A method for optimizing the optical storage capacity of a smart base station

Abstract

This invention relates to the field of power system technology and discloses a method for optimizing the configuration of photovoltaic (PV) and energy storage capacity in smart base stations. The method includes acquiring historical load and PV output data of the smart base station, constructing a relationship between temperature and the load power of high-power equipment; using minimum net expenditure as the objective function, incorporating electricity purchase costs, energy storage investment costs, energy storage loss costs, and the revenue from participating in grid peak-shaving ancillary services, constructing an optimization model for the operation of the PV-energy storage grid-connected system; solving the optimization model to determine the initial PV-energy storage capacity configuration scheme; and dynamically correcting the initial PV-energy storage capacity configuration scheme based on communication forward-looking predictions and the thermal inertia buffering effect of buildings to determine the optimal PV-energy storage capacity configuration scheme for the smart base station. This invention achieves precise configuration and efficient operation of the PV-energy storage system from multiple dimensions of communication, energy, and thermal environment, providing systematic technical support for the green, low-carbon, and intelligent development of smart base stations.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and more specifically, to a method for optimizing the configuration of optical and energy storage capacity in a smart base station. Background Technology

[0002] With the rapid deployment of 5G communication networks, the energy consumption and power supply reliability requirements of smart base stations are becoming increasingly prominent. Currently, the configuration of photovoltaic and energy storage systems for smart base stations is mainly based on static design methods using historical load averages and photovoltaic output, ensuring power supply through pre-set fixed-capacity photovoltaic panels and energy storage batteries. These solutions typically employ a "large-scale replacement of small-scale" redundancy configuration strategy, for example, determining the energy storage capacity based on the base station's maximum load demand and configuring the photovoltaic scale in conjunction with local average solar irradiance conditions. Some improved solutions introduce time-of-use pricing mechanisms, reducing electricity purchase costs by charging during off-peak hours and discharging during peak hours. However, existing technologies generally rely on fixed-parameter models and lack real-time response capabilities to dynamic changes in communication services and the coupling effects of the thermal environment, resulting in the following technical bottlenecks:

[0003] 1) The periodic and sudden characteristics of communication traffic were not taken into account. Especially during holidays and large-scale events, the base station load may surge by 30%-50%, which means that the statically configured energy storage capacity cannot meet the demand and it is necessary to rely on high-priced power grid to supplement the power.

[0004] 2) The load of high-power equipment (such as air conditioners and servers) in base stations is strongly correlated with indoor temperature, but existing solutions do not quantify the buffering capacity of building thermal inertia (such as wall heat storage) to load fluctuations. A surge in air conditioning load during high-temperature periods may trigger excessive discharge of energy storage, while the heat storage capacity of walls is not fully utilized, resulting in redundant or insufficient energy storage capacity;

[0005] 3) Traditional photovoltaic-storage configurations only focus on electricity purchase costs and do not comprehensively consider multiple objectives such as grid peak-shaving ancillary service revenue and photovoltaic absorption rate improvement, making it difficult to achieve a balance between economic benefits and grid interaction value;

[0006] 4) The lack of a real-time linkage correction model between communication traffic prediction and thermal inertia parameters results in the inability of the photovoltaic storage capacity configuration to adapt to long-term operating conditions such as seasonal changes and equipment aging.

[0007] The aforementioned problems have led to a common contradiction in existing photovoltaic and energy storage systems: "over-configuration" and "insufficient peak-shaving capacity." This is especially true in remote base stations, where it is difficult to balance the economic efficiency and reliability of photovoltaic and energy storage systems. Summary of the Invention

[0008] To address the problems in related technologies, this invention proposes a method for optimizing the optical storage capacity configuration of smart base stations, thereby overcoming the aforementioned technical problems existing in existing related technologies.

[0009] Therefore, the specific technical solution adopted by the present invention is as follows:

[0010] A method for optimizing the optical storage capacity configuration of a smart base station includes the following steps:

[0011] S1. Obtain historical load data and photovoltaic output data of smart base stations, analyze the power consumption characteristics of high-power equipment, and construct the relationship between temperature and load power of high-power equipment to visualize the changes in power load of high-power equipment.

[0012] S2. Taking the minimum net expenditure as the objective function, the system incorporates electricity purchase costs, energy storage investment costs, energy storage loss costs, and revenue from participating in grid peak shaving auxiliary services, and constructs an operation optimization model for the photovoltaic-storage grid-connected system in combination with constraints.

[0013] S3. Clarify the energy storage charging and discharging and peak shaving strategies. Based on the optimization algorithm, solve the operation optimization model of the photovoltaic-storage grid-connected system. Taking into account the grid's electricity purchase, electricity purchase expenditure and energy storage cost, determine the initial photovoltaic-storage capacity configuration scheme.

[0014] S4. Based on communication forward-looking prediction and the thermal inertia buffering effect of buildings, the initial optical storage capacity configuration scheme is dynamically modified to buffer the air conditioning load fluctuation through the ability of wall heat storage to delay indoor temperature changes, and to determine the optimal optical storage capacity configuration scheme for smart base stations.

[0015] Furthermore, acquiring historical load data and photovoltaic output data of smart base stations, analyzing the power consumption characteristics of high-power equipment, and constructing the relationship between temperature and the load power of high-power equipment, and visualizing the changes in the power load of high-power equipment includes the following steps:

[0016] S11. Obtain historical load data, photovoltaic output data, and temperature and humidity data of the smart base station, and standardize the obtained data.

[0017] S12. Use wavelet transform to decompose the load of high-power equipment into trend, periodic and random terms, and identify the proportion of each term; draw a heat map of load peak and valley periods, and mark the critical temperature point of load surge of high-power equipment under high temperature weather.

[0018] S13. A density-based clustering algorithm is used to cluster load curves, identify typical scenarios, and generate power feature vectors for each scenario to analyze the start-up and shutdown patterns of high-power equipment; a first-order dynamic model of load and temperature for high-power equipment is established based on heat transfer.

[0019] S14. Draw an interactive load-temperature curve to visually display the coupling periods of high temperature, high load and insufficient photovoltaic power; divide the intervals according to temperature, statistically analyze the power distribution of each interval, draw a kernel density estimation map, and mark the temperature critical point corresponding to the peak power.

