A charging and discharging control method suitable for power distribution network energy storage equipment

Abstract

The application discloses a kind of charging and discharging control methods suitable for power distribution network energy storage equipment, and the application relates to power distribution network technical field.The steps of the method include: collecting power distribution network historical load data and meteorological data, extracting and integrating load characteristics to form a sequence;Construct a load-meteorological dual-dimension model, input the characteristic sequence and meteorological data, and output the future 24-hour hourly load prediction value;According to the prediction value, calculate the load change rate and proportion of each period, and divide the flexible control period;According to the load characteristics of each period, match the charging and discharging strategy, combine to form the initial period strategy;Collect energy storage equipment state data, which includes state of charge, battery health state, battery temperature and cumulative cycle number, calculate health index and operating status index and synthesize state comprehensive index, and generate correction rule to optimize the initial strategy to obtain the correction period strategy.The application realizes the charging and discharging control of power distribution network energy storage equipment through the correction period strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power distribution network engineering, in particular to a charging and discharging control method suitable for power distribution network energy storage equipment. BACKGROUND

[0002] In the operation of the power distribution network, the charging and discharging control of the energy storage equipment is like a key hub. It can adjust power supply and demand in real time, smooth power fluctuations, and greatly improve power quality; it can ensure power supply during power grid failure or peak period and enhance power supply reliability; it can also efficiently store off-peak power and release it at peak times, optimizing energy utilization efficiency and helping the power grid to reduce costs and increase efficiency and green transformation, which is the core element of realizing the safe, economic and efficient operation of the intelligent power distribution network.

[0003] However, the existing energy storage control method generally lacks deep fusion analysis of the state of the equipment, and the state of charge, battery health, battery temperature and cumulative cycle number are not effectively involved in the strategy making, which is easy to cause excessive charging and discharging or state monitoring lag, and then affect the service life and operation safety of the energy storage equipment. SUMMARY

[0004] In view of the shortcomings of the prior art, the present application provides a charging and discharging control method suitable for power distribution network energy storage equipment, which solves the problem of lack of deep fusion analysis of the state of the equipment in the existing energy storage control method.

[0005] To achieve the above purpose, the present application is implemented by the following technical scheme: a charging and discharging control method suitable for power distribution network energy storage equipment, comprising the following steps:

[0006] Step S1: Collecting historical load data of the power distribution network, extracting and integrating the features of the historical load data of the power distribution network to obtain a load feature sequence; collecting meteorological data of the power distribution network;

[0007] Step S2: Constructing a load-meteorological dual-dimension prediction model, inputting the load feature sequence and the meteorological data of the power distribution network into the load-meteorological dual-dimension prediction model, and outputting a 24-hour hourly load prediction value;

[0008] Step S3: According to the 24-hour hourly load prediction value, the change rate and proportion of the load in different time periods are calculated, and according to the change rate and proportion of the load in different time periods, a day is divided into a plurality of flexible control periods;

[0009] Step S4: According to the load characteristics of the plurality of flexible control periods, the corresponding charging and discharging strategies are associated, and the initial period strategy is obtained by combining the corresponding charging and discharging strategies;

[0010] Step S5: Collect status data of energy storage devices in the distribution network. The status data includes: state of charge, battery health status, battery temperature, and cumulative cycle count. Calculate the device health index based on the state of charge and battery health status data. Calculate the operating status index based on the battery temperature and cumulative cycle count. Combine the device health index and the operating status index to calculate a comprehensive status index. Generate correction rules based on the comprehensive status index. Correct the initial time period strategy using the correction rules to obtain the corrected time period strategy.

[0011] Preferably, the step of collecting historical load data of the distribution network, extracting features from the historical load data of the distribution network, and integrating the data to obtain a load feature sequence includes:

[0012] Data collection:

[0013] Historical load data is extracted from the SCADA system, smart meters, and load monitoring terminals of the power distribution network, with a time span of at least one year, and divided into minute-level raw data and hour-level aggregated data according to time granularity.

[0014] Feature extraction:

[0015] Basic load characteristics: Extract daily peak load, daily minimum load, load fluctuation rate, and load cycle pattern;

[0016] Derivative features: Calculate the peak-to-valley load difference and the proportion of peak periods;

[0017] Sequence integration: Integrate the load characteristics of the previous 24 hours in chronological order to form an hourly load characteristic vector sequence. , , ..., ], each of which Include , Features include load volatility, load peak-to-valley difference, and peak period proportion.

[0018] Preferably, the construction of the load meteorological dual-dimensional prediction model includes:

[0019] Dataset partitioning: The dataset is divided into training set, validation set, and test set according to time sequence;

[0020] Temporal feature extraction layer:

[0021] Input layer: Receives the load characteristic sequence of the previous 24 hours. , , ..., ], with dimensions [24, ], This represents the number of load characteristics;

[0022] Hidden layers: Two LSTM layers are used. The first layer has 128 neurons and returns the sequence to preserve temporal features. The second layer has 64 neurons and outputs a load temporal feature vector. ;

[0023] Activation function: The default activation function that relies on the LSTM gating mechanism;

[0024] Meteorological impact modeling layer:

[0025] Using Extreme Gradient Boosting XGBoost:

[0026] A machine learning algorithm based on a gradient boosting framework improves the predictive performance of the model by integrating multiple decision trees;

[0027] Input: Hourly meteorological data for the next 24 hours predicted from the current moment, where the meteorological data for the first 24 hours includes the meteorological feature vector and extreme weather label code for each hour;

[0028] Model training: The learning rate was set to 0.05, the maximum tree depth to 4, and the optimal number of iterations was determined through cross-validation. For each hourly input meteorological data, a meteorological impact correction factor was output. ;

[0029] Joint prediction layer:

[0030] Fully Connected Layer: Splicing and The input is fed into two fully connected layers, and finally outputs hourly load forecasts for the next 24 hours through a linear layer. ;

[0031]

[0032] in, This is the hourly load forecast for the next 24 hours, where t is a small period within that 24 hours. It is the load time-series feature vector. It is a meteorological impact correction factor. It is the baseline load offset of historical data for the same period;

[0033] Loss function definition:

[0034] The weighted sum of the root mean square error and the mean absolute percentage error is used as the loss function:

[0035]

[0036] Where Loss is the loss function, used to comprehensively measure the overall difference between the model's predicted values ​​and the actual values. The smaller the value, the better the prediction effect; n: represents the number of samples, that is, the total number of observation data points involved in the calculation; : Represents the predicted value of the i-th sample, which is the model's estimate of the output for that sample; : Represents the true value of the i-th sample, which is the accurate data actually observed; The optimal balancing coefficient is determined through cross-validation on the validation set. This is used to adjust the weights of the two error components in the loss function to balance the model's attention to different types of errors. It needs to be adjusted and optimized according to the model's training status.

