A method for controlling the parallel operation stability of energy storage inverters

Abstract

The present application relates to the technical field of energy storage inverters, in particular to a stability control method for parallel operation of energy storage inverters, which comprises training a load prediction model based on historical load data and environmental data of a first region to obtain load prediction demand and total load demand of the first region within a first time; predicting total power generation output within the first time based on historical power generation data of renewable energy sources in the first region; storing excess power by the energy storage inverters when the total power generation output is greater than the total load demand; discharging the energy storage inverters to provide supplemental power when the total power generation output is less than the total load demand; obtaining the working state of the energy storage inverters in real time, and disconnecting the faulty energy storage inverters from the parallel group using a fault isolation mechanism when one and / or more energy storage inverters fail; obtaining the health of the remaining energy storage inverters, assigning output power to each energy storage inverter, and adjusting the power distribution strategy of the energy storage inverters in real time according to the load prediction demand.

Description

A method for stability control of parallel operation of energy storage inverters Technical Field

[0001] This invention relates to the field of energy storage inverter technology, specifically to a method for controlling the stability of parallel operation of energy storage inverters. Background Technology

[0002] In traditional power systems, electricity typically originates from centralized power plants and is dispatched and distributed through the power grid. However, in remote areas, due to the high costs of grid construction and maintenance, these regions often rely on energy storage systems and distributed energy systems for power. To ensure the efficient and stable operation of the system, parallel operation of energy storage inverters has gradually become an important means of solving this problem, especially when load demand fluctuates greatly and the supply of renewable energy is unstable.

[0003] Parallel operation of energy storage inverters can improve the reliability and stability of the system, but in practical applications, many challenges still exist. In particular, when faced with large load fluctuations, uneven load demand, and fluctuations in renewable energy power generation, existing parallel control methods for energy storage inverters are still insufficient in terms of system stability and reliability. At the same time, traditional methods for power allocation to energy storage inverters often ignore the health status of the inverters, leading to some inverters operating under overload or being idle, which affects the stability of the system.

[0004] To address this, a stability control method for parallel operation of energy storage inverters is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a stability control method for parallel operation of energy storage inverters. This method trains a load prediction model based on historical load data and environmental data of a first region to obtain the predicted load demand and total load demand of the first region within a first time period. Based on historical power generation data of renewable energy sources within the first region, it predicts the total power generation output within the first time period. When the total power generation output exceeds the total load demand, excess electricity is stored through the energy storage inverters. When the total power generation output is less than the total load demand, the energy storage inverters discharge to provide supplementary power. The method also acquires the operating status of the energy storage inverters in real time, and when one or more energy storage inverters fail, a fault isolation mechanism is used to disconnect the faulty energy storage inverter from the parallel group. Furthermore, it acquires the health status of the remaining energy storage inverters and allocates output power to each energy storage inverter. Finally, it adjusts the power allocation strategy of the energy storage inverters in real time according to the predicted load demand.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for stability control of parallel operation of energy storage inverters includes:

[0008] A load prediction model is trained based on historical load data and environmental data of the first region. The load prediction model is used to predict the load demand of the first region in the first time period, and the predicted load demand and total load demand at different time scales in the first time period are obtained.

[0009] Based on historical power generation data of renewable energy in the first region, the total power generation output for the first time period is generated.

[0010] When the total power generation output exceeds the total load demand, the excess power is stored through the energy storage inverter; when the total power generation output is less than the total load demand, the energy storage inverter discharges to provide supplementary power.

[0011] The specific process by which the energy storage inverter discharges to provide supplementary power is as follows:

[0012] The system acquires the real-time operating status of energy storage inverters and determines whether an inverter is malfunctioning based on its operating parameters and safety thresholds. When one or more energy storage inverters malfunction, a fault isolation mechanism is used to disconnect the faulty inverter from the parallel group and obtain the health status of the remaining inverters. When no inverter malfunctions, the system obtains the health status of the remaining inverters.

[0013] By assessing the health of the energy storage inverters, output power is allocated to each parallel energy storage inverter to obtain the output power matrix.

[0014] Based on the real-time acquired health status of the energy storage inverters, the output power matrix is ​​dynamically adjusted, and the output power of each energy storage inverter is adjusted accordingly.

[0015] Based on the health status of the energy storage inverters and the predicted load demand, the power allocation strategy of each energy storage inverter is adjusted in real time.

[0016] Preferably, the historical load data and environmental data are load demand, temperature, humidity, wind speed, solar radiation intensity, and rainfall within the same time period;

[0017] The power generation data refers to the power generation output data within the same time period;

[0018] The historical load data, environmental data, and power generation data include load curves and environmental curves at different time scales;

[0019] The historical load data, environmental data, and power generation data were obtained from the power company and weather station in the first region.

