Optical storage and charging energy allocation system based on data driving and control method

Abstract

The invention discloses a light storage and charging energy allocation system based on data driving and a control method, and relates to the technical field of light storage and charging energy allocation, and the method comprises the following steps: carrying out the real-time collection of the real-time power generation amount of a photovoltaic power generation system, the current power state of an energy storage battery, and the power consumption data of a load; and cleaning, normalization processing and time serialization are carried out on the collected data, so that the accuracy and continuity of the input data are ensured. According to the invention, by monitoring the power consumption data of the load in real time, combining with the long-term energy prediction model and dynamically adjusting the energy storage and discharge strategy, excessive discharge in a non-peak period is avoided, sufficient energy storage in a peak period is ensured, the power grid dependence is reduced, and the power consumption cost is reduced. Through load sudden increase detection and an adaptive scheduling algorithm, the system can quickly cope with sudden load changes, preferentially guarantee power supply of key loads, flexibly adjust a discharge mode, improve endurance time and system stability, and maximize green energy utilization efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of energy allocation for photovoltaic energy storage and charging, and particularly to a data-driven energy allocation system and control method for photovoltaic energy storage and charging. Background Art

[0002] Data-driven energy allocation control for photovoltaic energy storage and charging refers to the comprehensive collection and analysis of data on photovoltaic power generation, energy storage systems, and load electricity consumption, and the use of machine learning or prediction models to dynamically optimize the scheduling of future energy supply and demand, thereby achieving efficient collaborative management of photovoltaic power generation, battery energy storage, and grid power. Specifically, taking the photovoltaic energy storage and charging system of a base station as an example, by analyzing the hourly photovoltaic power generation and load electricity consumption data in the past 15 days, it is possible to predict the amount of electricity that the energy storage system needs to store when switching from normal time (low electricity price or low load period) to peak time (high electricity price or high load period). This can store energy in advance when photovoltaic power generation is sufficient, ensure that the load electricity consumption during peak hours mainly relies on the energy storage system for power supply, thereby minimizing the phenomenon of abandoned light, and at the same time reducing the dependence on the grid for electricity consumption. This energy allocation control strategy can achieve efficient utilization of photovoltaic power, reduce the overall electricity cost, and improve the green energy utilization rate and economic benefits of the system without affecting the normal operation of the base station.

[0003] The existing technologies have the following deficiencies:

[0004] When the existing photovoltaic energy storage and charging system performs energy allocation, it is prone to ignoring the risk of excessive discharge of the energy storage caused by a sudden increase in short-term load. Traditional energy allocation usually relies on historical photovoltaic power generation and load electricity consumption data for prediction, but in special scenarios, such as base station equipment failures, sudden increases in emergency communication services, or external grid power outages, the system cannot identify abnormal changes in the load in real time. This situation may lead to premature excessive discharge of the energy storage system during normal times, resulting in insufficient power reserve during peak times, forcing the system to rely on grid power supply, increasing the electricity cost, and even possibly causing a power outage of the base station, affecting the operation of critical services. To address this problem, a load mutation detection and prediction mechanism can be introduced to monitor the load fluctuation in real time and dynamically adjust the energy storage strategy to ensure that the energy storage system has sufficient power reserve in case of emergencies, thereby achieving more stable utilization of photovoltaic power and optimization of electricity costs.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a data-driven optical storage and charging energy allocation system and control method. By real-time monitoring of load power consumption data and combining with a long-term energy supply and demand prediction model, the discharge strategy of the energy storage system is dynamically adjusted to avoid over-discharge of the energy storage battery during off-peak hours, ensure sufficient energy storage during peak hours, reduce dependence on grid power, and optimize the electricity cost. Through a load sudden increase detection mechanism and an adaptive scheduling algorithm, the system can quickly respond to sudden load change scenarios, such as a surge in communication services or a power grid interruption, automatically adjust the energy storage strategy, ensure continuous power supply for critical loads, and enhance the endurance time by flexibly adjusting the discharge mode, enhance the system stability and power supply reliability, and maximize the utilization efficiency of green energy to solve the problems in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: A data-driven optical storage and charging energy allocation control method, comprising the following steps:

[0008] Real-time collect the real-time power generation of the photovoltaic power generation system, the current power state of the energy storage battery, and the load power consumption data, and clean, normalize, and time-series the collected data to ensure the accuracy and continuity of the input data;

[0009] Based on historical photovoltaic power generation, the charge and discharge records of the energy storage system, and the load power consumption pattern, use a machine learning model to construct a long-term energy supply and demand prediction model to predict future photovoltaic power generation trends and load demand changes, so as to plan the charge and discharge strategy of the energy storage system in advance;

[0010] Compare the real-time monitored load power consumption data with the predicted data, and detect short-term load sudden increase situations based on the change rate threshold and time window;

[0011] Introduce an energy storage strategy optimization algorithm, and calculate the optimal charge and discharge strategy of the energy storage system in real time according to the current photovoltaic power generation, the power state of the energy storage battery, and the predicted value of the load demand;

[0012] Through an anomaly recognition algorithm, correlate the load mutation signal with the external power grid state to identify abnormal scenarios. Based on the recognition result, execute an adaptive scheduling algorithm to give priority to ensuring the power supply demand of critical loads, and ensure the stable operation of the system and minimize the electricity cost in case of emergencies by adjusting the power exchange between the discharge rate of the energy storage system and the power grid.

[0013] Preferably, the specific steps of real-time collecting the real-time power generation of the photovoltaic power generation system, the current power state of the energy storage battery, and the load power consumption data, and cleaning, normalizing, and time-series the collected data to ensure the accuracy and continuity of the input data are as follows:

[0014] Realize real-time data collection and transmission of photovoltaic power generation, energy storage battery power and load power consumption through sensors and data transmission protocols to ensure data integrity and real-time performance;

[0015] Clean the collected data, remove outliers and supplement missing values, and unify the data frequency by resampling to ensure the continuity and rationality of the data;

[0016] Convert data from different sources to the same scale through normalization to avoid the impact of numerical differences between features on the accuracy of model decisions;

[0017] The processed data is arranged in chronological order as time series data to capture periodic and trend changes, providing a timing input basis for load forecasting and scheduling strategies.

[0018] Preferably, based on the historical photovoltaic power generation, the charging and discharging records of the energy storage system, and the load power consumption pattern, a machine learning model is used to build a long-term energy supply and demand forecasting model to predict future photovoltaic power generation trends and load demand changes, so as to plan the charging and discharging strategy of the energy storage system in advance. The specific steps are as follows:

[0019] By selecting key features to construct a time series dataset, high-quality input data is provided for model training;

[0020] Select a machine learning model suitable for time series data and train the model through cross-validation and parameter adjustment to improve prediction accuracy and robustness;

[0021] Evaluate model performance through various indicators, and optimize the model through hyperparameter adjustment and real-time evaluation to adapt to dynamic data changes;

[0022] The prediction results are used to plan the charging and discharging strategies of the energy storage system, and combined with electricity price optimization and abnormal load identification mechanisms to achieve electricity cost optimization and improve system stability.

[0023] Preferably, the specific steps of comparing the real-time monitored load power consumption data with the predicted data and detecting the short-term load surge based on the change rate threshold and the time window are as follows:

[0024] By comparing the deviation between the real-time collected load power consumption data and the predicted data, the actual load is identified to exceed the normal range, providing a basis for sudden increase detection;

[0025] By calculating the change rate and fluctuation range of load data, the trend and instability of load changes in a short period of time can be monitored, thereby providing early warning of sudden increase risks;

[0026] Compare the rate of change with the preset threshold. When the rate of change exceeds the set value, a load mutation signal is triggered and verified multiple times to improve accuracy.

[0027] After triggering the load mutation signal, according to the photovoltaic power generation, the state of charge of the energy storage battery, and the load demand, dynamically adjust the discharge strategy of the energy storage system to ensure continuous and stable power supply to the load.