[0037] Optimizer selection: Use the Adam optimizer with an initial learning rate of 0.001, which decays to 0.8 times the current value every 5 epochs. The batch size is set to 32, and the iteration is performed for 100 epochs.

[0038] Preferably, the step of calculating the rate of change and proportion of load in different time periods based on the hourly load forecast values ​​for the next 24 hours includes:

[0039] Load change rate The formula for calculating the difference between two consecutive hourly load forecasts is as follows:

[0040]

[0041] in, It is the rate of load change, where t is a small period within a 24-hour period. This is the load forecast value for the t-hour period. It is the load value in the t-hour period, that is... The adjacent load forecast values, with the last time period closing with the first time period;

[0042] Load ratio The load forecast value relative to the rated capacity of the distribution network The ratio is given by the formula:

[0043]

[0044] in, It is the load percentage. This refers to the rated capacity of the distribution network, where t is a small time period within a 24-hour period. This is the load forecast value for the t-hour period.

[0045] Preferably, the step of dividing a day into several flexible control periods based on the rate and proportion of load change at different times includes:

[0046] Flexible control of time period division rules:

[0047] Rapid rise phase: If a certain time period t satisfies and This is then classified as a period of rapid growth.

[0048] Stable fluctuation period: if and This indicates a period of stable fluctuation.

[0049] Peak risk period: when and This is divided into peak risk periods;

[0050] Recovery period from the trough: If and This is divided into a trough recovery period;

[0051] Iterate through the 24-hour load forecast data, calculate v(t) and a(t) for each period from t=1 to t=24, determine the type of each time period [t, t+1] according to the rules, ensure that each hourly period is divided, and finally form a complete 24-hour flexible control time period division result.

[0052] Preferably, the step of associating corresponding charging and discharging strategies with the load characteristics of the plurality of flexible control time periods, and combining the corresponding charging and discharging strategies to obtain the initial time period strategy includes:

[0053] Based on the load characteristics of each flexible control period, the most suitable charging and discharging strategy is matched for each period, enabling energy storage devices to respond to load fluctuations in advance and optimize the balance between power supply and demand.

[0054] Charge / discharge strategy parameters, the strategy parameters associated with each time period include:

[0055] Start-up threshold: The load threshold that triggers the charging / discharging of energy storage devices;

[0056] Preset power value: The amount of electricity that the energy storage device needs to store before the start of this period;

[0057] Pre-discharge window time: The time window for opening the discharge before the peak risk period;

[0058] Charging / discharging power: The charging / discharging power limit of the energy storage device during this period.

[0059] Preferably, the step of calculating the health index based on the state of charge and battery health data includes:

[0060] The health index of the energy storage device is calculated based on the state of charge (SOC) and the state of health (SOH).

[0061]

[0062] in, It is the equipment health index. The state of charge reflects the current charge level of the battery. yes The weights represent the emphasis on the current state of charge. It indicates the battery's health status and reflects the degree of battery aging. This is the weight of the battery's health status; this formula is used to comprehensively assess the health status of energy storage devices.

[0063] Preferably, the step of calculating the operating status index based on the battery temperature and the cumulative number of cycles includes:

[0064] Calculate the battery's operating state index based on the battery temperature T and the cumulative number of cycles N:

[0065]

[0066] in, It is an operating status index, where T is the battery temperature. At (ideal battery operating temperature), This part is valued N represents the cumulative number of battery cycles. For the rated number of cycles; when (New battery), this part is 0.5; when (Rated number of cycles reached), this part is 0; reflects the performance degradation of the battery due to cycle use. The more cycles, the more obvious the degradation. The closer the OI value is to 1, the closer the battery temperature is to the ideal value and the fewer the cumulative number of cycles, the better the operating condition. The closer the OI value is to 0, the closer the temperature is to the ideal range or the more the cumulative number of cycles, the worse the operating condition.

[0067] Preferably, the step of combining the health index and the operational status index to calculate the comprehensive status index includes:

[0068] A comprehensive condition index is obtained by combining the health index and operating status index of energy storage equipment:

[0069]

[0070] in, It is a comprehensive state index. It's a health index. Operating state index, numerator : Directly reflects the linear superposition of health index and operational status index, highlighting the combined effect of the two; denominator Introducing a nonlinear adjustment term ,when or As the value approaches zero, the molecular size decreases, inhibiting... The value; when and When both remainders approach 1, the denominator approaches 2, making... Approaching 1.

[0071] Preferably, the calculation of the waviness correction value by combining the waviness correction value and the environmental compensation value includes:

[0072] Based on the value of the state composite index, generate correction rules:

[0073] Start-up threshold relaxed by 5%, preset power value increased by 10%, pre-discharge window increased by 1 hour; The startup threshold is tightened by 5%, the preset power value is reduced by 10%, and the pre-discharge window is reduced by 1 hour. Based on the correction rules, the initial time period strategy is modified to obtain the modified time period strategy.

[0074] Beneficial effects

[0075] This invention provides a charging and discharging control method for energy storage devices in power distribution networks, applicable to both power supply and distribution systems of smart grids. It offers the following advantages:

[0076] (1) The charging and discharging control method applicable to energy storage equipment in distribution networks comprehensively considers the load characteristics in historical operating data and real-time meteorological factors, including meteorological feature vectors and extreme weather, and constructs a model from multiple dimensions to predict the load. Compared with the method based solely on historical load data, it can more comprehensively and accurately predict the hourly load situation in the next 24 hours, providing a more reliable basis for the formulation of subsequent energy storage control strategies.