[0020] Preferably, the load prediction model includes a historical load data analysis unit, an environmental data analysis unit, a time feature analysis unit, and a load prediction unit;

[0021] The historical load data analysis unit analyzes historical load data through moving averages to obtain load data information; the load data information includes long-term load trends and load cycles.

[0022] The environmental data analysis unit obtains the correlation between environmental factors and load demand through correlation analysis;

[0023] The time feature analysis unit generates time tags by dividing the time window;

[0024] The load forecasting unit generates load forecasting demand by comprehensively analyzing the load data information, the correlation between environmental factors and load demand, and the time stamp, and then smooths the load forecasting demand to obtain the load demand forecast value.

[0025] Preferably, the load prediction model includes a data preprocessing unit, an LSTM model training unit, and a load prediction unit;

[0026] The data preprocessing unit includes data cleaning, standardization, and feature extraction; it extracts time features, lag features, and moving averages.

[0027] The LSTM model training unit includes an input layer, hidden layers, fully connected layers, and an output layer. The input layer consists of multi-dimensional feature vectors of historical load data and environmental data. The hidden layer includes n LSTM hidden layers, each containing m LSTM units, used to obtain complex time dependencies and nonlinear features. The fully connected layer generates the final load prediction value by connecting the output of the LSTM layers to a fully connected layer. The output layer outputs the load demand prediction values ​​for multiple future time periods.

[0028] The load prediction unit is used to obtain the total load demand and smooth the total load demand.

[0029] The total power output is obtained by predicting the power generation data using an LSTM network.

[0030] Preferably, the specific steps for determining whether an energy storage inverter has malfunctioned are as follows:

[0031] The collected operating parameters of each energy storage inverter are converted into digital signals and preprocessed, synchronized, and stored; the operating parameters include voltage, current, frequency, temperature, power factor, and harmonic content;

[0032] Anomalies are determined by comparing real-time operating parameters with preset operating parameter safety thresholds; the anomaly determination includes single-parameter anomalies and multi-parameter anomalies.

[0033] Parameter correlation analysis is performed on the anomaly determination to confirm the anomaly.

[0034] Preferably, the formula for calculating the health of an energy storage inverter is:

[0035]

[0036] Where, α V For voltage weighting, V is the current output voltage, V nominal The rated output voltage, V max V is the upper limit of voltage. min As the lower limit of voltage, α I For current weighting, I is the current output current, I max For the rated output current, α f The output frequency weight is f, where f is the current output frequency. nominal The rated frequency is Δf, which is the allowable frequency deviation range, and α is the frequency. T For temperature weighting, HI T For temperature, α PT The power factor weight is PT, where PT is the current power factor, and α is the weight. HD The harmonic distortion rate weights are α, HD is the current harmonic distortion rate, and α is the harmonic distortion rate weight. OT The weighting is the runtime, where OT is the cumulative runtime, e is the natural constant, λ is the runtime decay coefficient, and α is the runtime weighting factor. FC Here, FC represents the number of failures, μ represents the failure rate decay coefficient, and T represents the current internal temperature. max This represents the upper limit of the temperature range.

[0037] Preferably, the specific process for allocating output power to each parallel energy storage inverter is as follows:

[0038] S10. Normalize the health of each energy storage inverter to obtain the first health of the energy storage inverter;

[0039] S20. Based on the first health status, calculate the first output power allocated to each energy storage inverter;

[0040] S30. Determine whether the first output power exceeds the rated power of the energy storage inverter;

[0041] S40. If an output power does not exceed the rated power, the operation shall be performed according to the allocated power.

[0042] S50. If the first output power exceeds the rated power, obtain the second output power that exceeds the range; obtain the first energy storage inverter that does not exceed the rated power, and allocate output power to the first energy storage inverter according to the second output power;

[0043] S60. Obtain the output power of all energy storage inverters and obtain the output power matrix.

[0044] Preferably, the process of obtaining the allocated output power of each parallel energy storage inverter further includes:

[0045] The health status of each energy storage inverter is acquired in real time, and the output power matrix is ​​dynamically adjusted based on the health status.

[0046] The output power of the energy storage inverter is adjusted according to the output power matrix.

[0047] Based on the load forecast demand at different time scales output by the load forecast model in the first time period, the power allocation strategy is adjusted in advance; the adjustment strategy includes:

[0048] S100. Based on the load curves of the predicted load demand at different time scales, prioritize allocating a higher proportion of output power to the energy storage inverter with higher health.