[0028] Preferably, introduce an energy storage strategy optimization algorithm. According to the current photovoltaic power generation, the state of charge of the energy storage battery, and the predicted value of the load demand, the specific steps for calculating the optimal charge and discharge strategy of the energy storage system are as follows:

[0029] First, calculate the current load demand gap, which is the difference between the actual load demand and the power provided by photovoltaic power generation, and is used to determine whether the energy storage system needs to charge or discharge. The calculation expression is as follows:

[0030] ΔP load =P demand -P solar

[0031] , where ΔP load is the load demand gap, P demand is the actual load demand, P solar is the current photovoltaic power generation;

[0032] According to the calculated load demand gap ΔP load , further calculate the charge and discharge power of the energy storage battery. The calculation expression is as follows:

[0033]

[0034] , where P battery is the charge and discharge power of the energy storage battery, P max is the maximum charge and discharge power of the energy storage battery, max(ΔP load , -P max ) is the maximum value limit, indicating that the load gap value shall not be less than the negative value of the maximum discharge power of the energy storage battery, and min(max(ΔP load , -P max ), P max ) is the minimum value limit, indicating that the load gap value shall not exceed the maximum charge power of the energy storage battery. SOC is the current state of charge of the energy storage battery, indicating the charging percentage of the battery.

[0035] Preferably, to ensure the long-term service life of the energy storage battery, comprehensively consider the influence of the battery health state on the charge and discharge power. The calculation expression is as follows:

[0036]

[0037] , where P battery_adjustedis the charge and discharge power of the energy storage battery after correcting the health state, and SOH is the health state of the energy storage battery;

[0038] When the energy storage system cannot fully meet the load demand, the dispatching decision of the power grid will be calculated by combining the real-time grid electricity price and the charge and discharge power of the energy storage battery, and it will be decided whether to purchase electricity from the grid or sell electricity to the grid. The calculation formula is as follows:

[0039] P grid =ΔP load -P battery_adjusted

[0040] , where P grid is the grid dispatching power;

[0041] If P grid >0, then purchase electricity from the grid;

[0042] If P grid <0, then sell electricity to the grid;

[0043] The grid electricity price is adjusted according to the real-time electricity price. The calculation formula is as follows:

[0044] C cost =P grid ×C grid

[0045] , where C grid is the real-time grid electricity price, and C cost is the cost of purchasing or selling electricity from the grid.

[0046] Preferably, through the anomaly recognition algorithm, the load mutation signal is associated with the external grid state for analysis to identify abnormal scenarios. Based on the recognition results, the adaptive dispatching algorithm is executed to preferentially ensure the power supply demand of critical loads, and by adjusting the power exchange between the discharge rate of the energy storage system and the grid, the specific steps to ensure the stable operation of the system and minimize the electricity cost in case of emergencies are as follows:

[0047] First, through the correlation analysis of the real-time monitored load sudden increase signal and the external grid state data, the load mutation intensity parameter is calculated to quantify the degree of the current load abnormal change. The calculation formula of the load mutation intensity parameter is as follows:

[0048]

[0049] , where I LT is the load mutation intensity parameter, ΔL is the load change amount within the time interval Δt, Δt is the time interval, and V dev is the deviation value of the grid voltage, that is, the difference between the actual voltage and the standard voltage, and F devis the deviation value of the grid frequency, that is, the difference between the actual frequency and the standard frequency. α, β, and γ are all weight parameters. α is the weight of the load change rate, indicating the degree of influence of load changes on the identification of sudden abnormal scenarios. β is the weight of the voltage deviation value, indicating the degree of influence of grid voltage fluctuations on the identification of sudden abnormal scenarios. γ is the weight of the frequency deviation value, indicating the degree of influence of grid frequency fluctuations on the identification of sudden abnormal scenarios;

[0050] After identifying an abnormal scenario, calculate the critical load priority power supply factor to determine the power supply ratio of the loads to be prioritized under abnormal conditions. The calculation expression is as follows:

[0051]

[0052] , where P critical is the critical load priority power supply factor, W critical is the total power demand of the critical loads, W total is the total power demand of all loads, and λ is an adjustment coefficient used to control the priority power supply weight of the critical loads under abnormal scenarios;

[0053] According to the critical load priority power supply factor P critical and the current state of charge of the current energy storage system, calculate the discharge rate adjustment parameter of the energy storage system and dynamically adjust the discharge strategy of the energy storage system to ensure the stable operation of the energy storage system and minimize the electricity cost. The calculation expression is as follows:

[0054]

[0055] , where R discharge is the discharge rate adjustment parameter of the energy storage system, P max is the maximum charge and discharge power of the energy storage battery, SOC is the current state of charge of the energy storage battery, indicating the charging percentage of the battery, δ is the cost adjustment coefficient, C grid is the real-time grid electricity price, C battery is the unit cost of battery discharge.

[0056] The data-driven optical storage and charging energy allocation system includes a data acquisition and preprocessing module, a long-term energy prediction module, a load surge detection and response module, an energy storage strategy optimization module, and an abnormal identification and adaptive scheduling module:

[0057] The data acquisition and preprocessing module collects the real-time power generation of the photovoltaic power generation system, the current state of charge of the energy storage battery, and the load power consumption data in real time, and cleans, normalizes, and time-series the collected data to ensure the accuracy and continuity of the input data;

[0058] Long-term energy prediction module, which constructs a long-term energy supply and demand prediction model using a machine learning model based on historical photovoltaic power generation, charge and discharge records of the energy storage system, and load power consumption patterns, predicts future photovoltaic power generation trends and load demand changes, and plans the charge and discharge strategies of the energy storage system in advance;

[0059] Load surge detection and response module, which compares the real-time monitored load power consumption data with the predicted data, and detects short-term load surges based on the change rate threshold and time window;

[0060] Energy storage strategy optimization module, which introduces an energy storage strategy optimization algorithm, and calculates the optimal charge and discharge strategy of the energy storage system in real time according to the current photovoltaic power generation, the state of charge of the energy storage battery, and the predicted value of load demand;

[0061] Abnormal recognition and adaptive scheduling module, which uses an abnormal recognition algorithm to perform correlation analysis on the load mutation signal and the external power grid state, identifies abnormal scenarios, and based on the recognition results, executes an adaptive scheduling algorithm to give priority to ensuring the power supply demand of critical loads, and adjusts the power exchange between the discharge rate of the energy storage system and the power grid to ensure the stable operation of the system in case of emergencies and minimize the electricity cost.

[0062] In the above technical solution, the technical effects and advantages provided by the present invention:

[0063] By real-time monitoring the load power consumption data and combining it with the long-term energy supply and demand prediction model, the present invention dynamically adjusts the discharge strategy of the energy storage system, thus effectively avoiding the problem of over-discharge of the energy storage battery during off-peak periods. Compared with the traditional fixed scheduling strategy, this solution can real-time identify load surges and flexibly adjust the discharge rate of the energy storage battery according to the surge signal, ensure sufficient energy storage power during peak periods, reduce the dependence on grid power, and reduce the overall electricity cost. At the same time, the system combines the grid electricity price optimization strategy to realize the strategy of storing electricity during low electricity price periods and discharging during high electricity price periods, greatly improving the economic benefits of the energy storage system, avoiding the phenomenon of abandoned light in photovoltaic power generation, and maximizing the utilization efficiency of green energy.

[0064] By introducing a load surge detection mechanism and an adaptive scheduling algorithm, the present invention can automatically trigger the adjustment strategy of the energy storage system when detecting abnormal load changes, and quickly respond to various sudden load change scenarios, such as a surge in communication services, equipment failures, or grid outages. This automatic response mechanism ensures that the energy storage system can give priority to ensuring the power supply demand of critical loads during abnormal load fluctuations, and avoid power outages caused by premature depletion of the energy storage battery's power. At the same time, the system can flexibly adjust the discharge mode according to the duration of the surge, and adopt a step-by-step discharge or partial discharge strategy to further improve the endurance of the energy storage battery, ensure the overall stability of the system and the reliability of power supply, and provide continuous power guarantee for the user's core business. Description of the Drawings

[0065] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0066] Figure 1 It is a method flow chart of the data-driven light storage charging energy allocation control method of the present invention.

