Energy scheduling method and device, equipment and storage medium

Abstract

The application provides an energy scheduling method and device, equipment and storage medium, wherein the method comprises: acquiring real-time data of a target base station and constructing a multi-dimensional feature data set, then inputting the same into a pre-trained XGBoost model, and predicting initial photovoltaic and load prediction values of multiple time scales in a current time window based on a preset rolling step; then a multi-objective optimization function containing peak clipping and valley filling effect, electricity cost and energy storage life loss is constructed; finally, based on the function and base station reliability rigid constraint, the adaptive weight particle swarm optimization algorithm is used to optimize the prediction result and energy storage power and state information, and the charge and discharge target scheduling instruction is obtained. The application improves the prediction accuracy through XGBoost rolling prediction, combines multi-objective optimization and reliability constraint, and can effectively reduce the electricity cost, stabilize the power grid load and prolong the energy storage life under the premise of ensuring the safety of base station power supply.

Description

Technical Field

[0001] This application relates to the field of energy dispatching technology, and more specifically, to an energy dispatching method, apparatus, equipment, and storage medium. Background Technology

[0002] With the widespread application of distributed photovoltaic (PV) power in communication base stations, how to achieve coordinated scheduling of PV, energy storage, and base station loads has become a research hotspot. Especially under the background of demand-side response for grid peak shaving and valley filling, it is necessary to optimize the electricity cost of base stations and improve energy utilization efficiency through intelligent scheduling methods while ensuring uninterrupted power supply to communication equipment.

[0003] The current mainstream base station peak shaving and valley filling scheduling technologies are mainly divided into three categories: First, fixed-period threshold-based energy storage scheduling schemes based on grid time-of-use pricing, which perform simple charging and discharging according to the electricity price period; second, photovoltaic-energy storage collaborative scheduling schemes based on traditional prediction models, which formulate 24-hour plans through day-ahead forecasts; and third, multi-objective optimization scheduling schemes that use intelligent optimization algorithms such as genetic algorithms to balance operating costs and grid load fluctuations.

[0004] However, existing fixed threshold schemes cannot adapt to the randomness and load fluctuations of photovoltaics and loads, which can easily lead to energy waste or power supply risks. Traditional prediction models have large prediction errors for nonlinear weather scenarios and day-ahead scheduling is lagging, which can easily cause reverse peak shaving. Multi-objective optimization schemes usually have high computational complexity and slow solution, making it difficult to meet the real-time scheduling requirements of low computing power and low latency at the edge of the base station, and often ignore the lifespan loss of energy storage throughout its entire life cycle. Summary of the Invention

[0005] The purpose of this application is to address the shortcomings of the prior art by providing an energy dispatching method, apparatus, equipment, and storage medium to solve the problems of the prior art being unable to adapt to photovoltaic and load fluctuations, dispatching lag, and difficulty in meeting real-time dispatching requirements.

[0006] To achieve the above objectives, the technical solution adopted in this application is as follows: In a first aspect, this application provides an energy dispatching method, the method comprising: Acquire real-time data from the target base station and construct a multi-dimensional feature dataset based on the real-time data, which includes load data, photovoltaic data, meteorological data, and power grid data. The multi-dimensional feature dataset is input into a pre-trained extreme gradient boosting XGBoost model. The XGBoost model predicts initial prediction results for multiple time scales under the current time window based on a preset rolling step size and the multi-dimensional feature dataset. The initial prediction results include: initial photovoltaic prediction value and initial load prediction value. A multi-objective optimization function is constructed, which includes peak shaving and valley filling effect, electricity cost and energy storage life loss. The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect objective, electricity cost objective, and energy storage life loss objective. Based on the multi-objective optimization function and the rigid constraint of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize each of the initial prediction results, the energy storage charging power, the energy storage discharging power and the energy storage status information corresponding to the initial prediction results, the target scheduling instruction is obtained. The target scheduling instruction includes: charging power instruction and discharging power instruction.

[0007] Optionally, the step of optimizing each initial prediction result, the corresponding energy storage charging power, energy storage discharging power, and energy storage state information based on the multi-objective optimization function and the rigid constraints of base station reliability, and using an adaptive weighted particle swarm optimization algorithm to obtain the target scheduling instruction includes: The initial input power is determined based on the initial prediction results, the energy storage charging power corresponding to the initial prediction results, and the energy storage discharging power. Based on the multi-objective optimization function and the rigid constraints of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize the initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information, the target scheduling instruction is obtained.

[0008] Optionally, determining the initial input power based on the initial prediction result, the energy storage charging power corresponding to the initial prediction result, and the energy storage discharging power includes: Using the initial prediction result, the energy storage charging power, and the energy storage discharging power as input parameters, the formula is used... The initial input power is calculated. in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the power input to the power grid at time τ. Let τ be the initial load prediction value of the target base station at time τ. The energy storage charging power at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the energy storage discharge power at time τ.

[0009] Optionally, the mathematical expression of the rigid constraint on base station reliability is:

[0010] in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the maximum power that the power grid can supply at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the maximum usable discharge power of the stored energy. Let τ be the initial load prediction value of the target base station at time τ. This represents the power supply safety margin at time τ.

[0011] Optionally, the step of optimizing the initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information based on the multi-objective optimization function and the rigid constraints of base station reliability, and using an adaptive weighted particle swarm optimization algorithm to obtain the target scheduling instruction, includes: The initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information are particle-coded to obtain multiple initial scheduling schemes; Based on the multi-objective optimization function and the rigid constraint of base station reliability, the adaptive weighted particle swarm optimization algorithm is used to optimize each of the initial scheduling schemes to obtain the target scheduling instruction.

[0012] Optionally, the step of optimizing each of the initial scheduling schemes based on the multi-objective optimization function and the rigid constraints of base station reliability, and obtaining the target scheduling instruction using an adaptive weighted particle swarm optimization algorithm, includes: Determine at least one candidate scheduling scheme from each of the initial scheduling schemes that satisfies the rigid constraint on base station reliability; Using each of the candidate scheduling schemes as input parameters, and employing an adaptive weighted particle swarm optimization algorithm to solve the multi-objective optimization function, the target scheduling instruction is obtained.