[0049] S200. Based on the real-time health status changes of the energy storage inverters, dynamically optimize the power allocation ratio of each energy storage inverter;

[0050] S300. When predicting an increase in future load demand, it proactively increases the output power of energy storage inverters with high health.

[0051] S400. When it is predicted that the future load demand will decrease, reduce the output power of part of the energy storage inverter and perform energy storage charging operation.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] 1. This invention acquires historical load and environmental data for a first region and employs a load forecasting model based on a long short-term memory network (LSTM) to achieve high-precision load demand forecasting. By capturing the long- and short-term dependencies and nonlinear characteristics in the load data, the accuracy of the forecast is improved. Based on the load forecasting capability, the energy storage inverter can more rationally allocate output power, optimize the utilization efficiency of power resources, and improve the overall operational stability of the system.

[0054] 2. This invention provides multi-parameter integrated monitoring and health assessment. By monitoring key operating parameters of the energy storage inverter in real time, such as voltage, current, frequency, and temperature, it identifies the inverter's operating status and potential faults. Multi-parameter analysis improves the accuracy and response speed of fault detection, reducing the incidence of false alarms and missed alarms. When an energy storage inverter malfunctions, an automatic fault isolation mechanism is triggered to quickly disconnect the faulty inverter from the parallel group, ensuring that other healthy inverters can adjust their power output in a timely manner, thus improving the stability of continuous operation.

[0055] 3. This invention achieves load balancing and optimized allocation among energy storage inverters through dynamic assessment of their health. Output power is allocated based on the health status of each inverter, thereby extending the inverter's lifespan and improving the overall performance of the energy storage inverter. Furthermore, by combining load forecasting results, the power allocation strategy is adjusted in advance to optimize response time and resource utilization, further enhancing the operating efficiency and stability of the energy storage system in complex power environments. Attached Figure Description

[0056] Figure 1 is a schematic flowchart of a method for controlling the stability of parallel operation of an energy storage inverter provided by the present invention.

[0057] Figure 2 is a flowchart of the stability of parallel operation of the energy storage inverter provided in an embodiment of the present invention;

[0058] Figure 3 is a flowchart of the process of allocating output power to an energy storage inverter according to an embodiment of the present invention. Detailed Implementation

[0059] 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.

[0060] Example 1

[0061] As one embodiment of the present invention, referring to Figures 1 and 2, the present invention provides a method for stability control of parallel operation of energy storage inverters, the technical solution of which is as follows:

[0062] A load prediction model is trained based on historical load data and environmental data of the first region. The load prediction model is used to predict the load demand of the first region in the first time period to obtain the predicted load demand and the total load demand in the first time period. The first time period is consistent with the time range of the predicted load demand.

[0063] Based on historical power generation data of renewable energy in the first region, the total power generation output is predicted and obtained in the first time period.

[0064] Furthermore, the historical load data and environmental data are load demand, temperature, humidity, wind speed, solar radiation intensity, and rainfall within the same time period;

[0065] The power generation data refers to the power generation output data within the same time period;

[0066] The historical load data, environmental data, and power generation data include load curves and environmental curves at different time scales; the load curves at different time scales include hourly, daily, and monthly load curves.

[0067] The historical load data, environmental data, and power generation data were obtained from the power company and weather station in the first region.

[0068] In this embodiment, the present invention fuses data from multiple time scales, including hours, days, and months, capturing both short-term fluctuations and long-term trends to ensure the comprehensiveness of the prediction results. Furthermore, the use of historical operational and meteorological data obtained from power companies and weather stations ensures the accuracy of the data, laying the foundation for model training and health status assessment of energy storage inverters.

[0069] Furthermore, the load prediction model includes a historical load data analysis unit, an environmental data analysis unit, a time characteristic analysis unit, and a load prediction unit;

[0070] The historical load data analysis unit analyzes historical load data through moving averages to obtain load data information; the load data information includes long-term load trends and load cycles.

[0071] The environmental data analysis unit obtains the correlation between environmental factors and load demand through correlation analysis;

[0072] The time feature analysis unit generates time tags by dividing the time window;

[0073] The load forecasting unit generates load forecasting demand by comprehensively analyzing the load data information, the correlation between environmental factors and load demand, and the time stamp, and then smooths the load forecasting demand to obtain the load demand forecast value.

[0074] In this invention, the load prediction model divides data analysis into different functional units, which facilitates system integration; by integrating multiple data sources such as historical load, environmental factors and time characteristics, the prediction model's ability to process multi-dimensional information is improved; and the prediction results are smoothed to effectively reduce noise interference caused by short-term fluctuations, making subsequent scheduling more stable and reliable.