[0067] Figure 2 It is a module schematic diagram of the data-driven light storage charging energy allocation system of the present invention. Detailed Embodiments

[0068] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0069] The present invention provides a data-driven light storage charging energy allocation control method as Figure 1 shown, including the following steps:

[0070] Real-time collect the real-time power generation of the photovoltaic power generation system, the current power state of the energy storage battery, and the load power consumption data, and perform cleaning, normalization processing, and time serialization on the collected data to eliminate data outliers and missing values, ensuring the accuracy and continuity of the input data;

[0071] The specific steps of real-time collecting the real-time power generation of the photovoltaic power generation system, the current power state of the energy storage battery, and the load power consumption data, and performing cleaning, normalization processing, and time serialization on the collected data to eliminate data outliers and missing values, ensuring the accuracy and continuity of the input data are as follows:

[0072] Realize real-time data collection and transmission of photovoltaic power generation, energy storage battery power, and load power consumption through sensors and data transmission protocols to ensure the integrity and real-time nature of the data;

[0073] First, real-time data is collected through sensors from a photovoltaic power generation system, a battery management system (BMS) of an energy storage battery, and an intelligent meter at the load end. This data includes the photovoltaic power generation (in kWh), the state of charge (SOC) of the energy storage battery, and the load power consumption (in kWh).

[0074] During the data collection process, to ensure the real-time nature of the data, a data transmission protocol (such as MQTT, Modbus, or HTTP) is used to transmit the collected raw data to the data processing center, minimizing the transmission delay of the data from the collection end to the processing end. In addition, to improve the integrity of the data, the system sets the data collection frequency, such as collecting data once per minute, and automatically generates timestamps, thus laying a foundation for subsequent time series processing.

[0075] Real-time data collection is the basis of the energy allocation system, ensuring that the system can obtain the latest status information of the photovoltaic power generation system and the load end. Through the combination of sensors and data transmission protocols, real-time data transmission and remote monitoring can be achieved, avoiding data lag and incompleteness.

[0076] Clean the collected data, remove outliers and fill in missing values, and at the same time unify the data frequency through resampling to ensure the continuity and rationality of the data;

[0077] After the data is transmitted to the processing center, the system first cleans the collected raw data to remove abnormal and incorrect data. For example, during the data cleaning process, values that do not fall within the normal range (such as negative photovoltaic power generation, SOC exceeding 100%) are identified and marked as outliers. For missing values, the system can use interpolation or moving average methods to fill in the missing data to ensure data integrity. In addition, the system also processes the problem of inconsistent data collection frequencies. For example, if the data collection frequency in a certain period is too high or too low, the data is unified to a preset sampling frequency (such as once per minute or once per hour) through resampling to avoid data being too sparse or redundant.

[0078] Data cleaning can significantly improve the quality of the input data and prevent abnormal data from affecting the accuracy of subsequent prediction models. For example, in practical applications, sensors may malfunction, resulting in abnormal or missing data. Through cleaning and filling, the system can maintain the continuity and rationality of the data, thus ensuring the reliability of subsequent analysis and prediction.

[0079] Convert data from different sources to the same scale through normalization methods to avoid the numerical differences between features from affecting the accuracy of model decision-making;

[0080] To make the collected data from different sources have the same scale, the system will perform normalization on the data. The units and value ranges of photovoltaic power generation, energy storage battery state of charge (SOC), and load power consumption are different. To enable these data to be input into the same model for analysis, the system adopts the min-max normalization method or the Z-score standardization method to scale the data to the range between 0 and 1.

[0081] For example, the data range of photovoltaic power generation may be from 0 to 100 kWh, while the range of SOC is from 0% to 100%. To make these data have the same influence in the model, the normalization process converts all data to the same scale. In this way, the situation where a certain feature dominates the model decision due to a large value is avoided.

[0082] Normalization is a crucial step before the data is input into the model, avoiding the scale differences between features from causing bias in the model. Especially in a machine learning-based energy allocation system, the weights of different features will directly affect the accuracy of the allocation strategy. Therefore, normalization can improve the efficiency and accuracy of model training.

[0083] Arrange the processed data in chronological order as time series data to capture periodic and trend changes, providing a time series input basis for load prediction and scheduling strategies;

[0084] After data cleaning and normalization are completed, the system will convert the processed data into time series data. Time serialization arranges the data in chronological order to ensure that each data point corresponds to a timestamp, thus providing a time series basis for subsequent prediction and scheduling.

[0085] Specifically, the system will perform time serialization processing on the data according to the data collection time interval (such as every minute or every hour) to generate a continuous time series dataset. For example, sort the photovoltaic power generation and load power consumption data within the past 15 days by hour to form a time series chart.

[0086] The data structure of time serialization can effectively capture the periodicity and trends of the data and provide basic data input for subsequent load prediction and scheduling algorithms. At the same time, the system will identify seasonal trends and abnormal fluctuations to make more accurate predictions for future energy allocation strategies.

[0087] Time serialization processing can help the system capture the changing trends and periodic characteristics of the data. In a photovoltaic power generation system, the power generation usually has daily, weekly, or monthly periodic changes. Through time serialization processing, the system can more accurately predict future power generation trends and load demands, improving the rationality and accuracy of energy allocation.

[0088] Based on historical photovoltaic power generation, charge and discharge records of the energy storage system, and load power consumption patterns, a long-term energy supply and demand prediction model is constructed using a machine learning model to predict future photovoltaic power generation trends and load demand changes, so as to plan the charge and discharge strategies of the energy storage system in advance;

[0089] Based on historical photovoltaic power generation, charge and discharge records of the energy storage system, and load power consumption patterns, the specific steps for constructing a long-term energy supply and demand prediction model using a machine learning model to predict future photovoltaic power generation trends and load demand changes, and to plan the charge and discharge strategies of the energy storage system in advance are as follows:

[0090] Construct a time series dataset by selecting key features to provide high-quality input data for model training;

[0091] After completing data cleaning and time serialization processing, key features are selected from the real-time data of the photovoltaic power generation system, including photovoltaic power generation, state of charge (SOC) of the energy storage system, grid electricity price, weather data (such as temperature, sunshine intensity), load power consumption, etc. To improve the prediction ability of the model, the system will construct a training dataset in the format of time series data for these features, and set the input window period (such as hourly data for the past 15 days) and prediction targets (such as power generation and load demand for the next 24 hours).

[0092] In addition, the dataset also needs to be divided into a training set, a validation set, and a test set to ensure that the model performs consistently on different datasets and avoid overfitting or underfitting.

[0093] Feature selection and dataset construction are the basis for building a prediction model. By reasonably selecting key features from historical data, it can help the machine learning model more accurately capture the changing laws of photovoltaic power generation and load demand. At the same time, dividing the dataset according to time series can effectively verify the generalization ability of the model and provide high-quality data input for subsequent model training.

[0094] Select a machine learning model suitable for time series data (such as LSTM), and train the model through cross-validation and parameter adjustment to improve prediction accuracy and robustness;

[0095] According to data characteristics and prediction requirements, select a suitable machine learning model to construct a long-term energy supply and demand prediction model. Commonly used models include long short-term memory network (LSTM), recurrent neural network (RNN), and XGBoost regression model, etc. These models are good at processing time series data and can effectively capture the periodicity and trend of historical data.

[0096] During the training process, the model adjusts its parameters and weights continuously to minimize the error between the predicted values and the actual values. To improve the robustness of the model, the system adopts the cross-validation method to train and validate the model multiple times, ensuring that the model performs consistently on different datasets. Additionally, to prevent the model from over-relying on recent data, a regularization term or an early stopping mechanism can be added to prevent overfitting.