[0013] Optionally, the multi-objective optimization function is:

[0014] in, This indicates the target effect of peak shaving and valley filling. Indicates the target electricity cost. Indicates the energy storage lifespan loss target. These are the weighting coefficients for the peak shaving and valley filling effect target, the electricity cost target, and the energy storage life loss target, respectively. The mathematical representation of the peak shaving and valley filling effect target is:

[0015] in, , These are sets of electricity price peak and valley periods, respectively. This represents the upper limit of the power input to the power grid during peak periods. The target power for filling the valley section. This represents the average power input to the power grid within the current time window. The coefficient is the smoothing penalty; max(0,·) indicates that only the portion exceeding the threshold is penalized. Let τ represent the grid input power at time τ; The mathematical representation of the electricity cost target is:

[0016] in, Let τ be the time-of-use electricity price. The scheduling time step; The energy storage lifetime loss target is expressed as:

[0017] in, The replacement cost over the entire lifecycle of the energy storage system, This represents the equivalent number of charge-discharge cycles generated within the current time window. For the battery's rated cycle life, This refers to the rated capacity of the energy storage.

[0018] Secondly, this application provides an energy dispatching device, the device comprising: The acquisition module is used to acquire real-time data of the target base station and construct a multi-dimensional feature dataset based on the real-time data. The real-time data includes: load data, photovoltaic data, meteorological data, and power grid data. The prediction module is used to input the multi-dimensional feature dataset into a pre-trained extreme gradient boosting XGBoost model. The XGBoost model predicts initial prediction results for multiple time scales under the current time window based on a preset rolling step size and the multi-dimensional feature dataset. The initial prediction results include: initial photovoltaic prediction value and initial load prediction value. The construction module is used to construct a multi-objective optimization function that includes peak shaving and valley filling effect, electricity cost and energy storage life loss. The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect objective, electricity cost objective, and energy storage life loss objective. The optimization module is used to optimize each of the initial prediction results, the energy storage charging power, the energy storage discharging power and the energy storage status information corresponding to the initial prediction results, based on the multi-objective optimization function and the rigid constraints of base station reliability, and using an adaptive weighted particle swarm optimization algorithm to obtain the target scheduling instruction, which includes a charging power instruction and a discharging power instruction.

[0019] Optionally, the optimization module is specifically used for: The initial input power is determined based on the initial prediction results, the energy storage charging power corresponding to the initial prediction results, and the energy storage discharging power. Based on the multi-objective optimization function and the rigid constraints of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize the initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information, the target scheduling instruction is obtained.

[0020] Optionally, the optimization module is specifically used for: Using the initial prediction result, the energy storage charging power, and the energy storage discharging power as input parameters, the formula is used... The initial input power is calculated. in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the power input to the power grid at time τ. Let τ be the initial load prediction value of the target base station at time τ. The energy storage charging power at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the energy storage discharge power at time τ.

[0021] Optionally, the mathematical expression of the rigid constraint on base station reliability is:

[0022] in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the maximum power that the power grid can supply at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the maximum usable discharge power of the stored energy. Let τ be the initial load prediction value of the target base station at time τ. This represents the power supply safety margin at time τ.

[0023] Optionally, the optimization module is specifically used for: The initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information are particle-coded to obtain multiple initial scheduling schemes; Based on the multi-objective optimization function and the rigid constraint of base station reliability, the adaptive weighted particle swarm optimization algorithm is used to optimize each of the initial scheduling schemes to obtain the target scheduling instruction.

[0024] Optionally, the optimization module is specifically used for: Determine at least one candidate scheduling scheme from each of the initial scheduling schemes that satisfies the rigid constraint on base station reliability; Using each of the candidate scheduling schemes as input parameters, and employing an adaptive weighted particle swarm optimization algorithm to solve the multi-objective optimization function, the target scheduling instruction is obtained.

[0025] Optionally, the multi-objective optimization function is:

[0026] in, This indicates the target effect of peak shaving and valley filling. Indicates the target electricity cost. Indicates the energy storage lifespan loss target. These are the weighting coefficients for the peak shaving and valley filling effect target, the electricity cost target, and the energy storage life loss target, respectively. The mathematical representation of the peak shaving and valley filling effect target is:

[0027] in, , These are sets of electricity price peak and valley periods, respectively. This represents the upper limit of the power input to the power grid during peak periods. The target power for filling the valley section. This represents the average power input to the power grid within the current time window. The coefficient is the smoothing penalty; max(0,·) indicates that only the portion exceeding the threshold is penalized. Let τ represent the grid input power at time τ; The mathematical representation of the electricity cost target is:

[0028] in, Let τ be the time-of-use electricity price. The scheduling time step; The energy storage lifetime loss target is expressed as:

[0029] in, The replacement cost over the entire lifecycle of the energy storage system, This represents the equivalent number of charge-discharge cycles generated within the current time window. For the battery's rated cycle life, This refers to the rated capacity of the energy storage.

[0030] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of an energy scheduling method as described in any one of the first aspects.

[0031] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of an energy scheduling method as described in any one of the first aspects.

[0032] The beneficial effects of this application are as follows: By integrating multi-dimensional data and XGBoost rolling prediction, the accuracy of photovoltaic and load forecasting is improved, overcoming the problem of large errors in traditional models under nonlinear scenarios, and providing an accurate data foundation for scheduling. By constructing a multi-objective optimization function that includes peak shaving and valley filling, cost, and lifespan, a comprehensive balance between economy, grid friendliness, and equipment lifespan is achieved, avoiding the one-sidedness of single-objective optimization. Combining rigid constraints on base station reliability with an adaptive weighted particle swarm optimization algorithm, the absolute safety of power supply to communication base stations is ensured, and the adaptive characteristics of the algorithm are used to quickly converge to the global optimum, solving the problem that traditional algorithms are prone to getting trapped in local optima and are computationally time-consuming, thus achieving efficient, safe, and low-cost scheduling of base station energy.

