Active balance control method for SOC (system on chip) of network construction type energy storage system

Abstract

The invention provides a networking type energy storage system SOC active balance control method, and relates to the technical field of networking type energy storage systems, and the method comprises the steps: S1, collecting the real-time state data and load prediction data of a battery; s2, judging whether the measurement data and the performance index exceed a preset safety range or not, if so, carrying out active equalization operation, and calculating a preliminary reference instruction for the battery according to the smaller value of the charging power and the discharging power; and S3, combining the load prediction result with the preliminary reference instruction according to a preset attention degree to generate an optimal power distribution instruction. And S4, performing power distribution on the battery according to the optimal power distribution instruction. According to the method, the optimal power distribution instruction is used for performing power distribution on the battery, so that the defects that a traditional system does not fully consider the influence of user behaviors and external environment factors on power demands and lacks prediction on future power demands are overcome, the system efficiency can be improved, the user experience can be improved, and meanwhile, the energy consumption of an energy storage system is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of grid-type energy storage systems, and in particular to a SOC active balancing control method for a grid-type energy storage system. Background Art

[0002] A grid-type energy storage system refers to a system that is connected to the power grid and can store and release energy. This system usually integrates a variety of energy storage technologies (such as lithium-ion batteries, lead-acid batteries, flywheel energy storage, supercapacitors, etc.) to balance the load of the power grid, adjust the power quality, and improve the reliability and stability of the power system. The grid-type energy storage system can release the stored energy during peak power demand periods or store excess power during low power demand periods, thereby providing a smooth power supply to the power grid. The early power grid architecture was mainly unidirectional, that is, power plants delivered electricity to end users, lacking flexibility and intelligence. With the rapid development of renewable energy (such as wind and solar energy), the power grid has begun to develop in an intelligent direction, emphasizing real-time monitoring and two-way energy flow, and energy storage systems are needed to enhance the flexibility and adaptability of the power grid. Chemical energy storage technologies such as lithium-ion batteries and lead-acid batteries continue to advance, and the energy density, cycle life and cost-effectiveness of batteries have been significantly improved, making them widely used in power energy storage systems. Technologies such as pumped storage and flywheel energy storage provide different energy storage solutions to adapt to different power demand scenarios.

[0003] Electromagnetic energy storage technologies such as supercapacitors are suitable for high-frequency power regulation due to their fast charging and discharging characteristics. Energy management systems integrate algorithms and control strategies to monitor and optimize the charging and discharging process of energy storage equipment in real time to achieve optimal support for the power grid. With the global emphasis on reducing carbon emissions and improving energy efficiency, countries have introduced policies to support the development of renewable energy and energy storage technologies, and to encourage investment and technological innovation. In energy storage systems, the SOC active balancing control method is a technology used to manage and optimize the charge state of each battery cell in the battery pack to improve the performance and safety of the overall system. This method is particularly important in the process of storing, using and releasing electrical energy, especially when the charge state of different cells in the battery pack may be different.

[0004] In traditional systems, fixed or simple rule-based power allocation strategies are often used, which cannot adapt to real-time changes in electricity demand and do not fully consider the impact of user behavior and external environmental factors on electricity demand. There is a lack of prediction of future electricity demand. It has achieved the benefits of improving system efficiency, extending battery life, improving system reliability and enhancing user experience, while reducing the energy consumption of the energy storage system, which helps to achieve energy conservation and emission reduction. Summary of the invention

[0005] The present invention provides a SOC active balancing control method for a grid-type energy storage system, which is used to solve the defects in the prior art that the system cannot adapt to real-time changing power demand, does not fully consider the impact of user behavior and external environmental factors on power demand, and lacks prediction of future power demand.

[0006] In one aspect, the present invention provides a method for actively balancing the SOC of a grid-type energy storage system, comprising:

[0007] S1: Collect real-time battery status data and load forecast data. The real-time battery status data includes battery measurement data and performance indicators.