[0097] Model selection and training are the core steps in constructing a prediction model. By selecting a deep learning model suitable for time series data, the system can capture complex change patterns in historical data, thereby improving the prediction accuracy of future energy supply and demand. At the same time, technical means such as cross-validation and regularization can effectively enhance the generalization ability of the model and reduce prediction errors.

[0098] Evaluate the model performance through various metrics, and optimize the model through hyperparameter tuning and real-time evaluation to make it adapt to dynamic data changes;

[0099] After completing the initial model training, evaluate and optimize the model performance to ensure the accuracy and stability of the prediction results. Evaluation metrics usually include Mean Squared Error (MSE), Mean Absolute Error (MAE), and Coefficient of Determination (R 2 ) etc. The system judges the prediction effect of the model based on these metrics and optimizes the model performance by adjusting the hyperparameters of the model (such as learning rate, number of hidden layers, number of training steps, etc.). In addition, the system also introduces a real-time evaluation mechanism to compare and analyze the prediction results of the model with the actual results, thereby continuously correcting the parameters of the model to improve the prediction accuracy. Through this dynamic optimization method, the model can adapt to the changes in photovoltaic power generation and load demand, and improve the reliability of long-term prediction.

[0100] Model optimization and evaluation are the key steps to ensure the actual application effect of the prediction model. Through multi-dimensional metric evaluation, the system can identify the deficiencies of the model and continuously improve the model performance by adjusting hyperparameters. At the same time, the real-time evaluation mechanism enables the model to adapt to the changes in the data environment and maintain the long-term stability of the prediction effect.

[0101] Use the prediction results for the charge and discharge strategy planning of the energy storage system, and combine with electricity price optimization and abnormal load identification mechanisms to achieve optimization of electricity consumption costs and improvement of system stability;

[0102] After the model optimization is completed, the predicted future photovoltaic power generation trends and the results of load demand changes predicted by the model are applied to the charge and discharge strategy planning of the energy storage system. The system will dynamically adjust the charge and discharge time and power threshold of the energy storage battery according to the prediction results. For example, when it is predicted that the photovoltaic power generation will be sufficient in the future, the system will charge in advance and give priority to using the energy storage system to supply power during high-load periods, thereby reducing the dependence on grid power. At the same time, the system will formulate an electricity price optimization strategy according to the fluctuations of the grid electricity price, charge during the low electricity price period, and discharge during the high electricity price period to minimize the electricity cost.

[0103] In addition, to cope with special scenarios (such as equipment failures or sudden load surges), the system will combine real-time monitoring data with prediction results, trigger an abnormal load identification mechanism, and dynamically adjust the energy storage strategy to ensure that the energy storage system has sufficient power reserves in case of emergencies and guarantee the normal operation of critical loads.

[0104] The application of the prediction results is the implementation link of the entire energy allocation control, ensuring that the energy storage system can plan the charge and discharge strategy in advance according to future supply and demand situations, thereby improving the utilization rate of photovoltaic power generation and reducing the phenomenon of abandoned light. At the same time, by introducing the electricity price optimization and abnormal load identification mechanisms, more flexible energy allocation can be achieved in special scenarios, thereby reducing the overall operating cost of the system.

[0105] Compare the real-time monitored load power consumption data with the predicted data, detect short-term load surges based on the change rate threshold and time window. When it is detected that the load surge exceeds the set threshold, trigger a load mutation signal and dynamically adjust the discharge strategy of the energy storage system to avoid excessive discharge of the energy storage system during off-peak hours;

[0106] Compare the real-time monitored load power consumption data with the predicted data, detect short-term load surges based on the change rate threshold and time window. When it is detected that the load surge exceeds the set threshold, trigger a load mutation signal and dynamically adjust the discharge strategy of the energy storage system to avoid excessive discharge of the energy storage system during off-peak hours. The specific steps are as follows:

[0107] By comparing the deviation between the real-time collected load power consumption data and the predicted data, identify the situation where the actual load exceeds the normal range, providing a basis for surge detection;

[0108] The system first compares the real-time collected load power consumption data with the load prediction data generated by the long-term energy supply and demand prediction model. To ensure the accuracy of the comparison, the system will synchronize the real-time data and the predicted data at the same time interval (such as every hour or every minute) to ensure that both have consistent timestamps.

[0109] During comparison, the system calculates the deviation value between the actual load data and the predicted load data, that is, subtracting the predicted data from the real-time data to obtain a deviation curve. If the deviation value continues to increase, it indicates that the situation where the actual load exceeds the predicted load is intensifying, and there may be a risk of sudden increase. To improve the sensitivity of the system, the system also compares the deviation change trends within multiple time windows, such as the load change situations in the past 5 minutes, 15 minutes, or 1 hour.

[0110] The comparison between real-time data and predicted data is the basis for sudden increase detection. By calculating the deviation value between the two, the system can quickly identify the situation where the load exceeds the normal range. This comparison method effectively utilizes the results of the long-term energy supply and demand prediction model, avoiding the problem that the system only relies on real-time data and ignores trend changes, thereby improving the accuracy of detection.

[0111] By calculating the change rate and fluctuation amplitude of the load data, the trend and instability of the load change within a short period are monitored, so as to early warn of the sudden increase risk;

[0112] After the comparison is completed, the system calculates the change rate of the deviation value, that is, calculates the change speed of the load data within a certain time window. The calculation formula for the change rate is: For example, if the load changes by 10 kW in the past 5 minutes, the change rate is 2 kW / min. The system will set multiple time windows (such as 5 minutes, 15 minutes, 30 minutes) to monitor the change rates in different time periods and analyze the fluctuation trend of the change rate. If the change rate rapidly increases within a short period, it indicates that there is a short-term sudden increase in the load. In addition, the system also introduces a fluctuation analysis algorithm to calculate the fluctuation amplitude of the change rate. If the fluctuation amplitude exceeds a certain threshold, it indicates that the system load is in an unstable state, and at this time, it is necessary to further detect the load mutation situation.

[0113] The change rate calculation and fluctuation analysis can effectively identify the sudden increase trend of the load. Compared with the simple comparison of load data, the change rate can more sensitively capture the rapid change of the load, thereby early warning of the sudden increase risk. At the same time, the fluctuation analysis of multiple time windows can effectively avoid false alarms caused by short-term noise or data anomalies, improving the accuracy and robustness of detection.

[0114] Compare the change rate with a preset threshold. When the change rate exceeds the set value, trigger a load mutation signal and verify it multiple times to improve the accuracy;

[0115] Based on the analysis results of the change rate and fluctuation amplitude, the system compares them with a preset change rate threshold. If the change rate exceeds the set threshold (for example, exceeds 5 kW / min), the system will trigger a load mutation signal, indicating that there is an abnormal sudden increase in the current load.

[0116] The threshold setting is usually based on the statistical analysis results of historical load data and is adjusted in combination with the importance of base station services and the stability of power grid power supply. For example, in the scenario of a sharp increase in communication services, a lower threshold can be set to detect sudden increases more sensitively. During normal business operations, a higher threshold can be set to avoid false triggering of signals. In addition, the system will also verify the mutation signal multiple times to ensure that it is not a misjudgment caused by short-term fluctuations or data noise.

[0117] Threshold judgment is the core step for triggering load mutation signals, directly determining the response speed and accuracy of the system to sudden load increases. A reasonable threshold setting can effectively balance the sensitivity and stability of the system, avoiding false alarms or missed reports. At the same time, the multiple verification mechanism can further improve the reliability of mutation signals, ensuring that the system has a high fault tolerance in practical applications.