[0033] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart of an energy dispatching method provided in an embodiment of this application is shown; Figure 2 This application provides a flowchart illustrating how to obtain a target scheduling instruction according to an embodiment of the present application. Figure 3 This invention provides a flowchart illustrating yet another method for obtaining a target scheduling instruction according to an embodiment of this application. Figure 4 This invention provides a flowchart illustrating yet another method for obtaining a target scheduling instruction according to an embodiment of this application. Figure 5 This paper shows a schematic diagram of the structure of an energy dispatching device provided in an embodiment of this application; Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0037] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0038] In application scenarios where communication base stations are equipped with distributed photovoltaic (PV) power generation systems and electrochemical energy storage devices, especially in grid areas implementing time-of-use policies, base stations are not only nodes in the communication network but also typical microgrid systems. The system needs to collect real-time data on PV output, base station load, and weather conditions. While ensuring uninterrupted power supply to communication equipment around the clock, it utilizes energy storage devices to store and release electrical energy. Its core objective is to reduce power draw from the grid during peak load periods and store energy during off-peak periods, thereby reducing base station operating costs, mitigating the impact of PV fluctuations on the grid, and extending the lifespan of energy storage batteries.

[0039] Existing technical solutions typically employ logic control based on fixed thresholds or traditional day-ahead scheduling strategies. The former simply performs charging and discharging switching based on preset upper and lower limits of SOC (State of Charge) or fixed time-of-use pricing periods; the latter relies on ordinary neural networks to predict photovoltaic and load conditions for the next day and formulate a 24-hour static scheduling plan.

[0040] However, the aforementioned existing technologies have significant technical drawbacks. First, the fixed threshold strategy lacks flexibility and cannot cope with random fluctuations in photovoltaic output and the dynamics of base station load, easily leading to phenomena such as charging failures or no discharge during peak hours, and even posing a risk of base station power outages. Second, traditional forecasting models have weak feature extraction capabilities for nonlinear meteorological data, resulting in low forecast accuracy, and the day-ahead static plan cannot be corrected based on real-time conditions, leading to large deviations between actual execution and the plan, resulting in reverse peak shaving. Finally, existing optimization algorithms often have excessively high computational complexity, making it difficult to achieve millisecond-level real-time response on edge gateways with limited computing power at base stations, and most of them ignore the cycle life loss of energy storage batteries, resulting in high long-term operation and maintenance costs.

[0041] Based on this, this application proposes an energy dispatching method. First, real-time data from the target base station is acquired and a multi-dimensional feature dataset is constructed. Then, this dataset is input into a pre-trained XGBoost model, which predicts initial photovoltaic and load values ​​at multiple time scales within the current time window based on a preset rolling step size. Next, a multi-objective optimization function is constructed, incorporating peak shaving and valley filling effects, electricity costs, and energy storage lifetime loss. Finally, based on this function and the rigid constraints of base station reliability, an adaptive weighted particle swarm optimization algorithm is used to optimize the prediction results, energy storage power, and state information to obtain the charging and discharging target dispatching command. This application improves prediction accuracy through rolling prediction using the XGBoost model. Combined with multi-objective optimization and reliability constraints, it can effectively reduce electricity costs, smooth grid load, and extend energy storage lifetime while ensuring the safety of base station power supply.

[0042] Figure 1 A flowchart of an energy dispatching method provided in an embodiment of this application is shown, with reference to... Figure 1 The subject executing this method can be an electronic device, such as... Figure 1 As shown, the method includes: S101. Obtain real-time data from the target base station and construct a multi-dimensional feature dataset based on the real-time data. The real-time data includes: load data, photovoltaic data, meteorological data, and power grid data.

[0043] A target base station refers to a communication base station equipped with a distributed photovoltaic power generation system and an electrochemical energy storage device. Real-time data refers to dynamic data reflecting the operating status of the base station collected at the current scheduling moment or within a very short time window.

[0044] Load data includes real-time power consumption data for base station communication equipment and auxiliary equipment such as air conditioners. Photovoltaic data includes real-time power, voltage, and current output from photovoltaic inverters. Meteorological data includes total horizontal radiation, direct radiation, diffuse radiation, ambient temperature, relative humidity, wind speed, wind direction, and hourly weather type, obtained through base station weather stations or external interfaces at the corresponding time. Grid data includes local time-of-use pricing periods, peak-valley-flat pricing standards, peak shaving and valley filling assessment indicators for grid demand-side response, and real-time grid power supply parameters.

[0045] After obtaining the real-time data, data preprocessing can be performed, including handling outliers and missing values, and data normalization. Optionally, the 3σ criterion can be used to remove outliers in photovoltaic output and base station load data. For missing values ​​caused by data acquisition interruption, Lagrange interpolation can be used to fill them in to ensure the continuity of the data sequence. Then, the Min-Max normalization method is used to map all input feature data to the [0,1] interval to eliminate the influence of different units on model training. The normalization formula is shown in the following formula (1): (1) Where x is the original feature value, These are the normalized eigenvalues. , These are the maximum and minimum values ​​of the feature sequence, respectively; when = When this happens, the normalized value of the feature is set to 0 or processed according to the default value of the project to avoid the denominator being 0.

[0046] After data preprocessing, a multi-dimensional feature dataset can be constructed based on the preprocessed real-time data. The multi-dimensional feature dataset includes basic features and derived features. Basic features include meteorological features, real-time photovoltaic output, real-time base station load, and real-time electricity price. Derivative features include time features (hours, minutes, whether it is a peak / valley / normal period, month, season), meteorological lag features (rate of change of radiation in the previous 1h / 3h, rate of change of temperature), photovoltaic historical features (output values ​​in the previous 15min / 30min / 1h, daily cumulative output), and load features (average load in the same period in history, rate of change of load), etc.

[0047] S102. Input the multi-dimensional feature dataset into the pre-trained Extreme Gradient Boosting (XGBoost) model. The XGBoost model predicts the initial prediction results for multiple time scales under the current time window based on the preset rolling step size and the multi-dimensional feature dataset. The initial prediction results include: initial photovoltaic prediction value and initial load prediction value.