[0075] Furthermore, the load prediction model includes a data preprocessing unit, an LSTM model training unit, and a load prediction unit;

[0076] The data preprocessing unit includes data cleaning, standardization, and feature extraction; it extracts time features, lag features, and moving averages.

[0077] The LSTM model training unit includes an input layer, hidden layers, fully connected layers, and an output layer. The input layer contains multi-dimensional feature vectors of historical load data and environmental data. The hidden layer includes two LSTM hidden layers, each containing 50 LSTM units, used to obtain complex time dependencies and nonlinear features. The fully connected layer generates the final load prediction value by connecting the output of the LSTM layers to a fully connected layer. The output layer outputs the load demand prediction values ​​for multiple future time points. In this embodiment, n is 2 and m is 50.

[0078] The load prediction unit is used to obtain the total load demand and smooth the total load demand.

[0079] The total power output is obtained by predicting the power generation data using an LSTM network.

[0080] To further illustrate the advantages of the above LSTM model in load forecasting, Table 1 below shows the comparison results of the present invention with other traditional forecasting methods in terms of forecasting accuracy and computational performance.

[0081] Table 1 Load Forecast Accuracy

[0082] Method MAPE (%) Prediction Time (s) Invention 2.1 0.15 Linear Regression 7.9 0.06 Traditional RNN 5.6 0.08 surface

[0083] As shown in Table 1, under the same dataset and hardware testing environment, the LSTM model used in this invention has a significant improvement in prediction accuracy, while the prediction time remains within an acceptable range.

[0084] Determine if the total power generation output exceeds the total load demand; when the total power generation output exceeds the total load demand, store the excess power through the energy storage inverter; when the total power generation output is less than the total load demand, provide supplementary power by discharging through the energy storage inverter.

[0085] The specific process by which the energy storage inverter discharges to provide supplementary power is as follows:

[0086] The system acquires the operating status of the energy storage inverter in real time and determines whether the energy storage inverter has a fault. When one or more energy storage inverters fail, a fault isolation mechanism is used to disconnect the faulty energy storage inverter from the parallel group and obtain the health status of the remaining energy storage inverters. When no energy storage inverters fail, the system obtains the health status of the energy storage inverters.

[0087] Furthermore, the specific steps to determine if an energy storage inverter has malfunctioned are as follows:

[0088] The collected operating parameters of each energy storage inverter are converted into digital signals and preprocessed, synchronized, and stored; the operating parameters include voltage, current, frequency, temperature, power factor, and harmonic content;

[0089] Anomalies are determined by comparing real-time operating parameters with preset operating parameter safety thresholds; the anomaly determination includes single-parameter anomalies and multi-parameter anomalies.

[0090] Parameter correlation analysis is performed on the anomaly determination to confirm the anomaly.

[0091] The safety thresholds are referenced from the technical manuals of energy storage inverter manufacturers and industry standards, and are also statistically optimized based on long-term operational data of this system in remote areas. First, the upper and lower limits of each parameter are initially determined based on standards such as rated voltage, upper temperature limit, and harmonic content. Then, by combining a large amount of historical data collected during actual operation, the fluctuation ranges of each parameter under normal and fault conditions are compared and analyzed to calibrate the accuracy of the thresholds. If necessary, multi-level warning and alarm ranges can be set for key parameters such as voltage, current, and temperature. For example, a 5% deviation of the voltage from the rated value triggers a minor alarm, while a deviation exceeding 10% is considered a serious anomaly. For harsh operating environments or environments with large load fluctuations, the safety thresholds can be updated periodically through an online learning mechanism to ensure accurate identification and timely warning of potential faults even as the inverter ages or the external environment changes.

[0092] Compared to traditional methods that only monitor voltage or current, this invention monitors multiple dimensions of information, including voltage, current, frequency, temperature, power factor, and harmonic content, making fault identification more accurate. Furthermore, by determining anomalies through single or multiple parameters and confirming faults through correlation analysis, it effectively reduces false alarms and missed alarms, thereby enhancing system safety and fault self-healing capabilities.

[0093] To further illustrate the advantages of the multi-parameter fault detection method in terms of accuracy and false alarm rate, Table 2 below provides comparative data between the present invention and different single-parameter or dual-parameter detection methods.