[0118] After triggering the load mutation signal, the discharge strategy of the energy storage system is dynamically adjusted according to the photovoltaic power generation, the state of charge of the energy storage battery, and the load demand to ensure continuous and stable power supply to the load;

[0119] After the load mutation signal is triggered, the system will dynamically adjust the discharge strategy of the energy storage system according to the current photovoltaic power generation, the state of charge (SOC) of the energy storage battery, and the grid electricity price. Specifically, the system will first reduce the discharge rate of the energy storage battery to ensure that the energy storage system has sufficient power reserve during off-peak periods.

[0120] At the same time, the system will also adjust the discharge strategy according to the duration of the sudden load increase and the availability of grid power supply. For example, if the sudden load increase lasts for a short time, the system will adopt a partial discharge mode to only meet the power consumption needs of critical loads; if the sudden load increase lasts for a long time, the system will adjust the discharge rate to avoid premature depletion of the battery power. In addition, when the sudden load increase exceeds a certain threshold, the system can also switch to the grid power supply mode to ensure continuous power supply to the load.

[0121] Dynamically adjusting the discharge strategy of the energy storage system is the ultimate goal of load mutation detection, aiming to ensure the stable operation of the system in case of emergencies. Through dynamic discharge adjustment, the system can effectively avoid over-discharge of the energy storage system during off-peak periods, thus ensuring that the energy storage battery has sufficient power reserve during peak periods, improving the overall optimization effect of electricity costs and the reliability of the system.

[0122] Introduce an energy storage strategy optimization algorithm to calculate the optimal charge and discharge strategy of the energy storage system in real time according to the current photovoltaic power generation, the state of charge of the energy storage battery, and the predicted load demand. This algorithm comprehensively considers the grid electricity price, changes in load demand, and the health status of the energy storage battery to achieve dynamic scheduling among photovoltaic power generation, energy storage, and grid power.

[0123] Introduce an energy storage strategy optimization algorithm. According to the current photovoltaic power generation, the state of charge of the energy storage battery, and the predicted value of the load demand, calculate the optimal charge and discharge strategy of the energy storage system in real time. This algorithm comprehensively considers the grid electricity price, the change of load demand, and the health state of the energy storage battery, and the specific steps to achieve the dynamic scheduling among photovoltaic power generation, energy storage, and grid power are as follows:

[0124] First, calculate the current load demand gap, that is, the difference between the actual load demand and the power provided by photovoltaic power generation, which represents the power that the energy storage system and the grid need to supplement, and is used to judge whether the energy storage system needs to charge or discharge. The calculation formula is as follows:

[0125] ΔP load =P demand -P solar

[0126] , where ΔP load is the load demand gap, that is, the load power demand that cannot be met by photovoltaic power generation, P demand is the actual load demand, obtained by the real-time data acquisition system, P solar is the current photovoltaic power generation, which is collected in real time by the sensors of the photovoltaic power generation system;

[0127] When ΔP load > 0, it means that the photovoltaic power generation is insufficient and power needs to be supplemented from the energy storage system or the grid; when ΔP load < 0, it means that there is a surplus of photovoltaic power generation and the energy storage system can be charged.

[0128] According to the calculated load demand gap ΔP load , further calculate the charge and discharge power of the energy storage battery. This step considers the current state of charge of the energy storage battery and the maximum charge and discharge power to ensure that the charge and discharge of the battery are within a reasonable range. The calculation formula is as follows:

[0129]

[0130] , where P battery is the charge and discharge power of the energy storage battery, a positive value indicates discharge, and a negative value indicates charge, P max is the maximum charge and discharge power of the energy storage battery, max(ΔP load , -P max ) is the maximum value limit, indicating that the load gap value shall not be less than the negative value of the maximum discharge power of the energy storage battery, min(max(ΔP load , -P max ), P max ) is the minimum value limit, indicating that the load gap value shall not exceed the maximum charge power of the energy storage battery, SOC is the current state of charge of the energy storage battery, indicating the charging percentage of the battery;

[0131] This step ensures that the charging and discharging power of the energy storage battery does not exceed its maximum limit, and at the same time dynamically adjusts the charging and discharging power according to the SOC state of the battery. As the SOC increases, the charging ability of the battery gradually weakens, thus avoiding the risk of overcharging.

[0132] To ensure the long-term service life of the energy storage battery, comprehensively consider the impact of the battery's health state on the charging and discharging power, which is used to measure the remaining life of the battery. The lower the health state, the smaller the available charging and discharging capacity of the battery. The calculation formula is as follows:

[0133]

[0134] , where P battery_adjusted is the charging and discharging power of the energy storage battery corrected by the health state, SOH is the health state of the energy storage battery, indicating the percentage of the remaining life of the battery;

[0135] This step dynamically adjusts the charging and discharging power of the energy storage battery to reduce the overuse of batteries with poor health states and extend the service life of the batteries.

[0136] When the energy storage system cannot fully meet the load demand, the dispatching decision of the power grid will be calculated by combining the real-time grid electricity price and the charging and discharging power of the energy storage battery, and it will be decided whether to purchase electricity from the grid or sell electricity to the grid. The calculation formula is as follows:

[0137] P grid = ΔP load - P battery_adjusted

[0138] , where P grid is the grid dispatching power, indicating the power of purchasing electricity from the grid or selling electricity to the grid;

[0139] If P grid > 0, then purchase electricity from the grid;

[0140] If P grid < 0, then sell electricity to the grid;

[0141] The grid electricity price is adjusted according to the real-time electricity price. The calculation formula is as follows:

[0142] C cost = P grid × C grid

[0143] , where C grid is the real-time grid electricity price, and C cost is the cost of purchasing or selling electricity from the grid.

[0144] This step combines the grid electricity price to adjust the grid dispatching strategy in real time, achieving the minimization of electricity consumption cost. When the grid electricity price is low, the system tends to purchase electricity from the grid; when the grid electricity price is high, the system tends to use the energy storage battery to supply power or sell electricity to the grid to obtain benefits.

[0145] Through the anomaly recognition algorithm, the load mutation signal is correlated with the external grid state for analysis to identify abnormal scenarios. Based on the recognition results, the adaptive dispatching algorithm is executed to prioritize ensuring the power supply demand of critical loads, and by adjusting the discharge rate of the energy storage system and the power exchange with the grid, the stable operation of the system in case of emergencies and the minimization of electricity consumption cost are ensured.

[0146] The specific steps for ensuring the stable operation of the system in case of emergencies and the minimization of electricity consumption cost through the anomaly recognition algorithm, correlating the load mutation signal with the external grid state for analysis to identify abnormal scenarios, and executing the adaptive dispatching algorithm based on the recognition results to prioritize ensuring the power supply demand of critical loads, and by adjusting the discharge rate of the energy storage system and the power exchange with the grid are as follows:

[0147] First, by correlating the real-time monitored load sudden increase signal with the external grid state data (such as grid frequency, voltage fluctuation, power supply interruption information, etc.), the load mutation intensity parameter is calculated to quantify the degree of the current abnormal load change. The calculation expression of the load mutation intensity parameter is as follows:

[0148]

[0149] , where, I LT is the load mutation intensity parameter, ΔL is the load change amount within the time interval Δt, Δt is the time interval, V dev is the deviation value of the grid voltage, that is, the difference between the actual voltage and the standard voltage, F dev is the deviation value of the grid frequency, that is, the difference between the actual frequency and the standard frequency, and α, β, and γ are all weight parameters. α is the weight of the load change rate, indicating the influence degree of the load change on the identification of sudden abnormal scenarios. If the load fluctuation is the main cause of the sudden anomaly, increase the value of α. For example, in a communication base station, the load fluctuation is usually a key abnormal signal. β is the weight of the voltage deviation value, indicating the influence degree of the grid voltage fluctuation on the identification of sudden abnormal scenarios. When the unstable grid voltage is the main cause of the system anomaly, increase the value of β. For example, in an industrial electricity consumption scenario, the voltage fluctuation may cause equipment damage. γ is the weight of the frequency deviation value, indicating the influence degree of the grid frequency fluctuation on the identification of sudden abnormal scenarios. When the unstable grid frequency is the main signal of the abnormal scenario, increase the value of γ. For example, in a large-scale power grid, the frequency fluctuation is usually an indication of the grid load imbalance;

[0150] By calculating the load mutation intensity parameter I LT , the system can comprehensively consider the changes in the load sudden increase signal and the external power grid status, and identify the intensity of the current abnormal scenario. For example, when the load suddenly increases accompanied by fluctuations in the grid voltage and frequency, the system will judge it as a grid fault or a load abnormal scenario and enter the adaptive scheduling mode.