[0077] (2) The charging and discharging control method applicable to energy storage equipment in distribution networks divides a day into multiple flexible control periods and associates them with corresponding charging and discharging strategies. Combined with energy storage equipment status assessment and strategy correction, the charging and discharging operations of energy storage equipment in different time periods are more targeted and reasonable, effectively improving the utilization efficiency of energy storage resources, avoiding unnecessary charging and discharging switching, and extending the life of energy storage equipment.

[0078] (3) The charging and discharging control method applicable to energy storage equipment in distribution network uses real-time monitoring of load operating status and attributes, clustering algorithm to identify load categories and determine priority weights, and combined with the modified strategy to control the charging and discharging of energy storage equipment. This enables the distribution network to better balance power supply and demand when facing different load characteristics and fluctuations, reduce voltage fluctuations and line losses, thereby enhancing the overall stability and reliability of the distribution network operation. Attached Figure Description

[0079] Figure 1 This is a flowchart of a charging and discharging control method for energy storage devices in power distribution networks, as proposed in this invention.

[0080] Figure 2 This is a flowchart illustrating the initial time period strategy obtained by the charging and discharging control method for energy storage devices in power distribution networks proposed in this invention.

[0081] Figure 3 This is a flowchart illustrating the modified time period strategy for a charging and discharging control method for energy storage devices in power distribution networks proposed in this invention. Detailed Implementation

[0082] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0083] Please see Figure 1 This invention provides a technical solution: a charging and discharging control method suitable for energy storage devices in power distribution networks. Specifically, the following charging and discharging control method suitable for energy storage devices in power distribution networks is provided; please refer to [link / reference]. Figure 1 The method includes the following steps:

[0084] Step S1: Collect historical load data of the distribution network, extract and integrate the features of the historical load data of the distribution network to obtain the load feature sequence; collect meteorological data of the distribution network.

[0085] Load data acquisition and feature extraction:

[0086] Data collection:

[0087] Historical load data is extracted from the SCADA system, smart meters, and load monitoring terminals of the power distribution network, with a time span of at least one year (including typical scenarios for all four seasons). The data is divided into minute-level raw data (for fluctuation frequency analysis) and hour-level aggregated data (for daily peak / valley statistics) based on time granularity.

[0088] Feature extraction:

[0089] Basic load characteristics: Extracting daily peak load ( ), daily minimum load ( ), load volatility ( ).

[0090] Derivative characteristics: Calculation of load peak-to-valley difference ( Peak period percentage (peak load duration / 24 hours).

[0091] Data preprocessing:

[0092] Missing value imputation: Time series interpolation methods (such as linear interpolation and cubic spline interpolation) are used to imput missing load data.

[0093] Outlier filtering: based on The principle or isolated forest algorithm is used to eliminate extreme load fluctuations caused by equipment failures, etc.

[0094] Sequence integration: Integrate the load characteristics of the previous 24 hours in chronological order to form an hourly load characteristic vector sequence. , , ..., ], each of which Include , Features include load volatility, load peak-to-valley difference, and peak period proportion.

[0095] Meteorological data collection and correlation mapping for power distribution networks:

[0096] Data collection:

[0097] Connect to the meteorological bureau's API or local weather stations to obtain real-time / historical weather data, including:

[0098] Basic meteorological factors: air temperature ( ),humidity( ), solar radiation intensity ( );

[0099] Extreme weather tags: heavy rain, high temperature ), cold wave ( ) and so on, using unique thermal coding (e.g., high temperature weather is coded as [1, 0, 0, 0]).

[0100] Data association and integration:

[0101] Following the time alignment principle, hourly meteorological data is matched with load data. A real-time meteorological feature vector is generated from the meteorological data. This data is combined with extreme weather label codes (such as [1, 0, 0, 0]) and used as part of the meteorological data input for subsequent models.

[0102] Through the above steps, the integration of load characteristic sequences and meteorological data was completed, providing structured and standardized input data for the next step of the load-meteorological dual-dimensional prediction model, ensuring that the model can effectively capture the correlation characteristics between load and meteorological factors.

[0103] Step S2: Construct a load-meteorological dual-dimensional prediction model. Input the load characteristic sequence and distribution network meteorological data into the load-meteorological dual-dimensional prediction model, and output the hourly load prediction values ​​for the next 24 hours.

[0104] Load-meteorological dual-dimensional modeling:

[0105] Dataset partitioning:

[0106] The dataset is divided into a training set (70%), a validation set (15%), and a test set (15%) in chronological order. For example, for one year of historical data, the first 8.4 months are used for training, the middle 1.8 months for validation, and the last 1.8 months for testing, ensuring the time series continuity of each dataset.

[0107] Temporal Feature Extraction Layer (LSTM Long Short-Term Memory Network)

[0108] Input layer: Receives the load characteristic sequence of the previous 24 hours. , , ..., ], with dimensions [24, ]( (This refers to the number of load characteristics).

[0109] Hidden layers: Two LSTM layers are used. The first layer has 128 neurons and returns the sequence to preserve temporal features. The second layer has 64 neurons and outputs a load temporal feature vector. .

[0110] Activation function: The default activation function that relies on the LSTM gating mechanism;

[0111] Meteorological impact modeling layer (limit gradient boosting XGBoost)

[0112] Using XGBoost:

[0113] It is a machine learning algorithm based on the gradient boosting framework, which improves the predictive performance of the model by integrating multiple decision trees.

[0114] Input: Hourly meteorological data for the next 24 hours predicted from the current moment, where the meteorological data includes a meteorological feature vector for each hour (e.g., ...). ) and future extreme weather label coding (unique thermal coding, such as high temperature weather as ), ).

[0115] Model training: The learning rate is set to 0.05, the maximum tree depth is 4, and the optimal number of iterations (e.g., 500 rounds) is determined through cross-validation. For each hourly input meteorological data, a meteorological impact correction factor is output. .