[0094] Table 2 Fault Detection Sensitivity

[0095] Fault Detection Method Detection Rate (%) False Alarm Rate (%) This Invention (Multi-Parameter Fault Detection) 98.0 1.0 Voltage monitoring only 81.3 3.9 Voltage and current monitoring 88.2 2.4 Output power monitoring only 85.7 3.1 surface

[0096] As shown in Table 2, the multi-parameter monitoring of the present invention has better performance in terms of detection rate and false alarm rate compared with single-parameter or dual-parameter monitoring, which greatly improves the accuracy and timeliness of fault identification.

[0097] Furthermore, the formula for calculating the health of an energy storage inverter is as follows:

[0098]

[0099] Where, α V For voltage weighting, V is the current output voltage, V nominal The rated output voltage, V max V is the upper limit of voltage. min As the lower limit of voltage, α I For current weighting, I is the current output current, I max For the rated output current, α f The output frequency weight is f, where f is the current output frequency. nominal The rated frequency is Δf, which is the allowable frequency deviation range, and α is the frequency. T For temperature weighting, HI T For temperature, α PT The power factor weight is PT, where PT is the current power factor, and α is the weight. HD The harmonic distortion rate weights are α, HD is the current harmonic distortion rate, and α is the harmonic distortion rate weight. OT The weighting is the runtime, where OT is the cumulative runtime, e is the natural constant, λ is the runtime decay coefficient, and α is the runtime weighting factor. FC Here, FC represents the number of failures, μ represents the failure rate decay coefficient, and T represents the current internal temperature. max This represents the upper limit of the temperature range.

[0100] The health calculation formula for energy storage inverters provided by this invention quantifies the health status of each energy storage inverter, providing a scientific basis for subsequent dynamic power allocation; and by weighting and dynamically adjusting key parameters for different scenarios, the health assessment process becomes more adaptable.

[0101] The output power is allocated to each parallel energy storage inverter based on the health status of the energy storage inverter; and the power allocation strategy of the energy storage inverter is adjusted in real time according to the predicted load demand.

[0102] Furthermore, the specific process for allocating output power to each parallel energy storage inverter is shown in Figure 3;

[0103] S10. Normalize the health of each energy storage inverter to obtain the first health of the energy storage inverter;

[0104] S20. Based on the first health status, calculate the first output power allocated to each energy storage inverter;

[0105] S30. Determine whether the first output power exceeds the rated power of the energy storage inverter;

[0106] S40. If an output power does not exceed the rated power, the operation shall be performed according to the allocated power.

[0107] S50. If the first output power exceeds the rated power, obtain the second output power that exceeds the range; obtain the first energy storage inverter that does not exceed the rated power, and allocate output power to the first energy storage inverter according to the second output power;

[0108] S60. Obtain the output power of all energy storage inverters and obtain the output power matrix.

[0109] This invention can reduce the load or reserve capacity of inverters with high health status by allocating more power to inverters with low health status, thereby effectively improving the overall system efficiency and extending the equipment life. When a new energy storage inverter is added or a faulty unit is removed, the system can automatically complete the normalization and redistribution.

[0110] Furthermore, the process of obtaining the allocated output power of each parallel energy storage inverter also includes:

[0111] The health status of each energy storage inverter is acquired in real time, and the output power matrix is ​​dynamically adjusted based on the health status.

[0112] The output power of the energy storage inverter is adjusted according to the output power matrix.

[0113] Based on the load forecast demand at different time scales output by the load forecast model in the first time period, the power allocation strategy is adjusted in advance; the adjustment strategy includes:

[0114] S100. Based on the load curves of the predicted load demand at different time scales, prioritize allocating a higher proportion of output power to the energy storage inverter with higher health.

[0115] S200. Based on the real-time health status changes of the energy storage inverters, dynamically optimize the power allocation ratio of each energy storage inverter to ensure the overall efficiency and stability of the system operation;

[0116] S300. When a significant increase in future load demand is predicted, the output power of a healthy energy storage inverter is increased in advance to meet the growth in load demand.

[0117] S400. When it is predicted that the future load demand will decrease, the output power of some energy storage inverters will be reduced or energy storage charging operations will be performed to optimize energy utilization efficiency.

[0118] In this embodiment, by combining real-time health status and prediction models, the output power matrix can be automatically corrected during system operation, improving the ability to adapt to sudden events and demand fluctuations; based on the predicted load and power generation, the system can perform advance scheduling to avoid energy waste or load collapse caused by delayed response.