[0020] Furthermore, wavelet transform is used to decompose the load of high-power equipment into trend, periodic, and random components, and the proportion of each component is identified through the following steps:

[0021] Wavelet transform is used to perform wavelet multi-resolution decomposition on the load signal. Through several levels of decomposition, approximation coefficients and detail coefficients at each level are obtained. Among them, the low-frequency approximation coefficients correspond to the load trend term, the mid-to-high frequency detail coefficients correspond to the periodic term, and the high-frequency detail coefficients correspond to the random term, thereby realizing the multi-component separation of the load signal.

[0022] The long-term trend term is reconstructed using low-frequency approximation coefficients, the periodic load component is reconstructed by merging mid-to-high-frequency detail coefficients, the random term of sudden fluctuations is reconstructed by merging high-frequency detail coefficients, and the proportion of each component in the total energy is calculated by the sum of squares of the energy of each component, thus quantifying the composition of each load.

[0023] Furthermore, the expression for the first-order dynamic model of load and temperature for high-power equipment is as follows:

[0024]

[0025] In the formula, P equip (t) represents the total power load of high-power devices in the base station at time t, and n represents the number of high-power devices. K represents the temperature weight of the i-th type of high-power device. i T represents the temperature sensitivity coefficient of the i-th type of high-power device. in (t) represents the actual indoor temperature of the base station at time t, T set,i P represents the temperature setpoint for the i-th type of high-power equipment. base,i P represents the base power consumption of the i-th type of high-power device. non-T(t) This represents the stationary load in a high-power device at time t that is completely independent of temperature.

[0026] Furthermore, the expression for the objective function in the optimization model of the photovoltaic-storage grid-connected system is:

[0027]

[0028] In the formula, F represents net expenditure cost, and P g (t) represents the grid interaction power at time t, C buy (t) represents the electricity purchase price at time t, Δt represents the time interval, and P bat (t) represents the rated energy storage power at time t, C PCS E represents the cost of an energy storage power conversion system. bat C represents the rated capacity of energy storage. bat L represents the energy cost of energy storage. f (R1) represents the load characteristic function, sng() represents the sign function, and Pp,sell (t) represents the photovoltaic power output at time t, C p (t) represents the peak-shaving ancillary service electricity price at time t.

[0029] Furthermore, the energy storage charging and discharging strategy is as follows: when the photovoltaic output is excessive, the energy storage is charged first; when the photovoltaic output is insufficient, the energy storage is used to supplement the load gap first.

[0030] The peak-shaving strategy is as follows: the peak-shaving ancillary service period is set to the entire process of photovoltaic power output, with the optimization objectives of peak-shaving ancillary services and net expenditure, so as to maximize the interactive benefits between the photovoltaic-storage system and the grid.

[0031] Furthermore, based on communication forward-looking forecasts and the thermal inertia buffering effect of buildings, the initial optical storage capacity configuration scheme is dynamically adjusted to buffer the air conditioning load fluctuations by delaying indoor temperature changes through wall heat storage. The optimal optical storage capacity configuration scheme for smart base stations is determined by the following steps:

[0032] S41. Predict communication service traffic using the trained LSTM-Transformer hybrid model, determine the load increment by combining the historical average load power of the smart base station, and correct the initial optical storage capacity configuration scheme based on the load increment.

[0033] S42. Construct a physical model of building heat transfer based on the thermal environment data of smart base stations, and determine thermal inertia parameters through parameter inversion to quantify the thermal inertia load buffering capacity. Based on the thermal inertia load buffering capacity, further correct the initial optical storage capacity after load increment correction to buffer the air conditioning load fluctuation through wall heat storage to delay indoor temperature changes, and obtain the optimal optical storage capacity configuration scheme for smart base stations.

[0034] Among them, thermal inertia parameters include building heat capacity, overall heat transfer coefficient, and thermal time constant.

[0035] Furthermore, the trained LSTM-Transformer hybrid model is used to predict communication service traffic, and the load increment is determined by combining the historical average load power of the smart base station. The initial optical storage capacity configuration scheme is then corrected based on the load increment, including the following steps:

[0036] S411. Construct a time series database based on the historical service traffic data of smart base stations, fit the relationship between traffic and power consumption by combining the device power consumption under different traffic conditions, and perform outlier removal, interpolation completion and normalization processing in sequence to generate traffic-power consumption training data.

[0037] S412. Train an LSTM-Transformer hybrid model using traffic-power training data, where the input data are historical service traffic, time features and environmental features, and the output data is the communication service traffic within a preset future time period.

[0038] S413. Use the trained LSTM-Transformer hybrid model to predict the communication service traffic within a preset time period in the future, and calculate the difference between the predicted peak communication service traffic and the historical average for the same period to obtain the load increment; adjust the initial optical storage capacity configuration scheme based on the load increment.

[0039] Furthermore, a physical model of building heat transfer is constructed based on the thermal environment data of the smart base station, and thermal inertia parameters are determined through parameter inversion to quantify the thermal inertia load buffering capacity. Based on the thermal inertia load buffering capacity, the initial optical storage capacity after load increment correction is further adjusted to obtain the optimal optical storage capacity configuration scheme for the smart base station, including the following steps:

[0040] S421. Based on the preprocessed thermal environment data of the smart base station, the smart base station is regarded as a single-node thermal network. A heat balance equation including heat capacity, total heat transfer coefficient, indoor and outdoor temperature and air conditioning power is constructed to obtain the physical model of heat transfer of the building.

[0041] S422. With the goal of minimizing the root mean square error between the measured temperature and the temperature predicted by the heat transfer physical model, the optimal heat capacity and heat transfer coefficient are solved by iterative optimization using a genetic algorithm. The thermal time constant is determined based on the ratio of the optimal heat capacity to the optimal heat transfer coefficient.

[0042] S423. Based on the preset fluctuation range of indoor temperature, calculate the load fluctuation that the building can buffer, and determine the thermal inertia-substitutable energy storage capacity in combination with the duration of high load; subtract the thermal inertia-substitutable energy storage capacity from the initial optical-storage capacity after load increment correction to obtain the optimal optical-storage capacity configuration scheme for the smart base station.

