standby generators. It establishes an energy dispatch priority sequence according to the gradient order of carbon emission factors from low to high. In each dispatch cycle, it prioritizes the energy with the lowest carbon emission factor to its output limit before dispatching the energy with the next lowest carbon emission factor. The model also dynamically adjusts the ranking position of grid power in the priority sequence based on the time-varying characteristics of the grid's real-time carbon emission factor.

3. The method according to claim 2, characterized in that, Based on time series alignment analysis of photovoltaic power output prediction curve and base station service load prediction curve, the mismatch interval between photovoltaic power output peak and service load trough is identified. Within the mismatch interval, energy storage charging scheduling is advanced to the photovoltaic power output ramp-up stage to improve the local photovoltaic consumption rate. The equivalent carbon emission factor of energy storage discharge in the carbon emission factor gradient sequence is dynamically corrected according to the charging and discharging status of energy storage. The equivalent carbon emission factor is equal to the product of the source weighted carbon emission factor of the stored electricity and the energy storage charging and discharging loss factor.

4. The method according to claim 3, characterized in that, The optimal power supply ratio is solved using a model predictive control framework. A finite-time optimal control problem is constructed within each rolling optimization time window. The objective function is a dynamic weighted sum of carbon emissions and energy costs. The weight coefficient of carbon emissions is adaptively adjusted based on the deviation between the current cumulative carbon emissions and the carbon emission quota threshold. When the cumulative carbon emissions approach the quota threshold, the carbon emission weight is automatically increased to trigger the priority consumption of green electricity. The rolling optimization adopts the first optimal solution at each time step and shifts the time window forward.

5. The method according to claim 4, characterized in that, The regional collaborative platform constructs a carbon quota urgency index based on the deviation ratio between the cumulative carbon emissions of each base station in the region and its respective carbon quota threshold. It establishes a priority receiving order for green electricity allocation according to the carbon quota urgency from high to low. The green electricity quota of the carbon quota surplus site is preferentially allocated to the site with the most carbon quota urgency. The allocation price is priced according to the carbon quota urgency gradient, so that the site with abundant carbon quota obtains the green electricity transfer revenue and the site with carbon quota urgency obtains the carbon emission reduction benefits.

6. The method according to claim 5, characterized in that, The coordinated scheduling adopts a two-stage coordinated framework of day-ahead planning and intraday rolling. In the day-ahead stage, the daily energy scheduling plan and regional green electricity allocation plan for each base station are generated based on the next day's weather forecast and historical service load patterns. In the intraday stage, the day-ahead plan is rolled over and corrected based on real-time photovoltaic output and service load deviation under the model prediction and control framework. When the actual cumulative carbon emissions deviate from the day-ahead plan by more than a preset deviation threshold, the regional coordinated platform is triggered to reallocate the green electricity allocation quota.

7. The method according to claim 1, characterized in that, The service load quality assurance requirements include: at any scheduling time, the total power supply of various energy sources is not less than the sum of the current service load power of the base station and the preset safety margin, and the power supply interruption time during the energy switching process does not exceed the preset service interruption tolerance threshold.

8. The method according to claim 1, characterized in that, The renewable energy output prediction adopts a time series prediction model trained based on historical meteorological data and real-time meteorological observation data. The prediction time granularity is at the minute level, and the predicted value and its confidence interval are output. The confidence interval is used to construct the robust constraint boundary of the optimization objective.

9. The method according to claim 1, characterized in that, Also includes: A carbon emission intensity ranking is generated based on the cumulative carbon emissions and energy cost data of each base station within a preset statistical period. For base stations whose carbon emission intensity exceeds a preset threshold, an energy-saving renovation report containing suggestions for energy ratio adjustment and equipment modification is automatically generated.

10. A base station energy consumption collaborative scheduling system based on carbon emission constraints, used to implement the method according to any one of claims 1-9, characterized in that, include: The carbon emission accounting module is used to establish carbon emission accounting models for the various power supply energy sources of the base station, calculate the carbon emissions corresponding to the power supply of each type of energy source in real time, and accumulate the carbon footprint of the base station. The energy dispatch optimization module is used to solve for the optimal power supply ratio of various energy sources in the current dispatch cycle, with the optimization objective being a weighted combination of minimizing carbon emissions and minimizing energy costs, and with constraints being the energy storage state of charge boundary, service quality assurance requirements for business loads, and energy output constraints. The forecast update module is used to update the renewable energy output forecast and the business load forecast at preset time intervals, and drive the energy dispatch optimization module to re-solve the optimal power supply ratio based on the updated forecasts. The regional coordination module is used to allocate surplus green electricity quotas to other base stations in the region through the regional coordination platform when there is a surplus in renewable energy generation at the base station, and to update the carbon emission quota allocation of each base station in sync.

Images