[0008] S2: Based on the real-time status data of the battery, determine whether the measured data and performance indicators exceed the preset safety range, and perform active balancing operations to restore the real-time status of the battery to the preset safety range, and then calculate the preliminary reference instructions for the battery based on the smaller value of the charging and discharging power.

[0009] S3: The decision tree model is trained with historical data obtained by integrating load forecast data to obtain a decision tree optimization model. The decision tree optimization model is used to predict power demand for future time periods to obtain load forecast results. The load forecast results are combined with preliminary reference instructions according to a preset degree of importance to generate optimal power allocation instructions.

[0010] S4: Allocate power to the battery according to the optimal power allocation instruction.

[0011] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S2, the active balancing operation includes:

[0012] The electric energy stored in the grid-type energy storage system is transferred from the battery with a high state of charge to the battery with a low state of charge. During the transfer process, the real-time status data of the battery is continuously monitored until the charge difference value is reduced to meet the preset safety range. The charge difference value is the difference between batteries with different charge states.

[0013] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S2, the formula of the preliminary reference instruction is expressed as:

[0014] P fi =min(P charge,i , P discharge,i )·SOH i ·f(T i )

[0015] Where P fi is the preliminary reference instruction of the ith battery cell, min(P charge,i , P discharge,i) means taking the smaller value of the charge and discharge power, SOH i is the compensation factor for health status, f(T i ) is the compensation factor for temperature.

[0016] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S3, the step of generating an optimal power allocation instruction includes:

[0017] S31: normalize the load forecast data and integrate them at daily time intervals to obtain historical data. The load forecast data includes historical electricity consumption data, consumer behavior research data, and external environmental factor data. The historical data includes daily total electricity consumption and average electricity consumption indicators.

[0018] S32: Build a decision tree model, use historical data to train the decision tree model, and perform cross-validation to obtain a decision tree optimization model.

[0019] S33: Use the decision tree optimization model to predict the power demand for the future time period and obtain the load prediction result.

[0020] S34: Combining the load forecast result with the preliminary reference instruction according to a preset degree of importance to generate an optimal power reference instruction.

[0021] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S31, the total daily power consumption is expressed as:

[0022]

[0023] Where P t is the power consumption at each time point in a day, n is the number of time points in that day, D total is the total daily electricity consumption.

[0024] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S31, the formula for the average power consumption index is expressed as:

[0025]

[0026] Where D aug The average daily electricity consumption.

[0027] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S32, the step of obtaining a decision tree optimization model includes:

[0028] S321: Divide the historical data into a training set and a test set.

[0029] S322: Use the machine learning library to build a decision tree model and select fitting parameters, including the depth of the tree, the minimum number of sample splits, and the minimum number of leaf node samples.

[0030] S323: Input the training set data into the decision tree model to obtain the association between the features and the target variable, and use K-fold cross validation to verify the decision tree model to obtain the average performance evaluation of the decision tree model.

[0031] S324: Adjust the decision tree model parameters according to the average performance evaluation to obtain a decision tree optimization model.

[0032] According to a method for active balancing control of SOC of a grid-type energy storage system provided by the present invention, in step S323, the K-fold cross-validation method is: divide the training set into K subsets, select one subset as the validation set each time, and the rest as the basic set, and repeat K times to obtain the average performance evaluation of the decision tree model.

[0033] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S33, the step of obtaining the load prediction result includes:

[0034] S331: Based on the decision tree optimization model, preliminary load forecasting results are obtained.

[0035] S332: Denormalize the preliminary load forecast result to restore it to the original power load value, and obtain the load forecast result, which is expressed as follows:

[0036] P original =f(X)×(P max -P min )+P min

[0037] Where P original is the load forecast result, f(X) is the preliminary load forecast result, P min is the minimum value of the training set data, P max is the maximum value of the training set data.

[0038] According to a method for active SOC balancing control of a grid-type energy storage system provided by the present invention, in step S34, the formula of the optimal power reference instruction is expressed as:

[0039] P optimal =σ·P original +(1-σ)·P fi

[0040] Where P optimal is the optimal power reference instruction, σ is the emphasis on the load forecast result, and (1-σ) is the emphasis on the preliminary reference instruction.