[0048] Optionally, an XGBoost model can be constructed using the XGBoost algorithm, wherein the base learner of the XGBoost model is a CART regression tree, and the objective function of the XGBoost model is a squared loss function with a regularization term. While ensuring fitting accuracy, the model complexity is controlled to avoid overfitting. The objective function is shown in the following equation (2): (2) Where Θ is the parameter set of the XGBoost model, n is the number of samples, and yi is the true photovoltaic output value of the i-th sample. Let K be the model prediction value for the i-th sample, and K be the number of regression trees. Let be the prediction function of the k-th regression tree; regularization term. The formula used to control model complexity is shown in equation (3) below: (3) Where T is the number of leaf nodes in the regression tree, and w is the weight vector of the leaf nodes. Generate penalty coefficients for leaf nodes. The weight regularization coefficient; , It is used only for XGBoost model training and is independent of the subsequent scheduling target weights.

[0049] The rolling step size refers to the time interval at which the prediction window slides forward. For example, the XGBoost model is called every 15 minutes (the rolling step size) to perform a prediction for the next 4 hours (the current time window), resulting in initial photovoltaic (PV) and load predictions. The initial prediction results refer to the model's output estimates of PV power output and base station load over a future period.

[0050] During the offline phase, historical meteorological, load, and photovoltaic (PV) data from multi-dimensional feature data can be used to train the XGBoost model, enabling it to understand the impact of weather changes on PV, as well as the impact of time and holidays on load. Then, during the online operation phase, a rolling time-domain strategy is employed to predict initial forecasts for multiple scheduling steps within the current time window. For example, if the forecast is set for the next 4 hours (time window), the input data is updated and the forecast is re-predicted every 15 minutes (scheduling step).

[0051] The current time window can be a period of time in the future after the current moment. The XGBoost model can output the predicted values ​​for the current time window at a series of future moments, thus obtaining the initial prediction results for multiple moments within the current time window.

[0052] In one possible implementation, incremental learning can be used to collect real-time data every 24 hours and add it to the training samples of the XGBoost model. This allows for incremental training of the model and rolling updates of the model's parameters, thereby adapting to changes in characteristics caused by seasonal changes, aging of photovoltaic modules, and adjustments to base station services, ensuring the long-term predictive accuracy of the model.

[0053] S103. Construct a multi-objective optimization function that includes peak shaving and valley filling effect, electricity cost and energy storage life loss. The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect objective, electricity cost objective, and energy storage life loss objective.

[0054] The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect, electricity cost, and energy storage lifespan loss. Peak shaving and valley filling effect refers to smoothing the load curve of the base station on the grid by reducing the power drawn from the grid during peak hours and increasing the power drawn from the grid during off-peak hours through energy storage regulation. Energy storage lifespan loss refers to the capacity decay that occurs during battery charging and discharging. Frequent and deep charging and discharging accelerates battery aging.

[0055] Optionally, constraints on peak shaving and valley filling effect, electricity cost, and energy storage life loss can be pre-constructed, and then a multi-objective optimization function including peak shaving and valley filling effect, electricity cost, and energy storage life loss can be constructed based on the constraint information.

[0056] Peak shaving and valley filling effect constraints are used to limit the power boundaries for peak shaving, valley filling, and smooth operation. Mathematically, they can be expressed as the following equations (4)-(6): (4) (5) (6) in, , They are sets of peak segments and valley segments, Let τ be the power input to the power grid. This represents the upper limit of the power input to the power grid during peak periods. The target power for filling the valley section. The average power input to the power grid over the rolling time domain. The threshold that allows the power input to the grid to deviate from the average value.

[0057] Electricity cost constraints are used to limit time-of-use pricing parameters and anti-reverse dispatch rules. Mathematically, they can be expressed as the following equations (7)-(9): (7) (8) (9) in, Let τ be the time-of-use electricity price. , , The electricity prices are peak, flat, and valley, respectively. During peak hours, energy storage is prohibited from charging from the grid when the predicted photovoltaic load is not higher than the predicted load. During valley hours, energy storage is prohibited from discharging when the grid and the predicted photovoltaic load can meet the load and safety margin, so as to avoid reverse scheduling of charging during peak hours and discharging during valley hours.

[0058] Energy storage lifespan loss constraints can be constraints on the safe operation of energy storage systems, including boundary, Dynamics, charge / discharge power, charge / discharge mutual exclusion, charge / discharge rate and Lifetime constraints can be mathematically expressed as equations (10)-(13): (10) (11) (12) (13) in, The energy storage state of charge at time τ, For example, the value can be set to 20%. For example, the value can be set to 90%; , τ represents the energy storage charging power and discharging power at time τ, and Pch_max and Pdis_max represent the rated maximum charging power and rated maximum discharging power of the energy storage, respectively. , These are charging efficiency and discharging efficiency, respectively. For the rated capacity of energy storage, The scheduling time step is set; the charge / discharge rate does not exceed the battery's rated rate, the cumulative equivalent charge / discharge cycle count does not exceed the battery's rated cycle life, and the cycle loss of a single scheduling operation is included in the optimization objective.

[0059] S104. Based on the multi-objective optimization function and the rigid constraints of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize each initial prediction result, the energy storage charging power, energy storage discharging power and energy storage status information corresponding to the initial prediction results, the target scheduling instruction is obtained. The target scheduling instruction includes: charging power instruction and discharging power instruction.

[0060] It should be noted that steps S102 and S104 can be executed at different times. Step S102 performs rolling prediction in different time windows to obtain the initial prediction results of each scheduling step in each time window. When a scheduling instruction needs to be generated, step S104 is executed to obtain the initial prediction results of the current time window and perform optimization processing based on the initial prediction results of the current time window to obtain the current target scheduling instruction.

[0061] In this process, while generating the initial prediction results, the energy storage charging power, energy storage discharging power, and energy storage status information corresponding to the initial prediction results can be obtained. For example, if the initial prediction results for 16 scheduling steps within the next 4 hours are generated, the energy storage charging power, energy storage discharging power, and energy storage status information corresponding to the times of these 16 scheduling steps within the next 4 hours can be obtained as the energy storage charging power, energy storage discharging power, and energy storage status information corresponding to the initial prediction results.