[0116] Joint prediction layer:

[0117] Fully Connected Layer: Splicing and The input is fed into two fully connected layers (128 neurons in the first layer and 64 neurons in the second layer), and finally outputs hourly load predictions for the next 24 hours through a linear layer. .

[0118]

[0119] in, This is the hourly load forecast for the next 24 hours, where t is a small period within that 24 hours. It is the load time-series feature vector. It is a meteorological impact correction factor. It is the baseline load offset of historical data for the same period.

[0120] For each hour t, select load data from the same historical hour and calculate the average of these historical data for the same period. 。:

[0121]

[0122] in, is the load value of the i-th year at hour t, and n is the number of years.

[0123] Offset calculation

[0124]

[0125] in, It is the reference load offset. It is the historical average load for the same period at time t. It is the current baseline load (set based on the power grid's operating experience and planning).

[0126] By following the steps above, the baseline load offset of historical data for the same period can be obtained. This offset reflects the difference between the historical average load and the current actual load at the same point in time, which helps to make more accurate corrections in load forecasting models.

[0127] Loss function definition:

[0128] The weighted sum of the root mean square error (RMSE) and the mean absolute percentage error (MAPE) is used as the loss function:

[0129]

[0130] Where Loss is the loss function, used to comprehensively measure the overall difference between the model's predicted values ​​and the actual values. The smaller the value, the better the prediction effect; n: represents the number of samples, that is, the total number of observation data points involved in the calculation; : Represents the predicted value of the i-th sample, which is the model's estimate of the output for that sample; : Represents the true value of the i-th sample, which is the accurate data actually observed; The balancing coefficient is a hyperparameter that is determined optimally through cross-validation on a validation set. By iterating over the validation set λ∈{0.01, 0.1, 1}, the value that minimizes the combined error of RMSE and MAPE is selected. This is used to adjust the weights of the two error components (root mean square error and mean absolute percentage error) in the loss function to balance the model's attention to different types of errors. It needs to be adjusted and optimized according to the model training situation.

[0131] Optimizer selection: Use the Adam optimizer with an initial learning rate of 0.001, which decays to 0.8 times the current value every 5 epochs. The batch size is set to 32, and the iteration is performed for 100 epochs.

[0132] Predictive execution:

[0133] The model is trained using historical load-meteorological data. After training, hourly meteorological data for the next 24 hours, predicted from the current moment, are input into the trained model, which then outputs hourly load forecasts for the next 24 hours. ( ).

[0134] By employing a "load-meteorological dual-dimensional model," the system can accurately reflect load change trends and provide refined data support. Forecast results are updated hourly, providing more detailed foundational data for time-of-use dispatching of the power grid. Dispatchers can then rationally arrange power generation plans and adjust grid operation modes based on load forecasts for different time periods, thereby improving grid operational efficiency.

[0135] Step S3: Based on the hourly load forecast values ​​for the next 24 hours, calculate the rate of change and proportion of load in different time periods. Based on the rate of change and proportion of load in different time periods, divide the day into several flexible control time periods.

[0136] Based on hourly load forecasts for the next 24 hours The rate and proportion of change are used to divide the 24-hour period into several flexible control periods (such as rapid rise period, peak risk period, etc.) to replace the traditional fixed peak and valley period.

[0137] Load change rate The formula for calculating the difference between two consecutive hourly load forecasts is as follows:

[0138]

[0139] in, It is the rate of load change, where t is a small period within a 24-hour period. This is the load forecast value for the t-hour period. It is the load value in the t+1 hour period, that is... The adjacent load forecast values, the last time period closes with the first time period (forming a 24-hour cycle: );

[0140] Load ratio The load forecast value relative to the rated capacity of the distribution network The ratio is given by the formula:

[0141]

[0142] in, This refers to the load share, which reflects the proportion of load relative to the grid capacity during a given period. This refers to the rated capacity of the distribution network, where t is a small time period within a 24-hour period. This is the load forecast value for the t-hour period;

[0143] Flexible control of time period division rules:

[0144] Rapid rise phase: If a certain time period t+1 satisfies and This is then classified as a rapid growth period. For example, hour, , , The conditions are met.

[0145] Stable fluctuation period: if and This indicates a period of stable fluctuation. For example... , , , This conforms to the definition.

[0146] Peak risk period: when and This is divided into peak risk periods. For example, , , ,but The period in question is a peak risk period.

[0147] Recovery period from the trough: If and This is divided into a trough recovery period. For example... , , This period is included here.

[0148] Iterate through the 24-hour load forecast data, calculate v(t) and a(t) one by one from t=1 to t=24, and determine the type of each time period [t, t+1] according to the rules (when t=24, the time period is [24, 1]) to ensure that each hourly segment is divided, and finally form a complete 24-hour flexible control time period division result.

[0149] Rapid rise period: [6:00-9:00]; [17:00-19:00];

[0150] Stable fluctuation period: [9:00-17:00];

[0151] Peak risk period: [19:00-22:00];

[0152] Recovery period from the trough: [22:00-6:00];

[0153] This provides comprehensive and accurate time-period data for subsequent daily strategy mapping, ensuring that the energy storage device's charging and discharging strategies are closely matched with load change characteristics, covering all time periods throughout the day.

[0154] Step S4: Based on the load characteristics of the several flexible control periods, associate the corresponding charging and discharging strategies, and combine the corresponding charging and discharging strategies to obtain the initial period strategy.

[0155] Construction of the policy mapping table:

[0156] Based on the load characteristics of each flexible control period (such as the load change rate, proportion, and duration of periods like the rapid rise period and peak risk period).

[0157] Matching the most suitable charging and discharging strategy to each time period enables energy storage devices to respond to load fluctuations in advance and optimize the balance between power supply and demand.

[0158] Charge / discharge strategy parameters, the strategy parameters associated with each time period include:

[0159] Start-up threshold: The load threshold that triggers the charging / discharging of the energy storage device (e.g., when the load reaches 80% of the rated capacity, discharge is initiated).