[0041] The SOC active balancing control method for a grid-type energy storage system provided by the present invention distributes power to the battery by using the optimal power allocation instruction, thereby solving the problems of unstable battery status, unreasonable power allocation and inaccurate power demand prediction, thereby improving system efficiency, extending battery life, improving system reliability and enhancing user experience, while reducing the energy consumption of the energy storage system, which helps to achieve the beneficial effects of energy conservation and emission reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 This is one of the flow charts of the SOC active balancing control method of the grid-type energy storage system provided in the embodiment of the present invention;

[0044] Figure 2 This is the second flow chart of the SOC active balancing control method of the grid-type energy storage system provided in the embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Combine the following Figure 1-Figure 2 The present invention describes the SOC active balancing control method of a grid-type energy storage system.

[0047] Figure 1 It is one of the flow charts of the SOC active balancing control method of the grid-type energy storage system provided in an embodiment of the present invention.

[0048] like Figure 1 As shown, the SOC active balancing control method of the grid-type energy storage system provided by the embodiment of the present invention mainly includes the following steps:

[0049] S1: Collect real-time battery status data and load forecast data. The real-time battery status data includes battery measurement data and performance indicators. These measurement data include battery voltage, current, and temperature, while performance indicators may involve battery health status (SOH), number of cycles, and charge and discharge efficiency. Through comprehensive analysis of these data, the operating status of the battery can be monitored in real time, thus providing a basis for subsequent fault diagnosis and performance optimization.

[0050] S2: According to the real-time status data of the battery, it is judged whether the measured data and performance indicators exceed the preset safety range. Active balancing operation is performed to restore the real-time status of the battery to the preset safety range, and then the preliminary reference instructions for the battery are calculated based on the smaller value of the charging and discharging power. According to the real-time status data of the battery, with the help of advanced monitoring systems and precise sensor networks, it is comprehensively and accurately judged whether the measured data and performance indicators exceed the preset safety range. Once it is determined to be exceeded, the intelligent active balancing control mechanism is immediately activated. This mechanism is based on an efficient microprocessor and runs a complex and sophisticated balancing algorithm. Through power regulation devices such as bidirectional DC-DC converters, it carries out meticulous charging and discharging adjustments for each battery cell in the battery pack. During the adjustment process, the key parameters such as battery voltage, current, and temperature are monitored in real time, and the balancing strategy is dynamically optimized until the real-time status of the battery is stably restored to the preset safety range. After the recovery is completed, the key factors such as the current SOC (state of charge), temperature, internal resistance, and remaining capacity of the battery are deeply analyzed, and the charging power and discharging power values ​​are accurately calculated by referring to the charging and discharging power limit curves pre-set by the battery management system based on battery characteristics and safety standards. Subsequently, we rigorously take the smaller value of the two as the basis, and comprehensively consider multiple additional conditions such as the real-time connection status of the current power grid, the change trend of the overall load demand of the system, and the historical operation data accumulated by the battery over a long period of time. We use professional mathematical models and data analysis methods to deeply correct and optimize the preliminary calculation results, and finally rigorously and scientifically calculate the preliminary reference instructions for the battery, providing a reliable basis for subsequent power dispatch and system operation.

[0051] In step S2, the active balancing operation includes:

[0052] During the operation of the grid-type energy storage system, when the system detects differences in the state of charge between the batteries, the charge balancing mechanism will be activated to gradually transfer the stored electrical energy from the battery with a high state of charge to the battery with a low state of charge. In this delicate and critical transfer process, with the help of a high-precision sensor network and an advanced data acquisition system, the real-time voltage, current, temperature and other multi-dimensional battery status data of each battery are continuously and uninterruptedly monitored, and these data are updated and analyzed at a very high frequency. As the electrical energy is transferred, the system constantly calculates the charge difference value, which is the difference between batteries with different states of charge, until it is reduced to a safe range that meets the system's pre-set requirements.