[0151] After identifying the abnormal scenario, calculate the critical load priority power supply factor to determine the power supply proportion of the load to be guaranteed first in the abnormal situation. The calculation formula is as follows:

[0152]

[0153] , where P critical is the critical load priority power supply factor, W critical is the total power demand of the critical load, W total is the total power demand of all loads, and λ is the adjustment coefficient used to control the priority power supply weight of the critical load in the abnormal scenario;

[0154] By calculating P critical , the system dynamically adjusts the power supply priority of the critical load. The more severe the abnormal scenario (i.e., the higher I LT ), the greater the priority power supply proportion of the system to the critical load, so as to ensure the continuous operation of critical services in case of power shortage or grid failure. For example, in a communication base station, the critical load may include communication core equipment, network transmission equipment, etc.

[0155] According to the critical load priority power supply factor P critical and the current state of charge of the energy storage system, calculate the discharge rate adjustment parameter of the energy storage system and dynamically adjust the discharge strategy of the energy storage system to ensure the stable operation of the energy storage system and minimize the power consumption cost. The calculation formula is as follows:

[0156]

[0157] , where R discharge is the discharge rate adjustment parameter of the energy storage system, P max is the maximum charge and discharge power of the energy storage battery, SOC is the current state of charge of the energy storage battery, indicating the charging percentage of the battery, δ is the cost adjustment coefficient used to balance the grid power supply cost and the battery discharge cost, C grid is the real-time grid electricity price, and C battery is the unit cost of battery discharge.

[0158] By calculating R discharge, the system can dynamically adjust the discharge rate of the energy storage system according to the severity of the abnormal scenario and the current state of the energy storage system. When the electricity price is high or there is a power grid fault, the system will give priority to using the energy storage battery for power supply; while when the electricity price is low, it will give priority to reducing the battery discharge, reducing the loss of the energy storage system and lowering the electricity cost.

[0159] Embodiment 1: In the photovoltaic energy storage charging system, in order to avoid the risk of over-discharge of the energy storage battery during the off-peak period, a dynamic energy storage management scheme based on short-term load surge detection is proposed. This scheme takes real-time data acquisition, load surge detection and dynamic discharge adjustment as the core, and realizes a fast response to abnormal load changes by continuously monitoring and adjusting the discharge strategy of the energy storage system, thereby ensuring the efficient operation and stable power supply of the system.

[0160] The system collects real-time photovoltaic power generation, the state of charge (SOC) of the energy storage battery and load power consumption data from the photovoltaic power generation system, the energy storage battery management system (BMS) and the smart meter at the load end. After being cleaned, normalized and time-serialized, these data are input into the long-term energy supply and demand prediction model to generate predicted values of future photovoltaic power generation trends and load demand changes. During actual operation, the system will compare the real-time collected load data with the predicted data and calculate the deviation value between the two. If the deviation value continues to increase and the change rate exceeds the preset threshold, it means that there is an abnormal fluctuation in the load power consumption situation, that is, a short-term load surge.

[0161] The system introduces change rate calculation and fluctuation analysis algorithms to dynamically monitor the change of real-time data. By setting a time window (such as the past 5 minutes, 15 minutes, 30 minutes), the change rate of the load data is calculated to identify the fluctuation trend of the load. If the change rate rises significantly within a short period of time, it means that there is a risk of load surge. In order to avoid false alarms caused by short-term noise or occasional events, the system will also conduct multiple verifications and analyze the fluctuation amplitude of the change rate to ensure the reliability of the load surge signal.

[0162] When the system detects that the load surge exceeds the threshold, it immediately triggers a load mutation signal and activates the dynamic discharge adjustment strategy of the energy storage system. During the adjustment process, the system will dynamically adjust the discharge rate of the energy storage system according to the current photovoltaic power generation, the state of charge of the energy storage battery and the load demand situation. For example, when a surge occurs, the system will give priority to reducing the power supply to non-critical loads and reserve more power for critical loads. At the same time, the system will reduce the discharge rate of the energy storage battery to extend the power supply time of the battery and avoid premature depletion of the battery power.

[0163] The advantages of this implementation mode lie in its strong real-time performance, high sensitivity, and strong adaptability. By monitoring the load change situation in real time, the system can quickly respond to sudden load increases and avoid over-discharging problems of energy storage batteries by dynamically adjusting the discharge strategy. At the same time, this solution effectively improves the stability and reliability of the system, ensuring that the base station and key equipment can operate continuously and stably in the event of sudden load changes.

[0164] Embodiment 2: To reduce the overall electricity cost and improve the economic benefits of the energy storage system, an energy allocation strategy combined with grid electricity price optimization is proposed. This solution dynamically adjusts the energy flow between photovoltaic power generation, energy storage, and grid power, rationally utilizes the electricity price fluctuations of the grid, and realizes the optimization of electricity costs and the efficient utilization of energy storage batteries.

[0165] The system will access the real-time electricity price data interface of the grid to obtain the electricity price information of the current period, including peak-valley electricity prices, time-of-use electricity prices, and dynamic electricity prices. A electricity price change model is established within the system. By analyzing historical electricity price data and current electricity price fluctuation trends, the future electricity price change situation is predicted. For example, the system can identify the low-valley electricity price period and peak electricity price period of each day and mark these periods as the charge and discharge time windows of the energy storage system.

[0166] According to the electricity price change model, the system will dynamically adjust the charge and discharge strategy of the energy storage system. During the low-valley electricity price period, the system will give priority to using grid power to charge the energy storage battery to maximize the power reserve of the energy storage battery; while during the peak electricity price period, the system will give priority to using the energy storage battery to supply power to the load, thereby reducing the use of high-price grid power. This strategy effectively reduces the overall electricity cost of the system, improves the utilization rate of photovoltaic power generation at the same time, and avoids the occurrence of light abandonment phenomenon.

[0167] In the case of sudden load increase, the system will first check the power status of the energy storage battery. If the power of the energy storage battery is insufficient to meet the current load demand, the system will automatically switch to the grid power supply mode to avoid power supply interruption caused by insufficient power. At the same time, the system will select an appropriate grid power period according to the electricity price model to minimize the use of high-price power and reduce the electricity cost.

[0168] The advantages of this solution lie in its high economy, strong flexibility, and high degree of intelligence. By introducing the electricity price optimization strategy, the system can rationally allocate energy during different electricity price periods to minimize the electricity cost. At the same time, in the event of sudden load changes, the system can flexibly switch to grid power supply to ensure the continuous and stable power supply of key loads, improving the reliability and economic benefits of the overall system.

[0169] Embodiment 3: To further improve the intelligence level and adaptive ability of the energy storage system, an optimization scheme for the charge and discharge strategy of an adaptive energy storage system based on machine learning is proposed. This scheme analyzes and learns historical data through machine learning algorithms, continuously optimizes the charge and discharge strategy of the energy storage system, and realizes adaptive adjustment to different scenarios.

[0170] The system uses the photovoltaic power generation, the state of charge of the energy storage battery, the load power consumption data, etc. collected and stored for a long time, and uses machine learning algorithms (such as LSTM, XGBoost, etc.) to model and analyze the historical data to identify the laws of energy supply and demand changes in different scenarios. For example, the system can identify the power generation laws in different seasons, the power consumption patterns in different time periods, etc., so as to adjust the strategy of the energy storage system in advance.