[0043] Furthermore, the revised expression for optical storage capacity is:

[0044]

[0045] The expression for the heat balance equation is:

[0046]

[0047] The expression for the optimal optical storage capacity is:

[0048]

[0049] In the formula, E adj E represents the corrected optical storage capacity. initThe initial optical storage capacity is represented by α, and the flow sensitivity coefficient is ΔP. load P represents the load increment of the predicted communication traffic. avg T represents the historical average load power of the smart base station during the same period, C represents the building's heat capacity, and T represents the average load power during the same period. in T out These represent the indoor and outdoor temperatures, respectively; L represents the overall heat transfer coefficient; and P represents the outdoor temperature. ac E represents the air conditioner's power. final E represents the optimal photovoltaic storage capacity. buffer This indicates the energy storage capacity that can be replaced by thermal inertia. T represents the energy of the building's thermally inertial buffered air conditioning load after unit conversion. peak Indicates the duration of high load. This represents the air conditioning load energy that can be buffered by the building's thermal inertia; ΔT represents the allowable range of indoor temperature fluctuations; and Δt represents the temperature regulation cycle. This represents the thermal time constant.

[0050] The beneficial effects of this invention are as follows:

[0051] 1) This invention constructs a dynamic configuration and collaborative scheduling system for the optical storage capacity of smart base stations by integrating communication traffic prediction, building thermal inertia buffering effect and multi-objective optimization model. This enables precise configuration and efficient operation of the optical storage system from multiple dimensions of "communication-energy-thermal environment", providing systematic technical support for the green, low-carbon and intelligent development of smart base stations.

[0052] 2) By integrating communication traffic prediction with the thermal inertia buffering effect of buildings, this invention can achieve dynamic correction of photovoltaic storage capacity, significantly reduce the dependence of base stations on the power grid, and adaptively adjust the energy storage configuration according to load changes in communication service fluctuation scenarios, thereby reducing the demand for high-priced power grid purchases from the source and effectively reducing electricity costs.

[0053] 3) Based on the coordinated control of the multi-objective optimization model and the heat transfer physical model, the photovoltaic storage system has significantly improved the local consumption capacity of photovoltaic power. By linking communication load prediction, thermal inertia buffering and energy storage scheduling strategies, the excess electricity during peak photovoltaic output periods can be fully utilized, reducing curtailment and promoting the development of base station energy supply towards a green and low-carbon direction.

[0054] 4) The optimized energy storage charging and discharging strategy of this invention, combined with the thermal inertia load buffering mechanism, can control the operating state of the energy storage system within the high-efficiency range, avoid losses caused by overcharging and over-discharging, significantly extend the service life of the energy storage equipment, and generate additional revenue by participating in grid peak shaving ancillary services, thereby improving the overall economic benefits of the photovoltaic-storage system.

[0055] 5) This invention quantifies the building’s thermal inertia’s ability to buffer load fluctuations and constructs a dynamic response optical storage capacity configuration model. This enables the building to delay indoor temperature changes through wall heat storage, thereby buffering the air conditioning load fluctuations. This allows the base station energy system to adapt to complex scenarios such as changes in communication services and seasonal changes, and to maintain stable power supply in extreme environments such as high temperature and high humidity, significantly enhancing the reliability and environmental adaptability of base station operation.

[0056] 6) Through a multi-dimensional dynamic optimization and power grid interaction mechanism, the photovoltaic-storage system can participate in peak shaving auxiliary services according to the power grid load status, effectively smoothing the peak-valley fluctuations of base station load, reducing the impact on the distribution network, providing support for power grid stability, and realizing the coordinated optimization of communication infrastructure and power system. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart of a method for optimizing the optical storage capacity configuration of a smart base station according to an embodiment of the present invention;

[0059] Figure 2 This is an interface diagram of the base station energy management system in an embodiment of the present invention;

[0060] Figure 3 This is an interface diagram of the optical storage capacity configuration in an embodiment of the present invention;

[0061] Figure 4 This is an interface diagram of the template editing method in an embodiment of the present invention. Detailed Implementation

[0062] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0063] According to an embodiment of the present invention, a method for optimizing the optical storage capacity configuration of a smart base station is provided.

[0064] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figures 1-4As shown, the method for optimizing the optical storage capacity configuration of a smart base station according to an embodiment of the present invention includes the following steps:

[0065] S1. Obtain historical load data and photovoltaic output data of smart base stations, analyze the power consumption characteristics of high-power equipment, and construct the relationship between temperature and load power of high-power equipment to visualize the changes in power load of high-power equipment.

[0066] The process of acquiring historical load data and photovoltaic output data of smart base stations, analyzing the power consumption characteristics of high-power equipment, constructing the relationship between temperature and the load power of high-power equipment, and visualizing the changes in the power load of high-power equipment includes the following steps:

[0067] S11. Acquire historical load data, photovoltaic output data, and temperature and humidity data of the smart base station, and standardize the acquired data; specifically including:

[0068] like Figure 2 The diagram shows the base station energy management system. First, minute-level load data for the past 1-3 years is extracted from the base station energy management system (EMS), with a focus on highlighting the sub-metering data of high-power equipment such as air conditioners and integrated photovoltaic storage units. Photovoltaic output data (including irradiance, panel temperature, and DC / AC power) for the synchronous phase is obtained through the photovoltaic inverter monitoring system. Temperature and humidity data (accuracy ±0.5℃) of the base station equipment room and outdoor environment are collected, and distributed sensors (such as DS18B20) are deployed or connected to the environmental monitoring system. Second, a unified timestamp format is established, and missing data is completed using cubic spline interpolation (time periods with a missing rate > 5% are marked as invalid). A data dictionary is created: {timestamp, air conditioner power, other equipment power, photovoltaic AC power, indoor temperature, outdoor temperature}, stored in Parquet format for subsequent processing.

[0069] S12. Use wavelet transform to decompose the load of high-power equipment into trend term (long-term operating baseline), periodic term (daily / weekly fluctuation) and random term (sudden start-up and shutdown), and identify the proportion of each term; draw a heat map of load peak and valley periods, and mark the critical temperature point of load surge of high-power equipment under high temperature weather (e.g., when the indoor temperature is > 26℃, the load increases by 80-120W for every 1℃ increase).