[0160] Preset power value: The amount of electricity that the energy storage device needs to store before the start of this period (e.g., charging to SOC=70% before the rapid rise period).

[0161] Pre-discharge window time: The time window for opening the discharge before the peak risk period;

[0162] Charging / discharging power: The charging / discharging power limit of the energy storage device during this period.

[0163] The strategy mapping table is as follows, serving as the initial time-period strategy:

[0164]

[0165] By customizing strategies based on time-period load characteristics, the lag inherent in traditional fixed peak-valley strategies is avoided, thus improving the real-time response of energy storage. Pre-setting electricity based on load forecasts reduces frequent start-ups and shutdowns of energy storage devices, extending their lifespan. It also provides a benchmark for dynamic adjustments: the initial time-period strategy provides a quantifiable parameter basis for subsequent state corrections and load priority adjustments, forming a hierarchical control framework.

[0166] Step S5: Collect status data of energy storage devices in the distribution network. The status data of energy storage devices in the distribution network includes: state of charge, battery health status, battery temperature, and cumulative cycle count. Calculate a health index based on the state of charge and battery health status data. Calculate an operating status index based on the battery temperature and cumulative cycle count. Combine the health index and operating status index to calculate a comprehensive status index. Generate correction rules based on the comprehensive status index. Correct the initial time period strategy using the correction rules to obtain the corrected time period strategy.

[0167] The status data of energy storage devices in the distribution network includes four core indicators: State of Charge (SOC), State of Health (SOH), Battery Temperature (T), and Cumulative Cycle Count (N).

[0168] State of charge (SOC) definition: The percentage of the battery's current charge relative to its rated capacity (reflecting real-time available capacity).

[0169] Data acquisition method: The data is obtained in real time by the battery management system (BMS) using the coulomb counting method or the Kalman filtering algorithm.

[0170] Battery State of Health (SOH) definition: The ratio of the battery's remaining capacity to its rated capacity (measures the degree of battery aging; for a new battery, SOH = 100%).

[0171] Data acquisition method: The BMS estimates SOH by monitoring parameters such as battery internal resistance and capacity decay curve, combined with the Arrhenius model; it is calibrated periodically (e.g., once a day) through charge-discharge tests, or indirectly derived based on the cumulative number of cycles (e.g., SOH decreases by 1%~3% every 100 cycles);

[0172] Battery temperature (T) definition: The real-time temperature of the energy storage battery pack (affecting charging and discharging efficiency and lifespan).

[0173] Data acquisition method: Hardware: Thermocouple sensors or thermistors are deployed between battery modules to directly measure the surface temperature of the battery cells; System: The temperature of each measuring point is collected through the data acquisition module (DAQ), and the average or maximum value is taken as the battery temperature.

[0174] Total number of loops (N):

[0175] Definition: The number of complete charge and discharge cycles a battery completes from full charge to full discharge (directly related to the rate of battery degradation).

[0176] Data collection method: The BMS has a built-in counter. When the SOC drops from ≥95% to ≤5% and is then recharged to ≥95%, it is counted as one complete cycle. Manual reset is supported (such as resetting the count after battery replacement).

[0177] Preprocessing process

[0178] Missing value handling: When temperature data is missing, time series interpolation (such as linear interpolation) is used to fill it in based on the values ​​at previous and subsequent times; when SOC / SOH / N is missing, it is marked as "invalid data" and relies on BMS retransmission or the next cycle of data to overwrite it.

[0179] Outlier filtering:

[0180] Temperature anomaly judgment: If the measured temperature is >60℃ or <-20℃ (outside the battery's safe operating range), it is considered a sensor malfunction, triggering hardware self-test and removing the data point; SOC / SOH logic verification: If SOC is >100% or <0%, or SOH is >100% (except for new batteries), it is judged as a calculation error, and the BMS algorithm is reset and data is collected again.

[0181] The health index of the energy storage device is calculated based on the state of charge (SOC) and the state of health (SOH).

[0182]

[0183] in, It is the health index of energy storage equipment. The state of charge reflects the current charge level of the battery. yes The weights represent the emphasis on the current state of charge. It indicates the battery's health status, reflecting the degree of battery aging (such as capacity decay and increased internal resistance). This is a weighted measure of battery health; this formula is used to comprehensively assess the health status of energy storage devices. For example, , ,but The closer the value is to 1, the better the battery health.

[0184] Calculate the battery's operating state index based on the battery temperature (T) and the cumulative number of cycles N:

[0185]

[0186] in, It is an operating status index, where T is the battery temperature. At (ideal battery operating temperature), This part is valued N represents the cumulative number of battery cycles. For the rated number of cycles; when (New battery), this part is 0.5; when (Rated number of cycles reached), this part is 0; it reflects the performance degradation of the battery due to cycle use. The more cycles, the more obvious the degradation.

[0187] A closer OI value to 1 indicates that the battery temperature is close to the ideal value and the cumulative number of cycles is low, indicating excellent operating condition; a closer OI value to 0 indicates that the temperature deviates from the ideal range or the cumulative number of cycles is high, indicating poor operating condition. This index provides a quantitative basis for evaluating the actual operating condition of the battery and can be used to dynamically adjust the charging and discharging strategies of energy storage devices.

[0188] A comprehensive condition index is obtained by combining the health index and operating status index of energy storage equipment:

[0189]

[0190] in, It is a comprehensive state index. It's a health index. Operating state index, numerator : Directly reflects the linear superposition of health index and operational status index, highlighting the combined effect of the two; denominator Introducing a nonlinear adjustment term ,when or As the value approaches zero, the molecular size decreases, inhibiting... The value; when and When both remainders approach 1, the denominator approaches 2, making... Approaching 1.

[0191] Based on the value of the state composite index, generate correction rules:

[0192] Start-up threshold relaxed by 5%, preset battery level -10%, pre-charging window -1 hour; Start-up threshold tightened by 5%, preset power value increased by 10%, pre-discharge window increased by 1 hour.

[0193]

[0194] Based on the correction rules, the initial time period strategy is modified to obtain the modified time period strategy as shown above.