[0119] This invention achieves high-precision prediction of load demand and renewable energy generation by combining Long Short-Term Memory (LSTM) networks, and introduces multi-parameter fault detection and inverter health assessment mechanisms into the parallel operation control of energy storage inverters. First, the LSTM model improves the accuracy of load and power generation predictions because it can capture the nonlinearity and long-short-term dependencies in time-series data. Second, addressing the diverse fault risks often faced by inverters in remote areas, this invention monitors and performs correlation analysis on multi-dimensional parameters such as voltage, current, frequency, temperature, and harmonics to identify single or multiple parameter anomalies and quickly isolates faulty inverters after a fault occurs, minimizing the impact on the system. Furthermore, the overall operating status of the inverters is quantified through a health calculation formula, and a flexible power allocation strategy is implemented based on the health status of each inverter, allocating smaller loads or reserving power to inverters with low health status, thereby improving system efficiency and extending equipment lifespan. Finally, by acquiring health status in real time and dynamically adjusting power based on load forecast demand, this invention achieves high efficiency and safety of energy storage systems even in remote areas with limited resources and complex operating environments.

[0120] Example 2

[0121] A load prediction model is trained based on historical load data and environmental data of the first region. The load prediction model is used to predict the load demand of the first region in the first time period to obtain the predicted load demand and the total load demand in the first time period.

[0122] Based on historical power generation data of renewable energy in the first region, the total power generation output is predicted and obtained in the first time period.

[0123] Determine if the total power generation output exceeds the total load demand; when the total power generation output exceeds the total load demand, store the excess power through the energy storage inverter; when the total power generation output is less than the total load demand, provide supplementary power by discharging through the energy storage inverter.

[0124] The specific process by which the energy storage inverter discharges to provide supplementary power is as follows:

[0125] The system acquires the operating status of the energy storage inverter in real time and determines whether the energy storage inverter has a fault. When one or more energy storage inverters fail, a fault isolation mechanism is used to disconnect the faulty energy storage inverter from the parallel group and obtain the health status of the remaining energy storage inverters. When no energy storage inverters fail, the system obtains the health status of the energy storage inverters.

[0126] Furthermore, the specific steps to determine if an energy storage inverter has malfunctioned are as follows:

[0127] The collected operating parameters of each energy storage inverter are converted into digital signals and preprocessed, synchronized, and stored; the operating parameters include voltage, current, frequency, temperature, power factor, and harmonic content;

[0128] Anomalies are determined by comparing real-time operating parameters with preset operating parameter safety thresholds; the anomaly determination includes single-parameter anomalies and multi-parameter anomalies.

[0129] Parameter correlation analysis is performed on the anomaly determination to confirm the anomaly.

[0130] Furthermore, the formula for calculating the health of an energy storage inverter is as follows:

[0131]

[0132] Where, α V For voltage weighting, V is the current output voltage, V nominal The rated output voltage, V max V is the upper limit of voltage. min As the lower limit of voltage, α I For current weighting, I is the current output current, I max For the rated output current, α f The output frequency weight is f, where f is the current output frequency. nominal The rated frequency is Δf, which is the allowable frequency deviation range, and α is the frequency. T For temperature weighting, HI T For temperature, α PT The power factor weight is PT, where PT is the current power factor, and α is the weight. HD The harmonic distortion rate weights are α, HD is the current harmonic distortion rate, and α is the harmonic distortion rate weight. OT The weighting is based on runtime, where OT is the cumulative runtime, λ is the runtime decay coefficient, and α is the runtime weighting factor. FC Here, FC represents the number of failures, μ represents the failure rate decay coefficient, and T represents the current internal temperature. max This represents the upper limit of the temperature range.

[0133] In this embodiment, the health status of the five energy storage inverters is obtained by acquiring and normalizing the data. For specific data, please refer to Table 3.

[0134] Table 3 Health Status of Energy Storage Inverters

[0135] Sample Number | Voltage | Current | Frequency | Temperature | Harmonics | Health Status 1 | 0.95 | 0.88 | 0.98 | 0.90 | 0.94 | 0.93 | 2 | 0.96 | 0.85 | 0.97 | 0.87 | 0.95 | 0.91 | 3 | 0.94 | 0.90 | 0.99 | 0.88 | 0.92 | 0.92 | 4 | 0.92 | 0.87 | 0.96 | 0.85 | 0.90 | 0.89 | 5 | 0.93 | 0.91 | 0.97 | 0.86 | 0.93 | 0.91 surface

[0136] By assessing the health of the energy storage inverters, output power is allocated to each parallel energy storage inverter to obtain the output power matrix.

[0137] Based on the real-time acquired health status of the energy storage inverters, the output power matrix is ​​dynamically adjusted, and the output power of each energy storage inverter is adjusted; based on the health status of the energy storage inverters and the predicted load demand, the power allocation strategy of each energy storage inverter is adjusted in real time.