[0070] Specifically, the load of high-power equipment is decomposed into trend, periodic, and random components using wavelet transform, and the proportion of each component is identified through the following steps:

[0071] The minute-level load data of high-power equipment is normalized to the [0,1] interval, and outliers (such as points exceeding 3 times the standard deviation) are removed. Linear interpolation is used to complete the missing data, and a 5-point moving average filter is applied for preliminary noise reduction.

[0072] Prioritize the use of db series wavelets with good compact support (such as db4 and db6), whose orthogonality and finite support characteristics are suitable for capturing abrupt changes in load signals; compare the decomposition effects of different wavelet bases (such as the continuity of the periodic term after db4 decomposition being better than that of Haar wavelets), with the principle of minimizing reconstruction error.

[0073] The number of decomposition layers is set according to the main periodic components of the load signal. The number of decomposition layers for daily periodic load is set to 3 (decomposition to the hourly scale), and the number of decomposition layers for weekly periodic load is set to 4 (decomposition to the daily scale).

[0074] Wavelet transform is used to perform wavelet multi-resolution decomposition on the load signal. Through several levels of decomposition, approximation coefficients and detail coefficients at each level are obtained. Among them, the low-frequency approximation coefficients correspond to the load trend term, the mid-to-high frequency detail coefficients correspond to the periodic term, and the high-frequency detail coefficients correspond to the random term, thereby realizing the multi-component separation of the load signal.

[0075] The long-term trend term is reconstructed using low-frequency approximation coefficients, the periodic load component is reconstructed by merging mid-to-high-frequency detail coefficients, the random term of sudden fluctuations is reconstructed by merging high-frequency detail coefficients, and the proportion of each component in the total energy is calculated by the sum of squares of the energy of each component, thus quantifying the composition of each load.

[0076] S13. Density-based clustering algorithms are used to cluster load curves, identify typical scenarios (such as "high temperature and high load" and "cloudy and rainy with low photovoltaic power"), and generate power feature vectors (mean, variance, peak-to-valley difference) for each scenario. The start-up and shutdown patterns of high-power equipment are analyzed (such as the correlation between the start-up and shutdown interval of air conditioning compressors and outdoor temperature). A first-order dynamic model of load and temperature of high-power equipment is established based on heat transfer.

[0077] Specifically, the expression for the first-order dynamic model of load and temperature for high-power equipment is as follows:

[0078]

[0079] In the formula, P equip (t) represents the total power load of high-power devices in the base station at time t, and n represents the number of high-power devices. K represents the temperature weight of the i-th type of high-power equipment. In this embodiment, the temperature weight of the air conditioning equipment ranges from 0.6 to 0.8, the temperature weight of the server / optical storage integrated machine ranges from 0.1 to 0.3, and the value of the equipment unrelated to temperature is 0. i This represents the temperature sensitivity coefficient of the i-th type of high-power device. In this embodiment, the temperature sensitivity coefficient of the air conditioner is 100-200W / ℃, and the temperature sensitivity coefficient of the server / optical storage integrated machine is 20-50W / ℃. in (t) represents the actual indoor temperature of the base station at time t, T set,iP represents the temperature setpoint for the i-th type of high-power equipment. base,i P represents the base power consumption of the i-th type of high-power device. non-T(t) This represents the fixed load in a high-power device at time t that is completely independent of temperature (such as the computing power of a server or the transmission power of a radio frequency signal).

[0080] S14. Draw an interactive load-temperature curve to visually display the coupling periods of high temperature, high load and insufficient photovoltaic power; divide the intervals according to temperature, statistically analyze the power distribution of each interval, draw a kernel density estimation map, and mark the temperature critical point corresponding to the peak power.

[0081] The above steps involve collecting and standardizing historical load, photovoltaic output, and temperature and humidity data from smart base stations. Combined with wavelet transform to decompose load components and clustering algorithms to identify typical scenarios, a dynamic relationship model between temperature and high-power equipment load is constructed, enabling visualization of load changes. This step accurately reveals the power consumption patterns, temperature sensitivity, and load fluctuation components (trend, periodic, and random components) of high-power equipment (such as air conditioners). It provides a foundation for load characteristic analysis based on actual operating data for photovoltaic-storage capacity configuration, enabling subsequent optimization to specifically address temperature-driven load fluctuations and improve the matching degree of the photovoltaic-storage system to actual load demands.

[0082] S2. Taking the minimum net expenditure as the objective function, the system incorporates electricity purchase costs, energy storage investment costs, energy storage loss costs, and revenue from participating in grid peak shaving auxiliary services, and constructs an operation optimization model for the photovoltaic-storage grid-connected system in combination with constraints.

[0083] Specifically, the expression for the objective function in the optimization model of the photovoltaic-storage grid-connected system is as follows:

[0084]

[0085] In the formula, F represents net expenditure cost, and P g (t) represents the grid interaction power at time t, C buy (t) represents the electricity purchase price at time t, Δt represents the time interval, and P bat (t) represents the rated energy storage power at time t, C PCS E represents the cost of an energy storage power conversion system. bat C represents the rated capacity of energy storage. bat L represents the energy cost of energy storage. f (R1) represents the load characteristic function (e.g., temperature-load relationship, where R1 is a key parameter, such as temperature), sng() represents the sign function, and P p,sell (t) represents the photovoltaic power output at time t, C p (t) represents the peak-shaving ancillary service electricity price at time t;

[0086] The first term in the objective function is the electricity purchase cost, used to calculate the daily electricity purchase cost; the second term is the power and energy cost of energy storage, which is the allocation of the investment cost of energy storage equipment; the third term is the energy storage loss cost, where a sign function is used to distinguish between charging and discharging losses, with sng being -1 during charging and 1 during discharging, to calculate the losses under different states; the fourth term is the revenue from participating in grid peak shaving ancillary services.