[0053] During the entire transfer period, the intelligent charge and discharge control unit accurately controls the size and direction of the charge and discharge current and the length of time, and dynamically adjusts these parameters based on the real-time status data of the battery to ensure that the electric energy can be transferred efficiently, stably and orderly. At the same time, strict protection logic is built in. Once the battery is detected to have a risk trend of overcharge or over-discharge, corresponding restrictive measures are immediately taken, such as timely adjustment of the current size or even suspension of the transfer process, so as to resolutely avoid any potential damage to the battery and ensure the health and life of the battery.

[0054] In addition, thanks to the comprehensive and accurate information feedback provided by real-time data monitoring, the system can flexibly re-plan and adjust the transfer strategy based on the real-time changes in battery status, such as changes in battery internal resistance caused by temperature fluctuations affecting the charge status assessment, or changes in charging and discharging performance due to changes in battery aging characteristics, with its powerful intelligent algorithms and adaptive control capabilities. It performs all-round optimization from multiple aspects such as current adjustment amplitude to time optimization allocation to maximize the safety and effectiveness of the balancing process, so that the grid-type energy storage system can always operate in an efficient, stable and safe state, providing solid and reliable guarantees for the stable supply and flexible adjustment of the entire power system.

[0055] The formula for the preliminary reference instruction is:

[0056] P fi =min(P charge,i , P discharge,i )·SOH i ·f(T i )

[0057] Where P fi is the preliminary reference instruction of the ith battery cell, min(P charge,i , P discharge,i ) means taking the smaller value of the charge and discharge power, SOH i is the compensation factor for health status, f(Ti ) is the compensation factor for temperature.

[0058] S3: The decision tree model is trained with historical data obtained by integrating load forecast data to obtain a decision tree optimization model. The decision tree optimization model is used to predict power demand for future time periods to obtain load forecast results. The load forecast results are combined with preliminary reference instructions according to a preset degree of importance to generate optimal power allocation instructions.

[0059] In step S3, the step of generating an optimal power allocation instruction includes:

[0060] S31: Normalize the load forecast data and integrate them at daily time intervals to obtain historical data. The load forecast data includes historical electricity consumption data, consumer behavior research data, and external environmental factor data. The historical data includes daily total electricity consumption and average electricity consumption indicators. The historical electricity consumption data is obtained by obtaining the user's historical electricity consumption records from the power company or monitoring system, usually including electricity consumption per hour or per minute. Consumer behavior data is obtained by collecting information such as consumers' electricity consumption habits, peak and valley periods, and holiday usage patterns through surveys, questionnaires, or smart meters. External environmental data is obtained by obtaining weather data (such as temperature, humidity, air pressure) and other relevant external factors (such as holidays, special events) data.

[0061] In step S31, the formula for the average power consumption index is expressed as:

[0062]

[0063] Where D aug The average daily electricity consumption.

[0064] S32: Build a decision tree model, use historical data to train the decision tree model, and perform cross-validation to obtain a decision tree optimization model.

[0065] In step S32, the step of obtaining the decision tree optimization model includes:

[0066] S321: Divide the historical data into a training set and a test set. The training set is used for model learning and parameter adjustment, while the test set is used to evaluate the performance of the model on unseen data. In addition, in order to improve the robustness of the model, a stratified sampling method can be adopted to ensure that the proportion of each category in the training set and the test set is similar, especially in the case of imbalanced data categories.

[0067] S322: Use the machine learning library to build a decision tree model and select fitting parameters, including:

[0068] Tree depth: Limit the maximum depth of the tree to prevent overfitting and keep the model simple.

[0069] Minimum number of samples to split: Set the minimum number of samples required when splitting a node to avoid creating overly granular branches.

[0070] Minimum number of leaf node samples: Defines the minimum number of samples required in a leaf node to ensure that each leaf node represents enough samples and enhance the stability of the model.