[0062] The rigid constraint on base station reliability is used to ensure that at any given time, the sum of the maximum power that the power grid can supply, the available output of photovoltaic power, and the maximum discharge power of energy storage is not less than the base station load and the reserved safety margin. The mathematical form can be expressed as the following formula (14): (14) in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the maximum power that the power grid can supply at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the maximum usable discharge power of the stored energy. Let τ be the initial load prediction value of the target base station at time τ. This represents the power supply safety margin at time τ.

[0063] Adaptive Weighted Particle Swarm Optimization (PSO) is an improved swarm intelligence optimization algorithm. PSO simulates the foraging behavior of bird flocks, updating the search position based on the historical best solutions of individuals and the group.

[0064] In one possible implementation, a solution space can be constructed based on each initial prediction result, the energy storage charging power, energy storage discharging power, and energy storage state information corresponding to the initial prediction results. An adaptive weighted particle swarm optimization algorithm is then used to perform optimization in the solution space. After multiple iterations, the optimal scheduling scheme that satisfies the multi-objective optimization function and the rigid constraints of base station reliability is found. Target scheduling instructions are then generated based on the optimal scheduling scheme.

[0065] The target scheduling instruction refers to the specific charging and discharging power setting value that is finally issued to the energy storage hardware device after algorithm optimization.

[0066] As one possible implementation, the optimal scheduling scheme consists of initial prediction results, the energy storage charging power and discharging power corresponding to the initial prediction results, energy storage status information, and scheduling instructions. After determining the optimal scheduling scheme, the scheduling instructions in the optimal scheduling scheme can be directly used as the target scheduling instructions.

[0067] In this embodiment, multi-dimensional data fusion and XGBoost rolling prediction significantly improve the accuracy of photovoltaic and load forecasting, overcoming the problem of large errors in traditional models under nonlinear scenarios, and providing an accurate data foundation for scheduling. By constructing a multi-objective optimization function that includes peak shaving and valley filling, cost, and lifespan, a comprehensive balance between economy, grid friendliness, and equipment lifespan is achieved, avoiding the one-sidedness of single-objective optimization. Combining rigid constraints on base station reliability with an adaptive weighted particle swarm optimization algorithm, the absolute security of power supply to communication base stations is ensured, and the adaptive characteristics of the algorithm are used to quickly converge to the global optimum, solving the problem that traditional algorithms are prone to getting trapped in local optima and are computationally time-consuming, thus achieving efficient, safe, and low-cost scheduling of base station energy.

[0068] Figure 2 This application provides a flowchart illustrating how to obtain a target scheduling instruction, as shown in the embodiment of the present application. Figure 2 As shown, the process described above, based on a multi-objective optimization function and rigid constraints on base station reliability, and using an adaptive weighted particle swarm optimization algorithm to optimize each initial prediction result, the corresponding energy storage charging power, energy storage discharging power, and energy storage state information, to obtain the target scheduling instruction includes: S201. Determine the initial input power based on the initial prediction results, the energy storage charging power corresponding to the initial prediction results, and the energy storage discharging power.

[0069] Here, initial input power refers to the net power actually obtained by the base station system from the external power grid at the current scheduling moment. Energy storage charging power refers to the power value of the energy storage system absorbing electrical energy for charging at the current moment. Energy storage discharging power refers to the power value of the energy storage system releasing electrical energy for discharging at the current moment.

[0070] Optionally, the initial prediction results, energy storage charging power, and energy storage discharging power can be used as input parameters, and calculated using the formula... The initial input power is calculated.

[0071] in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the power input to the power grid at time τ. Let τ be the initial load prediction value of the target base station at time τ. The energy storage charging power at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the energy storage discharge power at time τ.

[0072] S202. Based on the multi-objective optimization function and the rigid constraints of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize the initial input power, energy storage charging power, energy storage discharging power and energy storage status information, the target scheduling instruction is obtained.

[0073] Energy storage status information refers to the current physical operating status of an energy storage system, including the state of charge (SOC), which is the ratio of the battery's current remaining charge to its rated capacity, as well as battery temperature, health status, etc.

[0074] Optionally, the grid input power, energy storage charging and discharging power, and SOC state can be used as decision variables in the particle swarm optimization algorithm. An adaptive weighted particle swarm optimization algorithm iterates continuously in a multi-dimensional space to find the variable combination that minimizes the score of the multi-objective optimization function without violating the rigid constraints of base station reliability. This variable combination is the optimal scheduling scheme, and the scheduling instructions in the optimal scheduling scheme are used as the target scheduling instructions.

[0075] In this embodiment, the initial input power is determined by introducing power balance calculation, and the dynamic coupling relationship between the power grid, photovoltaics, energy storage, and load is mathematically quantified, providing accurate boundary conditions for subsequent optimization. Simultaneously, the power grid input power, charging / discharging power, and energy storage state information are used as joint decision variables in an adaptive weighted particle swarm optimization algorithm for optimization, achieving global collaborative solution of multiple variables under complex constraints. This not only ensures the physical feasibility and security of the scheduling strategy but also effectively improves the algorithm's convergence speed and optimization accuracy.

[0076] Figure 3 This application provides a flowchart illustrating another method for obtaining a target scheduling instruction, as shown in the embodiments of this application. Figure 3 As shown, the process described above, based on a multi-objective optimization function and rigid constraints on base station reliability, and using an adaptive weighted particle swarm optimization algorithm to optimize the initial input power, energy storage charging power, energy storage discharging power, and energy storage state information to obtain the target scheduling instruction, includes: S301. Perform particle coding on the initial input power, energy storage charging power, energy storage discharging power, and energy storage state information to obtain multiple initial scheduling schemes.

[0077] Initial input power, energy storage charging power, energy storage discharging power, and energy storage state information are decision variables at each time step. By performing particle encoding, the decision variables at each time step can be concatenated into a high-order vector to obtain the particle position. At the same time, a velocity vector is randomly initialized for each particle, and each particle represents an initial scheduling scheme.

[0078] By randomly generating a certain number of particles within a set boundary range, the adaptive weighted particle swarm optimization algorithm constructs a candidate solution population containing multiple initial scheduling schemes, covering various possible charging and discharging strategies.