[0138] Furthermore, the specific process for allocating output power to each parallel energy storage inverter is as follows:

[0139] S10. Normalize the health of each energy storage inverter to obtain the first health of the energy storage inverter;

[0140] S20. Based on the first health status, calculate the first output power allocated to each energy storage inverter;

[0141] The formula for obtaining the first output power is:

[0142]

[0143] Among them, P i Let HI be the first output power corresponding to the i-th energy storage inverter. i Let P be the health status of the i-th energy storage inverter, N be the total number of energy storage inverters, and P be the health status of the i-th energy storage inverter. total Total output power;

[0144] S30. Determine whether the first output power exceeds the rated power of the energy storage inverter;

[0145] S40. If an output power does not exceed the rated power, the operation shall be performed according to the allocated power.

[0146] S50. If the first output power exceeds the rated power, obtain the second output power that exceeds the range; obtain the first energy storage inverter that does not exceed the rated power, and allocate output power to the first energy storage inverter according to the second output power;

[0147] S60. Obtain the output power of all energy storage inverters and obtain the output power matrix.

[0148] Power is allocated to the five energy storage inverters provided in Table 3, with data referenced in Table 4;

[0149] As can be seen from Table 4, inverters with high health ratings can handle a larger power distribution with almost no overload; inverters with low health ratings can handle a relatively smaller power distribution or are prone to overload when the system pressure increases, thus highlighting the rationality and safety of the health-driven distribution strategy.

[0150] Table 4 Comparison of Power Allocation Efficiency Driven by Health Status

[0151] Sample Number | Health Status | Power Allocation | Average Load Rate | Overload Count | System Efficiency | 10.93 | 15.28 | 0.10 | 95.22 | 0.91 | 14.77 | 8.30 | 94.83 | 0.92 | 15.07 | 9.50 | 95.04 | 0.89 | 13.87 | 5.21 | 92.75 | 0.91 | 14.57 | 7.90 | 94.3 surface

[0152] Furthermore, the process of obtaining the allocated output power of each parallel energy storage inverter also includes:

[0153] The health status of each energy storage inverter is acquired in real time, and the output power matrix is ​​dynamically adjusted to adjust the output power of the energy storage inverter.

[0154] Based on the load demand forecast values ​​output by the load forecast model, the power allocation strategy is adjusted in advance.

[0155] This invention improves the operational stability and efficiency of parallel energy storage inverter systems through multi-parameter health calculation, fault detection, and LSTM load prediction. It utilizes historical load and environmental data to construct a high-precision prediction model, enabling accurate forecasting of demand across multiple time scales; it allocates power based on inverter health to avoid overload or fault escalation; and it automatically isolates and redistributes power in the event of a fault, ensuring continuous power supply. Its comprehensive application of threshold adaptation and multi-parameter correlation analysis effectively reduces false alarms and missed alarms, enhances fault self-healing capabilities and renewable energy utilization, making it suitable for power supply scenarios in remote areas.

[0156] 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 equivalents.

Claims

1. A method for stability control of parallel operation of energy storage inverters, characterized in that, include: A load prediction model is trained based on historical load and environmental data of the first region. This model then predicts the load demand of the first region within a specific time period, obtaining the predicted load demand and total load demand at different time scales within that time period. Based on historical renewable energy generation data within the first region, the total power generation output for the first time period is generated. When the total power generation output exceeds the total load demand, excess electricity is stored through energy storage inverters. When the total power generation output is less than the total load demand, supplementary power is provided by discharging through energy storage inverters. The process of providing supplementary power through energy storage inverter discharge involves: real-time acquisition of the operating status of the energy storage inverters; determining the fault status of the energy storage inverters based on their operating parameters and safety thresholds; and when one or more energy storage inverters fail, a fault isolation mechanism is used to disconnect the faulty energy storage inverter from the parallel group, and the health status of the remaining energy storage inverters is then assessed. The process involves: When no fault occurs in the energy storage inverter, the health status of the energy storage inverter is obtained; based on the health status, output power is allocated to each parallel energy storage inverter to obtain an output power matrix; the specific process for allocating output power to each parallel energy storage inverter is as follows: S10. Normalize the health status of each energy storage inverter to obtain its first health status; S20. Calculate the first output power allocated to each energy storage inverter based on the first health status; S30. Determine whether the first output power exceeds the rated power of the energy storage inverter; S40. If the first output power does not exceed the rated power, proceed according to the allocated power; S50. If the first output power exceeds the rated power, obtain the second output power that exceeds the range; obtain the first energy storage inverter that does not exceed the rated power, and allocate output power to the first energy storage inverter based on the second output power; S60. Obtain the output power of all energy storage inverters and obtain the output power matrix; dynamically adjust the output power matrix based on the real-time obtained energy storage inverter health status, and adjust the output power of each energy storage inverter; adjust the power allocation strategy of each energy storage inverter in real time based on the energy storage inverter health status and the predicted load demand.