[0071] S323: Input the training set data into the decision tree model to obtain the association between the features and the target variable, and use K-fold cross validation to verify the decision tree model to obtain the average performance evaluation of the decision tree model.

[0072] In step S323, the K-fold cross-validation method is: divide the training set into K subsets, select one subset as the validation set each time, and the rest as the basic set, and repeat K times to obtain the average performance evaluation of the decision tree model.

[0073] S324: Adjust the decision tree model parameters according to the average performance evaluation to obtain a decision tree optimization model.

[0074] S33: Use the decision tree optimization model to predict the power demand for the future time period and obtain the load prediction result.

[0075] In step S33, the steps of obtaining the load forecast result include:

[0076] S331: Based on the decision tree optimization model, a preliminary load forecast result is obtained. The trained and optimized decision tree model is used to predict future power demand, and relevant feature data of the future time period is input. The model generates a preliminary load forecast result. The preliminary result indicates the amount of power load that is expected to be used under given conditions.

[0077] S332: Denormalize the preliminary load forecast result to restore it to the original power load value, and obtain the load forecast result, which is expressed as follows:

[0078] P original =f(X)×(P max -P min )+P min

[0079] Where P original is the load forecast result, f(X) is the preliminary load forecast result, P min is the minimum value of the training set data, P max is the maximum value of the training set data.

[0080] S34: Combining the load forecast result with the preliminary reference instruction according to a preset degree of importance to generate an optimal power reference instruction.

[0081] In step S34, the formula of the optimal power reference instruction is expressed as:

[0082] P otimal =σ·P original +(1-σ)·P fi

[0083] Where P optimal is the optimal power reference instruction, σ is the emphasis on the load forecast result, and (1-σ) is the emphasis on the preliminary reference instruction.

[0084] S4: Allocate power to the battery according to the optimal power allocation instruction. First, receive the instruction and determine the current operating status of the battery to ensure that the state of charge, temperature, voltage and other parameters of each battery are within a safe range. Then, calculate the charging or discharging power of each battery according to the instruction, and control the battery's charging or discharging system accordingly for energy flow. The entire process requires real-time monitoring of the battery status to ensure that the battery will not be overcharged or over-discharged during the charging and discharging process, and dynamically adjust according to the actual operating data to optimize the overall energy management. Finally, the charging and discharging records are included in the system analysis to provide data support for subsequent decision-making.

[0085] This embodiment obtains a preliminary reference instruction for the battery by calculating the smaller value of the charging and discharging power and uses the decision tree optimization model to predict the power demand for the future time period to obtain a load forecast result, and then combines the load forecast result with the preliminary reference instruction according to a preset degree of importance to generate an optimal power allocation instruction, and uses the optimal power allocation instruction to allocate power to the battery, thereby solving the problems that the traditional system may adopt a fixed or simple rule-based power allocation strategy, cannot adapt to the real-time changing power demand, does not fully consider the impact of user behavior and external environmental factors on power demand, and lacks prediction of future power demand, thereby achieving the results of improving system efficiency, extending battery life, improving system reliability and enhancing user experience, while reducing the energy consumption of the energy storage system, which helps to achieve the beneficial effects of energy conservation and emission reduction.

[0086] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0087] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for actively balancing the SOC of a grid-type energy storage system, characterized in that: include: S1: Collecting real-time battery status data and load forecast data, wherein the real-time battery status data includes battery measurement data and performance indicators; S2: judging whether the measured data and performance indicators exceed a preset safety range according to the real-time status data of the battery, and performing active balancing operation to restore the real-time status of the battery to the preset safety range, and then calculating a preliminary reference instruction for the battery according to the smaller value of the charging and discharging power; S3: training the decision tree model with historical data obtained by integrating the load forecast data to obtain a decision tree optimization model, using the decision tree optimization model to predict power demand for future time periods to obtain load forecast results, combining the load forecast results with the preliminary reference instructions according to a preset degree of importance, and generating an optimal power allocation instruction; S4: Allocate power to the battery according to the optimal power allocation instruction.