[0079] S302. Based on the multi-objective optimization function and the rigid constraints of base station reliability, the adaptive weighted particle swarm optimization algorithm is used to optimize each initial scheduling scheme to obtain the target scheduling instruction.

[0080] During the optimization process, the comprehensive score of each particle in the population under the multi-objective optimization function can be calculated first. Simultaneously, it is verified whether the scheme violates the rigid constraints of base station reliability. For particles that violate the rigid constraints, a penalty function is applied or they are directly eliminated to ensure that the search process always proceeds within the safe and feasible region.

[0081] Before each iteration, the inertia weights are dynamically calculated and adjusted based on the fitness distribution of particles in the current population. In the early stages of the search or when particles are dispersed, the weights are increased to enhance global exploration capabilities and avoid getting trapped in local optima; in the later stages of the search or when particles tend to converge, the weights are decreased to enhance local exploration capabilities and accelerate convergence to the optimal solution. Then, based on the adaptive weights, individual historical optimal positions, and global optimal positions, the velocity update formula and position update formula are used to drive all particles to move in the solution space, generating a new generation of scheduling schemes.

[0082] Through multiple rounds of evaluation, weight adjustment, and state updates, until a preset termination condition is met, such as reaching the maximum number of iterations or the fitness error being less than a threshold, the position vector corresponding to the globally optimal particle represents the optimal scheduling strategy with the best overall benefit. This strategy can be decoded and output to obtain the final target scheduling instruction.

[0083] In this embodiment, an initial candidate solution space is constructed by transforming multidimensional scheduling variables into particle codes. Combined with an adaptive weighting mechanism, the algorithm can dynamically balance global exploration and local development capabilities based on the optimization progress, effectively avoiding getting trapped in local optima. Simultaneously, rigid constraints are used as safety boundaries for verification, ensuring the safety of the scheduling strategy.

[0084] Figure 4 This application provides a flowchart illustrating another method for obtaining a target scheduling instruction, as shown in the embodiments of this application. Figure 4 As shown, the process of obtaining the target scheduling instruction by optimizing each initial scheduling scheme based on a multi-objective optimization function and rigid constraints on base station reliability, and using an adaptive weighted particle swarm optimization algorithm, includes: S401. Determine at least one candidate scheduling scheme that satisfies the rigid constraints of base station reliability among the initial scheduling schemes.

[0085] In one possible implementation, after constructing a solution space containing all initial scheduling schemes, the initial scheduling schemes in the solution space can first be screened based on the rigid constraints of base station reliability. A point-by-point power balance calculation is performed on each initial scheduling scheme along the time axis. If it is found that the total power of the grid, photovoltaic, and energy storage at any given time cannot meet the rigid constraints, the scheme is directly deemed infeasible and eliminated or subject to a very large penalty. The initial scheduling schemes that pass the verification are used as candidate scheduling schemes. After verifying all initial scheduling schemes, the new solution space composed of all the candidate scheduling schemes is used as the solution space for the optimization process of the adaptive weighted particle swarm optimization algorithm.

[0086] S402. Using each candidate scheduling scheme as input parameters, and using the adaptive weighted particle swarm optimization algorithm to solve the multi-objective optimization function, the target scheduling instruction is obtained.

[0087] Optionally, the adaptive weighted particle swarm optimization algorithm can use candidate scheduling schemes as a starting point or search space. During the iteration process, the algorithm calculates the comprehensive score of each candidate scheme under the multi-objective optimization function, and uses adaptive weights to adjust the search step size, guiding the particle swarm to continuously evolve towards the direction with the highest comprehensive benefit among the candidate scheduling schemes, and finally converges to the global optimum, which is then decoded into specific charging power commands and discharging power commands.

[0088] The multi-objective optimization function is expressed as follows (15): (15) in, This indicates the target effect of peak shaving and valley filling. Indicates the target electricity cost. Indicates the energy storage lifespan loss target. These are the weighting coefficients for the peak shaving and valley filling effect target, the electricity cost target, and the energy storage life loss target, respectively. The mathematical expression for the peak shaving and valley filling effect is as follows (16): (16) in, , These are sets of electricity price peak and valley periods, respectively. This represents the upper limit of the power input to the power grid during peak periods. The target power for filling the valley section. This represents the average power input to the power grid within the current time window. The coefficient is the smoothing penalty; max(0,·) indicates that only the portion exceeding the threshold is penalized. Let τ represent the grid input power at time τ; The mathematical expression for the electricity cost target is as follows (17): (17) in, Let τ be the time-of-use electricity price. The scheduling time step; The energy storage life loss target is expressed as follows (18): (18) in, The replacement cost over the entire lifecycle of the energy storage system, This represents the equivalent number of charge-discharge cycles generated within the current time window. For the battery's rated cycle life, This refers to the rated capacity of the energy storage.

[0089] In one possible implementation, equation (19) can also be used as a multi-objective optimization function.

[0090] (19) in, Indicates power balance constraints, and , , This represents the dynamic constraints of SOC. This indicates the charging efficiency, and its value can be 0.95. This represents the discharge efficiency, and its value can be 0.93. Used to constrain business continuity, and . It is used to constrain schedulable capacity. Used to constrain the power of grid nodes. This is a penalty item for battery degradation. This is a penalty item for excessive leniency.

[0091] Scheduled capacity ,in, This represents the lower safety limit, and its value can be 20%. This indicates the upper limit of safety, and its value can be 90%. This indicates the current state of charge (SBC). It also considers battery health. and formula The capacity decay is corrected. The final schedulable capacity is updated to... .

[0092] If only the load value is predicted and scheduling is performed based on the load value, the objective function shown in equation (19) above can be optimized, and the objective function can be solved and optimized based on the adaptive hybrid particle swarm algorithm to obtain the target scheduling instruction.

[0093] After receiving and issuing the target scheduling instruction, the instruction can be corrected based on a pre-set deviation threshold trigger mechanism, taking into account the execution result of the target scheduling instruction and the deviation threshold. For example, when the peak grid input power exceeds the peak limit by 5%, or the deviation between the predicted and actual values ​​exceeds 15%, an emergency scheduling correction is immediately triggered, suspending the original scheduling plan and re-invoking the optimization model to solve for new scheduling instructions, thus forming a closed-loop scheduling system of prediction, optimization, execution, feedback, and correction.