[0195] Step S6: Collect load data from downstream of the distribution network, classify the loads according to preset classification rules and load data to obtain load categories, and perform weight analysis on the load categories using the hierarchical analysis method to calculate the load priority weights.

[0196] Collect load data, which includes running status and attribute data:

[0197] Data Source: Relevant data is collected from various load terminals downstream of the distribution network. This data can be obtained through devices such as smart meters and sensors. Data types include, but are not limited to, real-time power, running time, start-stop frequency, power factor, and other operating status data of the load, as well as attribute data such as load type (e.g., industrial equipment, commercial appliances, residential appliances), rated power, and importance level (if preset).

[0198] Data processing: Cleaning and processing the collected data. Removing outlier data points, such as unreasonable power readings due to sensor malfunction or communication interference. Performing appropriate estimation or interpolation on missing data to ensure data integrity and accuracy.

[0199] Classification rules: definition basis and attribute judgment;

[0200] Basic infrastructure loads are defined as equipment typically related to people's basic living needs, public safety, and critical social services. Examples include medical equipment, fire-fighting facilities, and emergency lighting in hospitals. These loads require priority power supply under all circumstances to ensure social stability.

[0201] Attributes: Requires extremely high power supply reliability; power outages are generally not permitted. Its operation must be continuous and uninterrupted.

[0202] Delayable loads are defined as devices whose power supply requirements are not so urgent and whose power supply can be delayed for a certain period of time. Examples include electric vehicle charging facilities and equipment in some non-critical production workshops.

[0203] Attribute: When the power grid is under strain or energy storage devices need to prioritize other more important loads, their power supply time can be appropriately delayed. The power demand of these loads has a certain degree of elasticity.

[0204] Intermittently operating loads are defined as loads whose operation is intermittent, meaning they do not run continuously. Examples include refrigeration equipment and air conditioning systems, which turn on and off intermittently based on external conditions such as temperature.

[0205] Attributes: Its power demand fluctuates greatly at different times. When controlling the charging and discharging of energy storage devices, it is necessary to consider its operating patterns so as to reasonably arrange the operation of energy storage devices when they are not running or when the power is low.

[0206] Classification and Determination: Using preset classification rules, the load is divided into:

[0207] Basic infrastructure (such as hospitals and communication base stations): requires extremely high power supply reliability and cannot be interrupted.

[0208] Delayable types (such as electric vehicle charging and household washing machines): Power supply periods can be flexibly adjusted.

[0209] Intermittent operation equipment (such as refrigeration equipment and elevators): Short-term power outages are permitted.

[0210] The load priority weights are calculated using the analytic hierarchy process (AHP).

[0211] Building a hierarchical structure:

[0212] Target level: Maximize the overall benefits of the power distribution network.

[0213] Criterion layer: Reliability (weight) ), economic ( ), environmental protection ( ).

[0214] Solution layer: Various load categories (basic protection category) Delayable class Intermittent operation type ).

[0215] Judgment matrix and weight calculation:

[0216] Experts compared each element in each layer pairwise to construct a judgment matrix (e.g., ).

[0217] Calculate the feature vectors and normalize them to obtain the load priority weights. Basic security category Delayable class Intermittent operation class .

[0218] Step S7: Based on the modified time period strategy and load priority weight, calculate the charging and discharging allocation index of the energy storage device, and control the charging and discharging of the energy storage device in the distribution network according to the charging and discharging allocation index of the energy storage device.

[0219] Adjusted time-period strategy factor:

[0220] Obtain the correction strategy parameters for each time period from the correction time period strategy, and define the strategy factors. The impact of time period on energy storage allocation:

[0221]

[0222] in: It is a strategy factor used to reflect the degree of influence of this period on energy storage allocation. The larger the value, the higher the weight of this period in energy storage allocation. It is a comprehensive index of equipment condition; For time period The load percentage (as a percentage of rated capacity), It is the load proportion influence function, and the input is the load proportion a(t) in time period t. For time period The rate of change of load, It is a load change rate influence function, with the load change rate in time period t as the input. (t); The weighting coefficient is used to adjust the relative influence of different factors on the strategy factor; t is the t-th hour within a 24-hour time period; where, the peak risk period is: (Prioritize protection, increase discharge); Rapid rise phase: (Standard response); Period of stable fluctuation: (Appropriate adjustments); Recovery period from the trough: (Prioritize charging, limit discharging).

[0223] Load ratio impact function:

[0224]

[0225] in, It is a load occupancy impact function, quantifying the impact of load occupancy on strategy factors. This represents the load percentage at time t, where t is the t-th hour within a 24-hour period. It is the basic value of the load proportion influence function. This is the rated capacity relative to the load percentage. The higher the load percentage (the closer to or exceeding the rated capacity), the greater the load percentage becomes. k is a coefficient greater than 0. When it increases, Will follow The increase is linear, indicating that more attention needs to be paid to energy storage allocation during this period (such as strengthening energy storage discharge during peak load periods).

[0226] Effect function of load change rate:

[0227]

[0228] in, It is the load change rate influence function, reflecting the effect of the speed of load change on the strategy factor. It is the rate of load change at time t. This is the threshold value for the load change rate, where t is the t-th hour within a 24-hour time period, and r is the base value of the load change rate influence function. It is an influence coefficient; if the load change rate is fast (reaching or exceeding the set threshold)... ), Will follow The linear increase in load indicates that the load fluctuates drastically during this period, requiring flexible adjustments to energy storage allocation (such as charging in advance when the load surges during a rapid rise).

[0229] Combining the strategy factor and load priority weight, the charge / discharge allocation index is obtained:

[0230]

[0231] in, Assigning an index to the charging and discharging of energy storage devices. It is the first Class load during time period The predicted power demand; It is the load priority weight. It is a strategy factor. It is a time period The total demand of the three load categories is used to balance the competition among loads (the higher the demand share, the lower the allocation index, to avoid a single load monopolizing electricity). i is the load type, (i=1) represents Class A load; (i=2) represents Class B load; (i=3) represents Class C load; j is the control period (peak risk period, rapid rise period, stable fluctuation period, and low recovery period).