[0171] Based on the analysis results of the model, the system can generate an adaptive charge and discharge strategy. During the period when the photovoltaic power generation is sufficient, the system will charge the energy storage battery in advance according to the prediction results; during the peak load demand period, the system will give priority to using the energy storage battery for power supply to reduce the dependence on grid power. At the same time, when there is a sudden load change, the system can dynamically adjust the discharge strategy to ensure that the power of the energy storage battery can continuously meet the needs of critical loads.

[0172] Through the anomaly detection function of the machine learning model, the system can identify abnormal situations such as equipment failures, external grid interruptions, and surges in communication services, and automatically adjust the strategy of the energy storage system according to the identification results. For example, when a grid interruption is detected, the system will automatically switch to the energy storage power supply mode and adjust the discharge rate to extend the power supply time to ensure the normal operation of critical equipment.

[0173] The advantages of this scheme are high intelligence, strong adaptability, and accurate prediction. Through the continuous optimization of machine learning algorithms, the system can flexibly adjust the energy storage strategy in different scenarios, realize the efficient utilization of photovoltaic power generation, and provide reliable power protection during abnormal load fluctuations. At the same time, the adaptive charge and discharge strategy can effectively reduce the energy waste of the energy storage system, improve the overall economic benefits and stability of the system.

[0174] The present invention effectively avoids the problem of over-discharge of the energy storage battery during off-peak periods by dynamically adjusting the discharge strategy of the energy storage system through real-time monitoring of the load power consumption data and combining it with a long-term energy supply and demand prediction model. Compared with the traditional fixed scheduling strategy, this scheme can real-time identify sudden increases in the load and flexibly adjust the discharge rate of the energy storage battery according to the sudden increase signal to ensure sufficient energy storage power during peak periods, reduce the dependence on grid power, and lower the overall power consumption cost. At the same time, by combining the grid electricity price optimization strategy, the system realizes the strategy of storing electricity during low electricity price periods and discharging during high electricity price periods, greatly improving the economic benefits of the energy storage system, avoiding the phenomenon of abandoned photovoltaic power generation, and maximizing the utilization efficiency of green energy.

[0175] By introducing a load surge detection mechanism and an adaptive scheduling algorithm, the present invention can automatically trigger the adjustment strategy of the energy storage system when detecting abnormal load changes, and quickly respond to various sudden load change scenarios, such as a sharp increase in communication services, equipment failures, or power grid outages. This automated response mechanism ensures that the energy storage system can prioritize the power supply needs of critical loads during abnormal load fluctuations, avoiding power outages caused by premature depletion of the energy storage battery. At the same time, the system can flexibly adjust the discharge mode according to the duration of the surge, adopting a step-by-step discharge or partial discharge strategy to further improve the endurance of the energy storage battery and ensure the overall stability of the system and the reliability of power supply, providing continuous power support for the user's core business.

[0176] The invention provides Figure 2 the data-driven optical storage and charging energy allocation system shown in the figure, including a data acquisition and preprocessing module, a long-term energy prediction module, a load surge detection and response module, an energy storage strategy optimization module, and an anomaly identification and adaptive scheduling module:

[0177] The data acquisition and preprocessing module collects real-time power generation of the photovoltaic power generation system, the current power state of the energy storage battery, and load power consumption data in real time, and cleans, normalizes, and time-series the collected data to eliminate data outliers and missing values, ensuring the accuracy and continuity of the input data;

[0178] The long-term energy prediction module constructs a long-term energy supply and demand prediction model using a machine learning model based on historical photovoltaic power generation, charge and discharge records of the energy storage system, and load power consumption patterns, predicting future photovoltaic power generation trends and load demand changes to plan the charge and discharge strategy of the energy storage system in advance;

[0179] The load surge detection and response module compares the real-time monitored load power consumption data with the predicted data, detects short-term load surges based on the change rate threshold and time window. When it detects that the load surge exceeds the set threshold, it triggers a load mutation signal and dynamically adjusts the discharge strategy of the energy storage system to avoid over-discharging of the energy storage system during off-peak hours;

[0180] The energy storage strategy optimization module introduces an energy storage strategy optimization algorithm, and calculates the optimal charge and discharge strategy of the energy storage system in real time according to the current photovoltaic power generation, the power state of the energy storage battery, and the predicted load demand. This algorithm comprehensively considers the grid electricity price, load demand changes, and the health state of the energy storage battery to achieve dynamic scheduling among photovoltaic power generation, energy storage, and grid power;

[0181] The anomaly recognition and adaptive scheduling module, through the anomaly recognition algorithm, correlates the load mutation signal with the external power grid status for associated analysis to identify abnormal scenarios. Based on the recognition results, it executes the adaptive scheduling algorithm to prioritize ensuring the power supply demand of critical loads, and adjusts the power exchange between the energy storage system's discharge rate and the power grid to ensure the stable operation of the system and minimize the electricity cost in case of emergencies.

[0182] The data-driven light-storage-charging energy allocation control method provided by the embodiments of the present invention is implemented through the above-mentioned data-driven light-storage-charging energy allocation system. The specific methods and processes of the data-driven light-storage-charging energy allocation system are detailed in the embodiments of the above-mentioned data-driven light-storage-charging energy allocation control method, and will not be elaborated here.

[0183] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0184] Only some exemplary embodiments of the present invention have been described by way of illustration above. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

[0185] It should be noted that in this article, if there are relational terms such as first and second, they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article or device including the element.

[0186] It should be understood that in various embodiments of the present application, the magnitude of the serial numbers of the above processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0187] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0188] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0189] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0190] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0191] As described above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0192] Only some exemplary embodiments of the present invention have been described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. A data-driven solar energy storage and charging energy allocation control method, characterized in that: The following steps are involved: Collect the real-time power generation of the photovoltaic power generation system, the current power status of the energy storage battery and the load power consumption data in real time, and clean, normalize and time-series the collected data to ensure the accuracy and continuity of the input data; Based on the historical photovoltaic power generation, the charging and discharging records of the energy storage system, and the load power consumption pattern, a machine learning model is used to build a long-term energy supply and demand forecasting model to predict future photovoltaic power generation trends and load demand changes, so as to plan the charging and discharging strategy of the energy storage system in advance; Compare the real-time monitored load power consumption data with the predicted data, and detect short-term load surges based on the change rate threshold and time window; Introducing an energy storage strategy optimization algorithm to calculate the optimal charging and discharging strategy of the energy storage system in real time based on the current photovoltaic power generation, energy storage battery power status and load demand forecast value; Through the abnormal recognition algorithm, the load mutation signal is correlated with the external power grid status and analyzed to identify abnormal scenarios. Based on the recognition results, an adaptive scheduling algorithm is executed to prioritize the power supply needs of key loads and ensure the stable operation of the system and minimize electricity costs in emergency situations by adjusting the discharge rate of the energy storage system and the power exchange between the power grid.

2. The data-driven photovoltaic energy storage and charging allocation control method according to claim 1 is characterized in that: The real-time power generation of the photovoltaic power generation system, the current power status of the energy storage battery and the load power consumption data are collected in real time, and the collected data are cleaned, normalized and time-series to ensure the accuracy and continuity of the input data. The specific steps are as follows: Realize real-time data collection and transmission of photovoltaic power generation, energy storage battery power and load power consumption through sensors and data transmission protocols to ensure data integrity and real-time performance; Clean the collected data, remove outliers and supplement missing values, and unify the data frequency by resampling to ensure the continuity and rationality of the data; Convert data from different sources to the same scale through normalization to avoid the impact of numerical differences between features on the accuracy of model decisions; The processed data is arranged in chronological order as time series data to capture periodic and trend changes, providing a timing input basis for load forecasting and scheduling strategies.