[0094] Based on the same inventive concept, this application also provides an energy dispatching device corresponding to the energy dispatching method. Since the principle of the device in this application is similar to the energy dispatching method described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0095] Figure 5 A schematic diagram of the structure of an energy dispatching device provided in an embodiment of this application is shown.

[0096] The acquisition module 501 is used to acquire real-time data of the target base station and construct a multi-dimensional feature dataset based on the real-time data. The real-time data includes: load data, photovoltaic data, meteorological data and power grid data. The prediction module 502 is used to input the multi-dimensional feature dataset into the pre-trained extreme gradient boosting XGBoost model. The XGBoost model predicts the initial prediction results of multiple time scales under the current time window based on the preset rolling step size and the multi-dimensional feature dataset. The initial prediction results include: initial photovoltaic prediction value and initial load prediction value. Module 503 is used to construct a multi-objective optimization function that includes peak shaving and valley filling effect, electricity cost and energy storage life loss. The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect objective, electricity cost objective, and energy storage life loss objective. The optimization module 504 is used to optimize each initial prediction result, the energy storage charging power, energy storage discharging power and energy storage status information corresponding to the initial prediction result based on the multi-objective optimization function and the rigid constraint of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to obtain the target scheduling instruction, which includes the charging power instruction and the discharging power instruction.

[0097] Optionally, the optimization module 504 is specifically used for: The initial input power is determined based on the initial prediction results, the corresponding energy storage charging power, and the energy storage discharging power. Based on a multi-objective optimization function and rigid constraints on base station reliability, and using an adaptive weighted particle swarm optimization algorithm to optimize the initial input power, energy storage charging power, energy storage discharging power, and energy storage status information, the target scheduling instruction is obtained.

[0098] Optionally, the optimization module 504 is specifically used for: Using the initial prediction results, energy storage charging power, and energy storage discharging power as input parameters, the formula is used... The initial input power is calculated; in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the power input to the power grid at time τ. Let τ be the initial load prediction value of the target base station at time τ. The energy storage charging power at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the energy storage discharge power at time τ.

[0099] Optionally, the mathematical expression for the rigid constraint on base station reliability is:

[0100] in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the maximum power that the power grid can supply at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the maximum usable discharge power of the stored energy. Let τ be the initial load prediction value of the target base station at time τ. This represents the power supply safety margin at time τ.

[0101] Optionally, the optimization module 504 is specifically used for: Particle coding is performed on the initial input power, energy storage charging power, energy storage discharging power, and energy storage state information to obtain multiple initial scheduling schemes; Based on a multi-objective optimization function and rigid constraints on base station reliability, an adaptive weighted particle swarm optimization algorithm is used to optimize each initial scheduling scheme to obtain the target scheduling instruction.

[0102] Optionally, the optimization module 504 is specifically used for: Identify at least one candidate scheduling scheme that satisfies the rigid constraints on base station reliability among the initial scheduling schemes; Using each candidate scheduling scheme as input parameters, and employing an adaptive weighted particle swarm optimization algorithm to solve the multi-objective optimization function, the target scheduling instruction is obtained.

[0103] Optionally, the multi-objective optimization function is:

[0104] in, This indicates the target effect of peak shaving and valley filling. Indicates the target electricity cost. Indicates the energy storage lifespan loss target. These are the weighting coefficients for the peak shaving and valley filling effect target, the electricity cost target, and the energy storage life loss target, respectively. The mathematical representation of the peak shaving and valley filling effect is:

[0105] in, , These are sets of electricity price peak and valley periods, respectively. This represents the upper limit of the power input to the power grid during peak periods. The target power for filling the valley section. This represents the average power input to the power grid within the current time window. The coefficient is the smoothing penalty; max(0,·) indicates that only the portion exceeding the threshold is penalized. Let τ represent the grid input power at time τ; The mathematical representation of the electricity cost target is:

[0106] in, Let τ be the time-of-use electricity price. The scheduling time step; The energy storage lifespan loss target is expressed as:

[0107] in, The replacement cost over the entire lifecycle of the energy storage system, This represents the equivalent number of charge-discharge cycles generated within the current time window. For the battery's rated cycle life, This refers to the rated capacity of the energy storage.

[0108] This application's embodiments significantly improve the accuracy of photovoltaic and load forecasting through multi-dimensional data fusion and XGBoost rolling prediction, overcoming the problem of large errors in traditional models under nonlinear scenarios and providing an accurate data foundation for scheduling. By constructing a multi-objective optimization function that includes peak shaving and valley filling, cost, and lifespan, a comprehensive balance between economy, grid friendliness, and equipment lifespan is achieved, avoiding the one-sidedness of single-objective optimization. Combining rigid constraints on base station reliability with an adaptive weighted particle swarm optimization algorithm, the absolute security of power supply to communication base stations is ensured, while the algorithm's adaptive characteristics are utilized to quickly converge to the global optimum. This solves the problem of traditional algorithms easily getting trapped in local optima and being computationally time-consuming, achieving efficient, safe, and low-cost scheduling of base station energy.

[0109] Figure 6 This illustration shows a schematic diagram of an electronic device provided in an embodiment of this application, including: a processor 601, a storage medium 602, and a bus 603. The storage medium 602 stores machine-readable instructions executable by the processor 601. When the electronic device runs an energy scheduling method as described in the embodiment, the processor 601 communicates with the storage medium 602 via the bus 603. The processor 601 executes the machine-readable instructions, and the preamble of the method item of the processor 601 executes the steps in the above-described energy scheduling method.

[0110] This application also provides a computer-readable storage medium storing a computer program that is executed by a processor, which performs the steps of the above-described energy scheduling method.

[0111] In this embodiment, the computer program, when run by the processor, can also execute other machine-readable instructions to perform other methods as described in the embodiments. For details on the specific execution steps and principles, please refer to the description of the embodiments, which will not be repeated here.