[0087] The constraints are:

[0088]

[0089] In the formula, P l (t) represents the base station load power at time t, P v (t) represents the photovoltaic output power at time t, P cha (t), P dist (t) represent the energy storage charging and discharging power, respectively; minSoC and maxSoC represent the minimum and maximum values ​​of the energy storage state of charge, respectively; SoC(t) represents the energy storage state of charge at time t; P cs P represents the rated power of the energy storage, λ1 and λ2 represent the energy storage charge and discharge coefficients, respectively. buy β1 and β2 represent the grid's rated power purchase capacity, respectively, and the grid's power purchase and sales coefficients.

[0090] The above steps construct an optimization model for the photovoltaic-storage grid-connected system with the goal of minimizing net expenditure. This model incorporates multiple factors, including electricity purchase costs, energy storage losses, and peak-shaving benefits, and ensures compliance with actual operating rules through constraints such as power balance and energy storage state of charge. This model achieves coordinated optimization of the "source-load-storage-grid" system, quantifying the balance between economic costs and grid interaction benefits while ensuring power supply reliability. It provides a mathematical optimization framework that is both economical and feasible for photovoltaic-storage capacity configuration, enabling subsequent solutions to obtain the optimal solution that balances cost control and grid service.

[0091] S3. Define the energy storage charging / discharging and peak-shaving strategies. Based on optimization algorithms, solve the operation optimization model of the photovoltaic-storage grid-connected system. Taking into account grid power purchases, power purchase expenditures, and energy storage costs, determine the initial photovoltaic-storage capacity configuration scheme. Specifically, this includes:

[0092] First, determine the energy storage charging and discharging and peak shaving strategies. The energy storage charging and discharging strategy is as follows: when the photovoltaic output is excessive, the energy storage is charged first; when the photovoltaic output is insufficient, the energy storage is used to supplement the load gap. The peak shaving strategy is as follows: the peak shaving ancillary service period is set to the entire photovoltaic output process, with the peak shaving ancillary service and net expenditure as the optimization objectives, to maximize the interaction benefits between the photovoltaic-storage system and the grid.

[0093] Secondly, Particle Swarm Optimization (PSO) or Mixed Integer Linear Programming (MILP) is used to initialize parameters (such as energy storage cost of 0.08 yuan / W and 1.3 yuan / Wh, initial SoC value of 0.3, and rated power of 4 kW), and iterative solutions are performed to optimize energy storage power, energy, and grid power, satisfying constraints and minimizing net expenditure costs.

[0094] Finally, the cost-benefit curves were plotted by comparing the electricity purchase expenditure, energy storage investment, and peak shaving benefits of different schemes. The SoC fluctuation (ensuring 0.1-0.9) and power supply capacity (meeting load demand) of energy storage under extreme scenarios (such as continuous rain) were verified. The absorption rate (≥95%) was calculated to ensure full utilization of photovoltaic power and finally determine the initial photovoltaic and energy storage capacity configuration scheme.

[0095] The above steps, based on an optimization algorithm, solve the photovoltaic-storage grid-connected model, clarifying the charging and discharging logic of "storage charging when photovoltaic power is surplus and discharging when photovoltaic power is insufficient." Combined with a full-time peak-shaving strategy, the initial photovoltaic-storage capacity is determined by comprehensively evaluating factors such as grid power purchase and energy storage costs. This step, through strategy optimization and algorithm solving, achieves efficient utilization of energy storage resources, maximizing photovoltaic absorption rate and reducing power purchase costs while meeting base station load demands. It forms an initial configuration scheme that balances basic power supply and peak-shaving services, providing an iteratively optimizeable baseline capacity for subsequent dynamic adjustments.

[0096] S4. Based on communication forward-looking prediction and the thermal inertia buffering effect of buildings, the initial optical storage capacity configuration scheme is dynamically modified to buffer the air conditioning load fluctuation through the ability of wall heat storage to delay indoor temperature changes, and to determine the optimal optical storage capacity configuration scheme for smart base stations.

[0097] The optimal configuration scheme for smart base stations, determined by dynamically adjusting the initial photovoltaic and energy storage capacity based on communication forecasting and the thermal inertia buffering effect of buildings, aims to mitigate air conditioning load fluctuations by using wall heat storage to delay indoor temperature changes. This process includes the following steps:

[0098] S41. Predict communication service traffic using the trained LSTM-Transformer hybrid model, determine the load increment by combining the historical average load power of the smart base station, and correct the initial optical storage capacity configuration scheme based on the load increment.

[0099] Specifically, the process involves using a trained LSTM-Transformer hybrid model to predict communication traffic, combining this with the historical average load power of the smart base station to determine the load increment, and then revising the initial optical storage capacity configuration scheme based on the load increment, including the following steps:

[0100] S411. Construct a time-series database based on historical service traffic data of smart base stations, fit the relationship between traffic and power consumption by combining device power consumption under different traffic conditions, and sequentially perform outlier removal, interpolation completion, and normalization processing to generate traffic-power consumption training data; specifically including:

[0101] First, historical traffic (number of users, throughput, etc., 15-minute granularity) is extracted from the base station communication system, covering all scenarios (seasons, holidays), to build a time series library. Second, device power consumption (e.g., RRU, BBU) under different traffic conditions is tested in the laboratory to fit the traffic-power consumption relationship, distinguishing between strongly / weakly correlated devices (RRU, BBU are strongly correlated, weight 0.7-0.8; power supplies, etc. are weakly correlated, weight 0.2-0.3). Finally, outliers are removed, interpolation is performed, and normalization is applied to generate traffic-power consumption training data, which must cover at least three complete communication cycles (e.g., quarters).

[0102] S412. Train an LSTM-Transformer hybrid model using traffic-power training data, where the input data are historical service traffic, time features and environmental features, and the output data is the communication service traffic within a preset future time period.

[0103] In the LSTM-Transformer hybrid model, LSTM is used to capture daily / weekly cycles (long-term dependencies), while Transformer is used to handle short-term mutations (such as surges in activity flow and self-attention-enhanced detail capture).

[0104] S413. Use the trained LSTM-Transformer hybrid model to predict the communication service traffic within a preset time period in the future, calculate the difference between the predicted peak communication service traffic and the historical average for the same period, and convert the difference into load increment; adjust the initial optical storage capacity configuration scheme based on the load increment.