[0232] Example: Basic protection class ( During peak periods ( Demand accounts for 30%: Allocation Index Delayable classes ( During the trough ( Demand accounts for 50%: Allocation Index ;

[0233] Ensure that the sum of the distribution indices of all loads is 1 to facilitate subsequent power allocation.

[0234]

[0235] in, It is the normalized charge / discharge distribution index of energy storage devices. The charging and discharging index is assigned to the energy storage device. i is the load type, (i=1) represents Class A load; (i=2) represents Class B load; (i=3) represents Class C load, and j is the control period (peak risk period, rapid rise period, stable fluctuation period, and trough recovery period).

[0236] Energy storage device constraints: Physical limiting parameters of energy storage devices to ensure charging and discharging safety. ; Current SOC: =65%, SOC operating range: [20%, 95%]

[0237] Power distribution calculation:

[0238] Charging power allocation:

[0239]

[0240] Discharge power distribution:

[0241]

[0242] in, and These are the charging power allocation index and the discharging power allocation index, respectively. and These are the maximum values ​​of charging power and discharging power (constraints). It is the normalized energy storage device charge and discharge distribution index, where i is the load type and j is the divided control period. The charge and discharge status is determined according to the correction period strategy.

[0243] For each load Send charging / discharging power command: command

[0244] This flexible power command mechanism optimizes grid operation. Under different load and energy storage conditions, energy storage devices can respond quickly and accurately, adjusting the power balance in the grid. This helps reduce peak-to-valley differences in the grid, improves power supply reliability and power quality, and also reduces grid operating costs.

[0245] This invention collects historical operating data of the distribution network to extract load characteristic sequences; simultaneously, it collects meteorological data closely related to load fluctuations to form real-time meteorological feature vectors and extreme weather label codes, constructing a dual-dimensional load-meteorological prediction model. The above data is input into the model, outputting hourly load prediction values ​​for the next 24 hours. Based on the prediction values, the load change rate and proportion for different time periods are calculated, dividing the day into several flexible control periods. According to the load characteristics of each flexible control period, corresponding charging and discharging strategies are associated to form an initial time period strategy mapping table containing multiple strategies. Status data of energy storage devices in the distribution network are collected, including state of charge, battery health status, battery temperature, and cumulative cycle count. A health index, operating status index, and comprehensive status index are calculated to comprehensively assess the status of the energy storage devices. Based on the comprehensive status index, correction rules are generated to modify the initial time period strategies, resulting in modified time period strategies that better match the actual status of the energy storage devices. The operating status and attributes of downstream loads in the distribution network are monitored in real time to identify load categories. Load priority weights are determined using the analytic hierarchy process (AHP), and combined with the modified time period strategies and these weights, the charging and discharging allocation index of the energy storage devices is calculated, ultimately achieving precise charging and discharging control of the energy storage devices in the distribution network.

[0246] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0247] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

Claims

1. A charging and discharging control method for energy storage devices in power distribution networks, characterized in that, Includes the following steps: Step S1: Collect historical load data of the distribution network, extract and integrate the features of the historical load data of the distribution network to obtain the load feature sequence; Collect meteorological data for power distribution networks; Step S2: Construct a load-meteorological dual-dimensional prediction model. Input the load characteristic sequence and distribution network meteorological data into the load-meteorological dual-dimensional prediction model, and output the hourly load prediction values ​​for the next 24 hours. Step S3: Based on the hourly load forecast values ​​for the next 24 hours, calculate the rate of change and proportion of load in different time periods, and divide the day into several flexible control time periods based on the rate of change and proportion of load in different time periods. Step S4: Based on the load characteristics of the several flexible control periods, associate the corresponding charging and discharging strategies, and combine the corresponding charging and discharging strategies to obtain the initial period strategy; Step S5: Collect status data of energy storage devices in the distribution network. The status data includes: state of charge, battery health status, battery temperature, and cumulative cycle count. Calculate the device health index based on the state of charge and battery health status data. Calculate the operating status index based on the battery temperature and cumulative cycle count. Combine the device health index and the operating status index to calculate a comprehensive status index. Generate correction rules based on the comprehensive status index. Correct the initial time period strategy using the correction rules to obtain the corrected time period strategy. The comprehensive status index is calculated by combining the health index and the operational status index, including: A comprehensive condition index is obtained by combining the health index and operating status index of energy storage equipment: ； in, It is a comprehensive state index. It's a health index. Operating state index, numerator : Directly reflects the linear superposition of health index and operational status index, highlighting the combined effect of the two; denominator Introducing a nonlinear adjustment term ,when or As the value approaches zero, the molecular size decreases, inhibiting... The value; when and When both remainders approach 1, the denominator approaches 2, making... Approaching 1; Based on the state comprehensive index, a correction rule is generated. The initial time period strategy is then corrected using the correction rule to obtain the corrected time period strategy, including: Based on the value of the state composite index, generate correction rules: Start-up threshold relaxed by 5%, preset power value increased by 10%, pre-discharge window increased by 1 hour; The activation threshold is tightened by 5%, and the preset power value is reduced by 10%. Based on the correction rules, the initial time period strategy is corrected to obtain the corrected time period strategy.

2. The charging and discharging control method for energy storage devices in power distribution networks according to claim 1, characterized in that, The process involves collecting historical load data from the distribution network, extracting and integrating features from the historical load data to obtain a load feature sequence, including: Data collection: Historical load data is extracted from the SCADA system, smart meters, and load monitoring terminals of the power distribution network, with a time span of at least one year, and divided into minute-level raw data and hour-level aggregated data according to time granularity. Feature extraction: Basic load characteristics: Extract daily peak load, daily minimum load, load fluctuation rate, and load cycle pattern; Derivative features: Calculate the peak-to-valley load difference and the proportion of peak periods; Sequence integration: Integrate the load characteristics of the previous 24 hours in chronological order to form an hourly load characteristic vector sequence. , , ..., ], each of which Include , Features include load volatility, load peak-to-valley difference, and peak period proportion.