[0112] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0115] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0116] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0117] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. An energy dispatching method, characterized in that, include: Acquire real-time data from the target base station and construct a multi-dimensional feature dataset based on the real-time data, which includes load data, photovoltaic data, meteorological data, and power grid data. The multi-dimensional feature dataset is input into a pre-trained extreme gradient boosting XGBoost model. The XGBoost model predicts initial prediction results for multiple time scales under the current time window based on a preset rolling step size and the multi-dimensional feature dataset. The initial prediction results include: initial photovoltaic prediction value and initial load prediction value. A multi-objective optimization function is constructed, which includes peak shaving and valley filling effect, electricity cost and energy storage life loss. The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect objective, electricity cost objective, and energy storage life loss objective. Based on the multi-objective optimization function and the rigid constraint of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize each of the initial prediction results, the energy storage charging power, the energy storage discharging power and the energy storage status information corresponding to the initial prediction results, the target scheduling instruction is obtained. The target scheduling instruction includes: charging power instruction and discharging power instruction.

2. The method according to claim 1, characterized in that, The process involves optimizing the initial prediction results, the corresponding energy storage charging power, energy storage discharging power, and energy storage status information based on the multi-objective optimization function and the rigid constraints of base station reliability using an adaptive weighted particle swarm optimization algorithm. The resulting target scheduling instruction includes: The initial input power is determined based on the initial prediction results, the energy storage charging power corresponding to the initial prediction results, and the energy storage discharging power. Based on the multi-objective optimization function and the rigid constraints of base station reliability, and using the adaptive weighted particle swarm optimization algorithm to optimize the initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information, the target scheduling instruction is obtained.

3. The method according to claim 2, characterized in that, The step of determining the initial input power based on the initial prediction result, the energy storage charging power corresponding to the initial prediction result, and the energy storage discharging power includes: Using the initial prediction result, the energy storage charging power, and the energy storage discharging power as input parameters, the formula is used... The initial input power is calculated. in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the power input to the power grid at time τ. Let τ be the initial load prediction value of the target base station at time τ. The energy storage charging power at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the energy storage discharge power at time τ.

4. The method according to claim 1, characterized in that, The mathematical expression for the rigid constraint on base station reliability is: in, This indicates the scrolling time domain of the current time window. This represents the start time of the current time window, N is the number of time-domain scheduling steps in the current time window, and τ is any scheduling time within this time domain. Let τ be the maximum power that the power grid can supply at time τ. Let τ be the initial photovoltaic prediction value of the target base station. Let τ be the maximum usable discharge power of the stored energy. Let τ be the initial load prediction value of the target base station at time τ. This represents the power supply safety margin at time τ.

5. The method according to claim 2, characterized in that, The process, based on the multi-objective optimization function and the rigid constraints of base station reliability, uses an adaptive weighted particle swarm optimization algorithm to optimize the initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information to obtain the target scheduling instruction, including: The initial input power, the energy storage charging power, the energy storage discharging power, and the energy storage state information are particle-coded to obtain multiple initial scheduling schemes; Based on the multi-objective optimization function and the rigid constraint of base station reliability, the adaptive weighted particle swarm optimization algorithm is used to optimize each of the initial scheduling schemes to obtain the target scheduling instruction.

6. The method according to claim 5, characterized in that, The process of optimizing each initial scheduling scheme based on the multi-objective optimization function and the rigid constraints of base station reliability, and using an adaptive weighted particle swarm optimization algorithm to obtain the target scheduling instruction, includes: Determine at least one candidate scheduling scheme from each of the initial scheduling schemes that satisfies the rigid constraint on base station reliability; Using each of the candidate scheduling schemes as input parameters, and employing an adaptive weighted particle swarm optimization algorithm to solve the multi-objective optimization function, the target scheduling instruction is obtained.

7. The method according to any one of claims 1-6, characterized in that, The multi-objective optimization function is: in, This indicates the target effect of peak shaving and valley filling. Indicates the target electricity cost. Indicates the energy storage lifespan loss target. These are the weighting coefficients for the peak shaving and valley filling effect target, the electricity cost target, and the energy storage life loss target, respectively. The mathematical representation of the peak shaving and valley filling effect target is: in, , These are sets of electricity price peak and valley periods, respectively. This represents the upper limit of the power input to the power grid during peak periods. The target power for filling the valley section. This represents the average power input to the power grid within the current time window. The coefficient is the smoothing penalty; max(0,·) indicates that only the portion exceeding the threshold is penalized. Let τ represent the grid input power at time τ; The mathematical representation of the electricity cost target is: in, Let τ be the time-of-use electricity price. The scheduling time step; The energy storage lifetime loss target is expressed as: in, The replacement cost over the entire lifecycle of the energy storage system, This represents the equivalent number of charge-discharge cycles generated within the current time window. For the battery's rated cycle life, This refers to the rated capacity of the energy storage.

8. An energy dispatching device, characterized in that, include: The acquisition module is used to acquire real-time data of the target base station and construct a multi-dimensional feature dataset based on the real-time data. The real-time data includes: load data, photovoltaic data, meteorological data, and power grid data. The prediction module is used to input the multi-dimensional feature dataset into a pre-trained extreme gradient boosting XGBoost model. The XGBoost model predicts initial prediction results for multiple time scales under the current time window based on a preset rolling step size and the multi-dimensional feature dataset. The initial prediction results include: initial photovoltaic prediction value and initial load prediction value. The construction module is used to construct a multi-objective optimization function that includes peak shaving and valley filling effect, electricity cost and energy storage life loss. The optimization objectives of the multi-objective optimization function include: peak shaving and valley filling effect objective, electricity cost objective, and energy storage life loss objective. The optimization module is used to optimize each of the initial prediction results, the energy storage charging power, the energy storage discharging power and the energy storage status information corresponding to the initial prediction results, based on the multi-objective optimization function and the rigid constraints of base station reliability, and using an adaptive weighted particle swarm optimization algorithm to obtain the target scheduling instruction, which includes a charging power instruction and a discharging power instruction.

9. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of an energy scheduling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of an energy dispatching method as described in any one of claims 1 to 7.

Images