[0105] The revised expression for optical storage capacity is:

[0106]

[0107] In the formula, E adj E represents the corrected optical storage capacity. init The initial optical storage capacity is represented by α, and the flow sensitivity coefficient is represented by α. In this embodiment, the flow sensitivity coefficient of strongly correlated devices (such as air conditioners) is 0.6-0.8, and the flow sensitivity coefficient of weakly correlated devices (such as server computing units) is 0.3-0.5. ΔP load P represents the load increment of the predicted communication traffic. avg This represents the historical average load power of the smart base station during the same period.

[0108] S42. Construct a physical model of building heat transfer based on the thermal environment data of the smart base station, and determine the thermal inertia parameters (including building heat capacity, total heat transfer coefficient and thermal time constant) through parameter inversion to quantify the thermal inertia load buffering capacity; based on the thermal inertia load buffering capacity, further correct the initial optical storage capacity after load increment correction to buffer the air conditioning load fluctuation through wall heat storage to delay indoor temperature changes, and obtain the optimal optical storage capacity configuration scheme for the smart base station.

[0109] Specifically, a physical model of building heat transfer is constructed based on the thermal environment data of the smart base station, and thermal inertia parameters are determined through parameter inversion to quantify the thermal inertia load buffering capacity. Based on the thermal inertia load buffering capacity, the initial optical storage capacity after load increment correction is further adjusted to buffer the air conditioning load fluctuation through wall heat storage to delay indoor temperature changes. The optimal optical storage capacity configuration scheme for the smart base station includes the following steps:

[0110] S421. Based on the preprocessed thermal environment data of the smart base station, the smart base station is regarded as a single-node thermal network. A heat balance equation including heat capacity, total heat transfer coefficient, indoor and outdoor temperature and air conditioning power is constructed to obtain the physical model of heat transfer of the building.

[0111] The collection of thermal environment data includes: collecting indoor and outdoor temperatures (every 15 minutes) using temperature sensors (accuracy ±0.5℃), covering typical seasons throughout the year (summer / winter / transitional seasons); synchronously recording the real-time power of the air conditioner (10W resolution) using smart meters and marking the air conditioner's on / off status; collecting base station building structure parameters (wall materials, thickness, window area, etc.) for initial thermal model building; removing abnormal data (such as temperature jumps caused by sensor failures) and using linear interpolation to complete missing values; and extracting temperature-power data for "air conditioner on periods" according to air conditioner operating status (to avoid interference with thermal inertia analysis during off periods).

[0112] The expression for the heat balance equation is:

[0113]

[0114] In the formula, C represents the building's heat capacity, and T... in T out These represent the indoor and outdoor temperatures, respectively; L represents the overall heat transfer coefficient; and P represents the outdoor temperature. ac Indicates the air conditioner's power;

[0115] S422. With the goal of minimizing the root mean square error between the measured temperature and the temperature predicted by the heat transfer physical model, the optimal heat capacity and heat transfer coefficient are solved by iterative optimization using a genetic algorithm. The thermal time constant is determined based on the ratio of the optimal heat capacity to the optimal heat transfer coefficient.

[0116] Specifically, the optimal heat capacity and heat transfer coefficient are solved iteratively using a genetic algorithm. The thermal time constant is determined based on the ratio of the optimal heat capacity to the optimal heat transfer coefficient, including:

[0117] First, the search range for heat capacity and heat transfer coefficient is set: 500–2000 kJ / K for heat capacity and 200–800 W / K for heat transfer coefficient. Real-number encoding is used to encode these two parameters using genes, generating an initial population (e.g., 100 random parameter combinations). Each gene set corresponds to a candidate value for heat capacity and heat transfer coefficient. Then, for each parameter set in the initial population, the predicted temperature is calculated using a heat transfer model. The predicted temperature is compared with the measured temperature to calculate the RMSE (Recovery Time Sequence). The reciprocal of the RMSE is used as the fitness function value; higher fitness indicates a better parameter combination. Finally, the population is evaluated based on the fitness value. A selection operation (such as roulette wheel selection) is performed to retain high-fitness parameter combinations and eliminate low-fitness combinations. Crossover (such as arithmetic crossover) and mutation (such as Gaussian mutation) operations are then performed on the selected parameters to generate a new generation of population. The first two steps are then repeated iteratively to optimize the population parameters until the optimal fitness improvement of the population for 10 consecutive generations is less than 1% or a preset number of iterations (such as 200 generations) is reached. The iteration is then terminated, and the heat capacity and heat transfer coefficient corresponding to the optimal fitness are extracted. Finally, the optimal heat capacity and heat transfer coefficient are substituted into a formula to calculate the thermal time constant, which is equal to the ratio of heat capacity to heat transfer coefficient. This constant is used to quantify the building's thermal inertia's buffering capacity against temperature changes.

[0118] S423. Based on the preset fluctuation range of indoor temperature, calculate the load fluctuation that the building can buffer, and determine the thermally inertial replaceable energy storage capacity in conjunction with the duration of high load. Subtract the thermally inertial replaceable energy storage capacity from the initial optical-storage capacity after load increment correction to obtain the optimal optical-storage capacity configuration scheme for the smart base station. Configure the optical-storage capacity according to the determined optimal optical-storage capacity configuration scheme for the smart base station in conjunction with the preset editing template. The specific configuration interface is as follows: Figure 3 As shown, the editing template is as follows: Figure 4 As shown.

[0119] The expression for the optimal optical storage capacity is:

[0120]

[0121] In the formula, E final E represents the optimal photovoltaic storage capacity. buffer This indicates the energy storage capacity that can be replaced by thermal inertia. T represents the energy of the building's thermally inertial buffered air conditioning load after unit conversion. peak Indicates the duration of high load. This represents the air conditioning load energy that can be buffered by the building's thermal inertia; ΔT represents the allowable range of indoor temperature fluctuations; and Δt represents the temperature regulation cycle. This represents the thermal time constant.

[0122] The above steps integrate communication traffic prediction with the building's thermal inertia buffering effect. By using an LSTM-Transformer model to anticipate load increments and combining this with thermal inertia parameters to quantify the air conditioning load buffering capacity, the initial photovoltaic and energy storage capacity is dynamically adjusted. This step overcomes the limitations of traditional static configurations, enabling the photovoltaic and energy storage system to adapt to fluctuations in communication services and changes in the thermal environment. While reducing electricity costs, it improves photovoltaic absorption rate and grid peak-shaving benefits, forming an optimal configuration scheme that balances communication reliability, energy efficiency, and economy. This significantly enhances the environmental adaptability and overall benefits of smart base station photovoltaic and energy storage systems.