2. The method for stability control of parallel operation of energy storage inverters according to claim 1, characterized in that: The historical load data and environmental data are load demand, temperature, humidity, wind speed, solar radiation intensity, and rainfall within the same time period; the power generation data are power generation output data within the same time period; the historical load data, environmental data, and power generation data include load curves and environmental curves at different time scales; the historical load data, environmental data, and power generation data are obtained from the power company and meteorological station in the first region.

3. The method for stability control of parallel operation of energy storage inverters according to claim 1, characterized in that: The load forecasting model includes a historical load data analysis unit, an environmental data analysis unit, a time feature analysis unit, and a load forecasting unit. The historical load data analysis unit analyzes historical load data using moving averages to obtain load data information, including long-term load trends and load cycles. The environmental data analysis unit obtains the correlation between environmental factors and load demand through correlation analysis. The time feature analysis unit generates time tags by dividing time windows. The load forecasting unit generates load forecast demand by comprehensively analyzing the load data information, the correlation between environmental factors and load demand, and the time tags, and then smooths the load forecast demand to obtain the predicted load demand value.

4. The method for stability control of parallel operation of energy storage inverters according to claim 1, characterized in that: The load forecasting model includes a data preprocessing unit, an LSTM model training unit, and a load forecasting unit. The data preprocessing unit includes data cleaning, standardization, and feature extraction; it extracts time features, lag features, and moving averages. The LSTM model training unit includes an input layer, hidden layers, fully connected layers, and an output layer. The input layer takes into account multi-dimensional feature vectors from historical load data and environmental data. The hidden layer includes n LSTM hidden layers, each containing m LSTM units. The fully connected layer generates the final load forecast value by connecting the outputs of the LSTM layers to a fully connected layer. The output layer outputs predicted load demand values ​​for multiple future time periods. The load forecasting unit is used to obtain the total load demand and smooth it. The total power generation output is obtained by predicting power generation data using an LSTM network.

5. The method for stability control of parallel operation of energy storage inverters according to claim 1, characterized in that: The specific steps for determining a fault in an energy storage inverter are as follows: The collected operating parameters of each energy storage inverter are converted into digital signals and preprocessed, synchronized, and stored; the operating parameters include voltage, current, frequency, temperature, power factor, and harmonic content; anomalies are determined by comparing real-time operating parameters with preset safety thresholds; anomaly determination includes single-parameter anomalies and multi-parameter anomalies; parameter correlation analysis is performed on the anomaly determinations to confirm the anomalies.

6. The method for stability control of parallel operation of energy storage inverters according to claim 1, characterized in that: The formula for calculating the health of an energy storage inverter is: ； ;in, For voltage weighting, This is the current output voltage. This is the rated output voltage. This is the upper limit of voltage. This is the lower limit of voltage. Weighted by current, This is the current output current. Rated output current, For output frequency weights, The current output frequency, For the rated frequency, For the allowable frequency deviation range, Temperature weighting, For temperature, For power factor weights, The current power factor, For harmonic distortion rate weights, This represents the current harmonic distortion rate. Weighted by runtime, For cumulative running time, It is a natural constant. This is the runtime decay coefficient. As the weight of the number of failures, Number of failures This is the failure rate attenuation coefficient. The current internal temperature. This represents the upper limit of the temperature range.

7. The method for stability control of parallel operation of energy storage inverters according to claim 1, characterized in that: The process of obtaining the allocated output power for each parallel energy storage inverter also includes: obtaining the health status of each energy storage inverter in real time, and dynamically adjusting the output power matrix based on the health status; adjusting the output power of the energy storage inverters based on the output power matrix; and adjusting the power allocation strategy in advance based on the load forecast demand at different time scales in the first time period output by the load forecast model. The adjustment strategy includes: S100. Prioritizing the allocation of a higher proportion of output power to energy storage inverters with higher health status based on the load curves of the load forecast demand at different time scales; S200. Dynamically optimizing the power allocation ratio of each energy storage inverter based on the real-time health status changes of the energy storage inverters; S300. In advance increasing the output power of energy storage inverters with higher health status when it is predicted that future load demand will increase; S400. In advance reducing the output power of some energy storage inverters and performing energy storage charging operations when it is predicted that future load demand will decrease.

Images