3. The charging and discharging control method for energy storage devices in power distribution networks according to claim 2, characterized in that, The construction of the load meteorological dual-dimensional prediction model includes: Dataset partitioning: The dataset is divided into training set, validation set, and test set according to time sequence; Temporal feature extraction layer: Input layer: Receives the load characteristic sequence of the previous 24 hours. , , ..., ], with dimensions [24, ], This represents the number of load characteristics; Hidden layers: Two LSTM layers are used. The first layer has 128 neurons and returns the sequence to preserve temporal features. The second layer has 64 neurons and outputs a load temporal feature vector. ; Activation function: The default activation function that relies on the LSTM gating mechanism; Meteorological impact modeling layer: Using Extreme Gradient Boosting XGBoost: A machine learning algorithm based on a gradient boosting framework improves the predictive performance of the model by integrating multiple decision trees; Input: Hourly meteorological data for the next 24 hours predicted from the current moment, where the meteorological data for the first 24 hours includes the meteorological feature vector and extreme weather label code for each hour; Model training: The learning rate was set to 0.05, the maximum tree depth to 4, and the optimal number of iterations was determined through cross-validation. For each hourly input meteorological data, a meteorological impact correction factor was output. ; Joint prediction layer: Fully Connected Layer: Splicing and The input is fed into two fully connected layers, and finally outputs hourly load forecasts for the next 24 hours through a linear layer. ; ； in, This is the hourly load forecast for the next 24 hours, where t is a small period within that 24 hours. It is the load time-series feature vector. It is a meteorological impact correction factor. It is the baseline load offset of historical data for the same period; Loss function definition: The weighted sum of the root mean square error and the mean absolute percentage error is used as the loss function: ； Where Loss is the loss function, which is used to comprehensively measure the overall difference between the model's predicted values ​​and the actual values. The smaller the value, the better the prediction effect; n: represents the number of samples, that is, the total number of observation data points involved in the calculation; : Represents the predicted value of the i-th sample, which is the model's estimate of the output for that sample; : Represents the true value of the i-th sample, which is the accurate data actually observed; The optimal balancing coefficient is determined through cross-validation on the validation set. This is used to adjust the weights of the two error components in the loss function to balance the model's attention to different types of errors. It needs to be adjusted and optimized according to the model's training status. Optimizer selection: Use the Adam optimizer with an initial learning rate of 0.001, which decays to 0.8 times the current value every 5 epochs. The batch size is set to 32, and the iteration is performed for 100 epochs.

4. The charging and discharging control method for energy storage devices in power distribution networks according to claim 3, characterized in that, The calculation of the load change rate and proportion for different time periods based on the hourly load forecast values ​​for the next 24 hours includes: Load change rate The formula for calculating the difference between two consecutive hourly load forecasts is as follows: ； in, It is the rate of load change, where t is a small period within a 24-hour period. This is the load forecast value for the t-hour period. It is the load value in the t-hour period, that is... The adjacent load forecast values, with the last time period closing with the first time period; Load ratio The load forecast value relative to the rated capacity of the distribution network The ratio is given by the formula: ； in, It is the load percentage. This refers to the rated capacity of the distribution network, where t is a small time period within a 24-hour period. This is the load forecast value for the t-hour period.

5. The charging and discharging control method for energy storage devices in power distribution networks according to claim 4, characterized in that, The method divides a day into several flexible control periods based on the rate and proportion of load change at different times, including: Flexible control of time period division rules: Rapid rise phase: If a certain time period t satisfies and This is then classified as a period of rapid growth. Stable fluctuation period: if and This indicates a period of stable fluctuation. Peak risk period: when and This is divided into peak risk periods; Recovery period from the trough: If and This is divided into a trough recovery period; Iterate through the 24-hour load forecast data, calculate v(t) and a(t) for each period from t=1 to t=24, determine the type of each time period [t, t+1] according to the rules, ensure that each hourly period is divided, and finally form a complete 24-hour flexible control time period division result.

6. The charging and discharging control method for energy storage devices in power distribution networks according to claim 5, characterized in that, The step of associating the corresponding charging and discharging strategies with the load characteristics of the several flexible control periods, and combining the corresponding charging and discharging strategies to obtain the initial period strategy includes: Based on the load characteristics of each flexible control period, the most suitable charging and discharging strategy is matched for each period, enabling energy storage devices to respond to load fluctuations in advance and optimize the balance between power supply and demand. Charge / discharge strategy parameters, the strategy parameters associated with each time period include: Start-up threshold: The load threshold that triggers the charging / discharging of energy storage devices; Preset power value: The amount of electricity that the energy storage device needs to store before the start of this period; Pre-discharge window time: The time window for opening the discharge before the peak risk period; Charging / discharging power: The charging / discharging power limit of the energy storage device during this period.

7. The charging and discharging control method for energy storage devices in power distribution networks according to claim 6, characterized in that, The health index is calculated based on the state of charge and battery health data, including: The health index of the energy storage device is calculated based on the state of charge (SOC) and the state of health (SOH). ； in, It is the equipment health index. The state of charge reflects the current charge level of the battery. yes The weights represent the emphasis on the current state of charge. It indicates the battery's health status and reflects the degree of battery aging. This is the weight of the battery's health status; this formula is used to comprehensively assess the health status of energy storage devices.

8. The charging and discharging control method for energy storage devices in power distribution networks according to claim 7, characterized in that, The operating status index is calculated based on the battery temperature and the cumulative number of cycles, including: Calculate the battery's operating state index based on the battery temperature T and the cumulative number of cycles N: ； in, It is an operating status index, where T is the battery temperature. hour, This part is valued N represents the cumulative number of battery cycles. For the rated number of cycles; when This part is 0.5; when This part is 0; it reflects the performance degradation of the battery due to cycle use. The more cycles, the more obvious the degradation. The closer the OI value is to 1, the closer the battery temperature is to the ideal value and the fewer the cumulative number of cycles, indicating good operating condition. The closer the OI value is to 0, the closer the temperature is to the ideal range or the more the cumulative number of cycles, indicating poor operating condition.

Images