2. The SOC active balancing control method of a grid-type energy storage system according to claim 1 is characterized in that: In step S2, the active balancing operation includes: The electric energy stored in the grid-type energy storage system is transferred from a battery with a high state of charge to a battery with a low state of charge; during the transfer process, the real-time status data of the battery is continuously monitored until the charge difference value is reduced to meet the preset safety range, and the charge difference value is the difference value between batteries with different charge states.

3. The SOC active balancing control method of a grid-type energy storage system according to claim 1 is characterized in that: In step S2, the formula of the preliminary reference instruction is expressed as: P fi =min(P charge,i ,P discharge,i )·SOH i ·f(T i ) Where P fi is the preliminary reference instruction of the ith battery cell, min(P charge,i , P discharge,i ) means taking the smaller value of the charge and discharge power, SOH i is the compensation factor for health status, f(T i ) is the compensation factor for temperature.

4. The SOC active balancing control method of a grid-type energy storage system according to claim 1 is characterized in that: In step S3, the step of generating the optimal power allocation instruction includes: S31: normalizing the load forecast data and integrating them at daily time intervals to obtain historical data, wherein the load forecast data includes historical power consumption data, consumer behavior research data, and external environmental factor data, and the historical data includes daily total power consumption and average power consumption indicators; S32: constructing a decision tree model, using the historical data to train the decision tree model, and performing cross-validation to obtain a decision tree optimization model; S33: Use the decision tree optimization model to predict the power demand for a future time period to obtain a load prediction result; S34: Combining the load forecast result with the preliminary reference instruction according to a preset degree of importance to generate the optimal power reference instruction.

5. The SOC active balancing control method of a grid-type energy storage system according to claim 4 is characterized in that: In step S31, the total daily power consumption is expressed as: Where P t is the power consumption at each time point in a day, n is the number of time points in that day, D total is the total daily electricity consumption.

6. The SOC active balancing control method of a grid-type energy storage system according to claim 5 is characterized in that: In step S31, the formula for the average power consumption index is expressed as: Where D aug The average daily electricity consumption.

7. The SOC active balancing control method of a grid-type energy storage system according to claim 5 is characterized in that: In step S32, the step of obtaining the decision tree optimization model includes: S321: Divide the historical data into a training set and a test set; S322: construct a decision tree model and select fitting parameters, including the depth of the tree, the minimum number of sample splits, and the minimum number of leaf node samples; S323: inputting the data of the training set into the decision tree model to obtain the association between the features and the target variable, and validating the decision tree model using K-fold cross validation to obtain an average performance evaluation of the decision tree model; S324: Adjust the decision tree model parameters according to the average performance evaluation to obtain the decision tree optimization model.

8. The SOC active balancing control method of a grid-type energy storage system according to claim 7 is characterized in that: In step S323, the K-fold cross-validation method is: divide the training set into K subsets, select one subset as the validation set each time, and the rest as the basic set, and repeat K times to obtain the average performance evaluation of the decision tree model.

9. The SOC active balancing control method of a grid-type energy storage system according to claim 5, characterized in that: In step S33, the step of obtaining the load forecast result includes: S331: Obtaining a preliminary load forecast result based on the decision tree optimization model; S332: Denormalize the preliminary load forecast result to restore it to the original power load value, and obtain the load forecast result, which is expressed as follows: P original =f(X)×(P max -P min )+P min Where P original is the load forecast result, f(X) is the preliminary load forecast result, P min is the minimum value of the training set data, P max is the maximum value of the training set data.

10. The SOC active balancing control method of a grid-type energy storage system according to claim 5, characterized in that: In step S34, the formula of the optimal power reference instruction is expressed as: P optimal =σ·P original +(1-σ)·P fi Where P optimal is the optimal power reference instruction, σ is the emphasis on the load forecast result, and (1-σ) is the emphasis on the preliminary reference instruction.

Images