[0123] In summary, by utilizing the above-mentioned technical solutions of this invention, this invention constructs a dynamic configuration and collaborative scheduling system for the optical storage capacity of smart base stations by integrating communication traffic prediction, building thermal inertia buffering effect, and multi-objective optimization model. This enables precise configuration and efficient operation of the optical storage system from multiple dimensions of "communication-energy-thermal environment," providing systematic technical support for the green, low-carbon, and intelligent development of smart base stations.

[0124] Furthermore, by integrating communication traffic prediction with the thermal inertia buffering effect of buildings, this invention can achieve dynamic correction of photovoltaic storage capacity, significantly reducing the dependence of base stations on the power grid. In scenarios with fluctuating communication services, it can adaptively adjust energy storage configuration according to load changes, thereby reducing the need for high-priced power purchases from the grid and effectively lowering electricity costs.

[0125] Furthermore, based on the coordinated control of a multi-objective optimization model and a heat transfer physics model, this invention significantly improves the on-site consumption capacity of photovoltaic power by the photovoltaic-storage system. By linking communication load forecasting, thermal inertia buffering, and energy storage scheduling strategies, the excess electricity during peak photovoltaic output periods can be fully utilized, reducing curtailment and promoting the green and low-carbon development of base station energy supply.

[0126] Furthermore, the optimized energy storage charging and discharging strategy of this invention, combined with a thermal inertia load buffering mechanism, can control the operating state of the energy storage system within the high-efficiency range, avoid losses caused by overcharging and over-discharging, significantly extend the service life of the energy storage equipment, and generate additional revenue by participating in grid peak shaving ancillary services, thereby improving the overall economic benefits of the photovoltaic-energy storage system.

[0127] Furthermore, by quantifying the building's thermal inertia to buffer load fluctuations, this invention constructs a dynamic response optical storage capacity configuration model, enabling the building to delay indoor temperature changes through wall heat storage, thereby buffering air conditioning load fluctuations. This allows the base station energy system to adapt to complex scenarios such as changes in communication services and seasonal changes, and to maintain stable power supply even in extreme environments such as high temperature and high humidity, significantly enhancing the reliability and environmental adaptability of base station operation.

[0128] Furthermore, through a multi-dimensional dynamic optimization and grid interaction mechanism, the photovoltaic-storage system can participate in peak shaving auxiliary services according to the grid load status, effectively smoothing the peak-valley fluctuations of base station load, reducing the impact on the distribution network, providing support for grid stability, and realizing the coordinated optimization of communication infrastructure and power system.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing the optical storage capacity configuration of a smart base station, characterized in that, Includes the following steps: S1. Obtain historical load data and photovoltaic output data of smart base stations, analyze the power consumption characteristics of high-power equipment, and construct the relationship between temperature and load power of high-power equipment to visualize the changes in power load of high-power equipment. S2. Taking the minimum net expenditure as the objective function, the system incorporates electricity purchase costs, energy storage investment costs, energy storage loss costs, and revenue from participating in grid peak shaving auxiliary services, and constructs an operation optimization model for the photovoltaic-storage grid-connected system in combination with constraints. The objective function in the optimization model for the operation of the photovoltaic-storage grid-connected system is expressed as follows: ； In the formula, F represents net expenditure cost, and P g (t) represents the grid interaction power at time t, C buy (t) represents the electricity purchase price at time t, Δt represents the time interval, and P bat (t) represents the rated energy storage power at time t, C PCS E represents the cost of an energy storage power conversion system. bat C represents the rated capacity of energy storage. bat L represents the energy cost of energy storage. f (R1) represents the load characteristic function, sng() represents the sign function, and P p,sell (t) represents the photovoltaic power output at time t, C p (t) represents the peak-shaving ancillary service electricity price at time t; S3. Clarify the energy storage charging and discharging and peak shaving strategies. Based on the optimization algorithm, solve the operation optimization model of the photovoltaic-storage grid-connected system. Taking into account the grid's electricity purchase, electricity purchase expenditure and energy storage cost, determine the initial photovoltaic-storage capacity configuration scheme. S4. Based on communication forward-looking prediction and the thermal inertia buffering effect of buildings, the initial optical storage capacity configuration scheme is dynamically modified to buffer the air conditioning load fluctuation through the ability of wall heat storage to delay indoor temperature changes, and to determine the optimal optical storage capacity configuration scheme for smart base stations.

2. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 1, characterized in that, The process of acquiring historical load data and photovoltaic output data of smart base stations, analyzing the power consumption characteristics of high-power equipment, constructing the relationship between temperature and the load power of high-power equipment, and visualizing the changes in the power load of high-power equipment includes the following steps: S11. Obtain historical load data, photovoltaic output data, and temperature and humidity data of the smart base station, and standardize the obtained data. S12. Use wavelet transform to decompose the load of high-power equipment into trend, periodic and random terms, and identify the proportion of each term; draw a heat map of load peak and valley periods, and mark the critical temperature point of load surge of high-power equipment under high temperature weather. S13. A density-based clustering algorithm is used to cluster load curves, identify typical scenarios, and generate power feature vectors for each scenario to analyze the start-up and shutdown patterns of high-power equipment; a first-order dynamic model of load and temperature for high-power equipment is established based on heat transfer. S14. Draw an interactive load-temperature curve to visually display the coupling periods of high temperature, high load and insufficient photovoltaic power; divide the intervals according to temperature, statistically analyze the power distribution of each interval, draw a kernel density estimation map, and mark the temperature critical point corresponding to the peak power.

3. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 2, characterized in that, The process of using wavelet transform to decompose the load of high-power equipment into trend, periodic, and random components, and identifying the proportion of each component, includes the following steps: Wavelet transform is used to perform wavelet multi-resolution decomposition on the load signal. Through several levels of decomposition, approximation coefficients and detail coefficients at each level are obtained. Among them, the low-frequency approximation coefficients correspond to the load trend term, the mid-to-high frequency detail coefficients correspond to the periodic term, and the high-frequency detail coefficients correspond to the random term, thereby realizing the multi-component separation of the load signal. The long-term trend term is reconstructed using low-frequency approximation coefficients, the periodic load component is reconstructed by merging mid-to-high-frequency detail coefficients, the random term of sudden fluctuations is reconstructed by merging high-frequency detail coefficients, and the proportion of each component in the total energy is calculated by the sum of squares of the energy of each component, thus quantifying the composition of each load.

4. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 2, characterized in that, The expression for the first-order dynamic model of the load and temperature of the high-power equipment is as follows: ； In the formula, P equip (t) represents the total power load of high-power devices in the base station at time t, and n represents the number of high-power devices. K represents the temperature weight of the i-th type of high-power device. i T represents the temperature sensitivity coefficient of the i-th type of high-power device. in (t) represents the actual indoor temperature of the base station at time t, T set,i P represents the temperature setpoint for the i-th type of high-power equipment. base,i P represents the base power consumption of the i-th type of high-power device. non-T(t) This represents the stationary load in a high-power device at time t that is completely independent of temperature.

5. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 1, characterized in that, The energy storage charging and discharging strategy is as follows: when the photovoltaic output is excessive, the energy storage is charged first; when the photovoltaic output is insufficient, the energy storage is used to supplement the load gap first. The peak-shaving strategy is as follows: the peak-shaving ancillary service period is set to the entire process of photovoltaic power output, with the optimization objectives of peak-shaving ancillary services and net expenditure, so as to maximize the interactive benefits between the photovoltaic-storage system and the grid.

6. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 1, characterized in that, The method of dynamically adjusting the initial optical storage capacity configuration scheme based on communication forward-looking prediction and the thermal inertia buffering effect of buildings, in order to buffer the air conditioning load fluctuation through the ability of wall heat storage to delay indoor temperature changes, and determining the optimal optical storage capacity configuration scheme for smart base stations includes the following steps: S41. Predict communication service traffic using the trained LSTM-Transformer hybrid model, determine the load increment by combining the historical average load power of the smart base station, and correct the initial optical storage capacity configuration scheme based on the load increment. S42. Construct a physical model of building heat transfer based on the thermal environment data of smart base stations, and determine thermal inertia parameters through parameter inversion to quantify the thermal inertia load buffering capacity. Based on the thermal inertia load buffering capacity, further correct the initial optical storage capacity after load increment correction to buffer the air conditioning load fluctuation through wall heat storage to delay indoor temperature changes, and obtain the optimal optical storage capacity configuration scheme for smart base stations. Among them, thermal inertia parameters include building heat capacity, overall heat transfer coefficient, and thermal time constant.

7. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 6, characterized in that, The process of using a trained LSTM-Transformer hybrid model to predict communication service traffic, combining this with the historical average load power of the smart base station to determine the load increment, and then revising the initial optical storage capacity configuration scheme based on the load increment includes the following steps: S411. Construct a time series database based on the historical service traffic data of smart base stations, fit the relationship between traffic and power consumption by combining the device power consumption under different traffic conditions, and perform outlier removal, interpolation completion and normalization processing in sequence to generate traffic-power consumption training data. S412. Train an LSTM-Transformer hybrid model using traffic-power training data, where the input data are historical service traffic, time features and environmental features, and the output data is the communication service traffic within a preset future time period. S413. Use the trained LSTM-Transformer hybrid model to predict the communication service traffic within a preset time period in the future, and calculate the difference between the predicted peak communication service traffic and the historical average for the same period to obtain the load increment; adjust the initial optical storage capacity configuration scheme based on the load increment.

8. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 7, characterized in that, The building heat transfer physical model is constructed based on the thermal environment data of the smart base station, and the thermal inertia parameters are determined through parameter inversion to quantify the thermal inertia load buffering capacity. The initial optical storage capacity, after load increment correction, is further adjusted based on the thermal inertia load buffering capacity to obtain the optimal optical storage capacity configuration scheme for smart base stations, including the following steps: S421. Based on the preprocessed thermal environment data of the smart base station, the smart base station is regarded as a single-node thermal network. A heat balance equation including heat capacity, total heat transfer coefficient, indoor and outdoor temperature and air conditioning power is constructed to obtain the physical model of heat transfer of the building. S422. With the goal of minimizing the root mean square error between the measured temperature and the temperature predicted by the heat transfer physical model, the optimal heat capacity and heat transfer coefficient are solved by iterative optimization using a genetic algorithm. The thermal time constant is determined based on the ratio of the optimal heat capacity to the optimal heat transfer coefficient. S423. Based on the preset fluctuation range of indoor temperature, calculate the load fluctuation that the building can buffer, and determine the thermal inertia-substitutable energy storage capacity in combination with the duration of high load; subtract the thermal inertia-substitutable energy storage capacity from the initial optical-storage capacity after load increment correction to obtain the optimal optical-storage capacity configuration scheme for the smart base station.

9. The method for optimizing the optical storage capacity configuration of a smart base station according to claim 8, characterized in that, The revised expression for optical storage capacity is: ； The expression for the heat balance equation is: ； The expression for the optimal optical storage capacity is: ； In the formula, E adj E represents the corrected optical storage capacity. init The initial optical storage capacity is represented by α, and the flow sensitivity coefficient is ΔP. load P represents the load increment of the predicted communication traffic. avg T represents the historical average load power of the smart base station during the same period, C represents the building's heat capacity, and T represents the average load power during the same period. in T out These represent the indoor and outdoor temperatures, respectively; L represents the overall heat transfer coefficient; and P represents the outdoor temperature. ac E represents the air conditioner's power. final E represents the optimal photovoltaic storage capacity. buffer This indicates the energy storage capacity that can be replaced by thermal inertia. T represents the energy of the building's thermally inertial buffered air conditioning load after unit conversion. peak Indicates the duration of high load. This represents the air conditioning load energy that can be buffered by the building's thermal inertia; ΔT represents the allowable range of indoor temperature fluctuations; and Δt represents the temperature regulation cycle. This represents the thermal time constant.

Images