3. The data-driven photovoltaic energy storage and charging allocation control method according to claim 1 is characterized in that: Based on the historical photovoltaic power generation, the charging and discharging records of the energy storage system, and the load power consumption pattern, a machine learning model is used to build a long-term energy supply and demand forecasting model to predict future photovoltaic power generation trends and load demand changes, so as to plan the charging and discharging strategy of the energy storage system in advance. The specific steps are as follows: By selecting key features to construct a time series dataset, high-quality input data is provided for model training; Select a machine learning model suitable for time series data and train the model through cross-validation and parameter adjustment to improve prediction accuracy and robustness; Evaluate model performance through various indicators, and optimize the model through hyperparameter adjustment and real-time evaluation to adapt to dynamic data changes; The prediction results are used to plan the charging and discharging strategies of the energy storage system, and combined with electricity price optimization and abnormal load identification mechanisms to achieve electricity cost optimization and improve system stability.

4. The data-driven photovoltaic energy storage and charging allocation control method according to claim 1 is characterized in that: The specific steps for comparing the real-time monitored load power consumption data with the predicted data and detecting short-term load surges based on the change rate threshold and time window are as follows: By comparing the deviation between the real-time collected load power consumption data and the predicted data, the actual load is identified to exceed the normal range, providing a basis for sudden increase detection; By calculating the change rate and fluctuation range of load data, the trend and instability of load changes in a short period of time can be monitored, thereby providing early warning of sudden increase risks; Compare the rate of change with the preset threshold. When the rate of change exceeds the set value, a load mutation signal is triggered and verified multiple times to improve accuracy. After the load mutation signal is triggered, the discharge strategy of the energy storage system is dynamically adjusted according to the photovoltaic power generation, the power status of the energy storage battery and the load demand to ensure continuous and stable power supply to the load.

5. The data-driven photovoltaic energy storage and charging allocation control method according to claim 1 is characterized in that: The energy storage strategy optimization algorithm is introduced. The specific steps of calculating the optimal charging and discharging strategy of the energy storage system in real time based on the current photovoltaic power generation, energy storage battery power status and load demand forecast value are as follows: First, the current load demand gap is calculated, that is, the difference between the actual load demand and the power provided by photovoltaic power generation, which is used to determine whether the energy storage system needs to be charged or discharged. The calculation expression is as follows: ΔP load =P demand -P solar In the formula, ΔP load is the load demand gap, P demand is the actual load demand, P solar is the current photovoltaic power generation; Based on the calculated load demand gap ΔP load , further calculate the charging and discharging power of the energy storage battery, the calculation expression is as follows: Where P battery is the charging and discharging power of the energy storage battery, P max is the maximum charge and discharge power of the energy storage battery, max(ΔP load , -P max ) is the maximum value limit, indicating that the load gap value must not be less than the negative value of the maximum discharge power of the energy storage battery, min(max(ΔP load , -P max ), P max ) is the minimum value limit, which means that the load gap value must not exceed the maximum charging power of the energy storage battery. SOC is the current power state of the energy storage battery, which indicates the charging percentage of the battery.

6. The data-driven photovoltaic energy storage and charging allocation control method according to claim 5 is characterized in that: In order to ensure the long-term service life of the energy storage battery, the influence of the battery health status on the charging and discharging power is comprehensively considered, and the calculation expression is as follows: Where P battery_adjusted It is the charge and discharge power of the energy storage battery after correction of the health status, SOH is the health status of the energy storage battery; When the energy storage system cannot fully meet the load demand, the grid dispatch decision will be calculated based on the real-time grid electricity price and the charging and discharging power of the energy storage battery to choose whether to purchase electricity from the grid or sell electricity to the grid. The calculation expression is as follows: P grid =ΔP load -P battery_adjusted Where P grid is the grid dispatching power; If P grid >0, then purchase electricity from the grid; If P grid <0, then sell electricity to the grid; The grid electricity price is adjusted according to the real-time electricity price, and the calculation formula is as follows: C cost =P grid ×C grid In the formula, C grid is the real-time grid electricity price, C cost It is the cost of purchasing or selling electricity to the power grid.

7. The data-driven photovoltaic energy storage and charging allocation control method according to claim 1 is characterized in that: Through the abnormal recognition algorithm, the load mutation signal is correlated with the external power grid state to analyze and identify abnormal scenarios. Based on the recognition results, the adaptive scheduling algorithm is executed to prioritize the power supply needs of key loads and adjust the discharge rate of the energy storage system and the power exchange between the power grid to ensure the stable operation of the system in emergency situations and minimize the electricity cost. The specific steps are as follows: First, the load sudden change intensity parameter is calculated by correlating the real-time monitored load surge signal with the external power grid status data to quantify the degree of abnormal load change. The calculation expression of the load sudden change intensity parameter is as follows: In the formula, I LT is the load mutation intensity parameter, ΔL is the load change within the time interval Δt, Δt is the time interval, V dev is the deviation value of the grid voltage, that is, the difference between the actual voltage and the standard voltage, F dev is the deviation value of the grid frequency, that is, the difference between the actual frequency and the standard frequency. α, β and γ are all weight parameters. α is the weight of the load change rate, which indicates the influence of the load change on the recognition of sudden abnormal scenes. β is the weight of the voltage deviation value, which indicates the influence of the grid voltage fluctuation on the recognition of sudden abnormal scenes. γ is the weight of the frequency deviation value, which indicates the influence of the grid frequency fluctuation on the recognition of sudden abnormal scenes. After identifying the abnormal scenario, calculate the priority power supply factor of the critical load to determine the proportion of load power supply that is prioritized in abnormal situations. The calculation expression is as follows: Where P critical is the priority power supply factor for critical loads, W critical is the total power demand of the critical loads, W total is the sum of the power requirements of all loads, and λ is the adjustment coefficient, which is used to control the priority power supply weight of key loads in abnormal scenarios; Prioritize power supply factor P according to key loads critical And the current power state of the energy storage system, calculate the discharge rate adjustment parameters of the energy storage system, and dynamically adjust the discharge strategy of the energy storage system to ensure the stable operation of the energy storage system and minimize the electricity cost. The calculation expression is as follows: In the formula, R discharge is the discharge rate adjustment parameter of the energy storage system, P max is the maximum charge and discharge power of the energy storage battery, SOC is the current state of charge of the energy storage battery, indicating the battery charging percentage, δ is the cost adjustment coefficient, C grid is the real-time grid electricity price, C battery is the unit cost of battery discharge.

8. A data-driven photovoltaic energy storage and charging allocation system, used to implement the data-driven photovoltaic energy storage and charging allocation control method according to any one of claims 1 to 7, characterized in that: It includes data acquisition and preprocessing module, long-term energy prediction module, load surge detection and response module, energy storage strategy optimization module and abnormality identification and adaptive scheduling module: The data collection and preprocessing module collects the real-time power generation of the photovoltaic power generation system, the current power status of the energy storage battery, and the load power consumption data in real time, and cleans, normalizes and time-series the collected data to ensure the accuracy and continuity of the input data; The long-term energy prediction module uses a machine learning model to build a long-term energy supply and demand prediction model based on historical photovoltaic power generation, charging and discharging records of the energy storage system, and load power consumption patterns. It predicts future photovoltaic power generation trends and load demand changes, so as to plan the charging and discharging strategies of the energy storage system in advance. The load surge detection and response module compares the real-time monitored load power consumption data with the predicted data, and detects short-term load surges based on the change rate threshold and time window; The energy storage strategy optimization module introduces the energy storage strategy optimization algorithm to calculate the optimal charging and discharging strategy of the energy storage system in real time based on the current photovoltaic power generation, energy storage battery power status and load demand forecast value; The anomaly identification and adaptive scheduling module uses an anomaly identification algorithm to correlate and analyze load mutation signals with the external grid status, identify abnormal scenarios, and execute an adaptive scheduling algorithm based on the identification results to prioritize the power supply needs of key loads. It also ensures the stable operation of the system and minimizes electricity costs in emergency situations by adjusting the discharge rate of the energy storage system and the power exchange between the grid.

Images