Micro-grid real-time scheduling system and method based on big data analysis

Abstract

The present application relates to the field of micro-grid operation control and power electronics, in particular to a micro-grid real-time scheduling system and method based on big data analysis; the method is executed by a local controller in the micro-grid system, comprising: receiving and storing the droop coefficient matrix issued by the cloud server; obtaining the real-time grid frequency and the real-time battery surface temperature; when the frequency deviates from the rated value, based on the frequency deviation and the real-time battery surface temperature, interpolation calculation is performed in the droop coefficient matrix to generate transient droop coefficient, adjust the pulse width modulation signal duty cycle of the inverter module, to change the active current amplitude injected into the micro-grid; and after scheduling, according to the frequency recovery time and the average temperature rise rate of the battery, it is judged whether to generate a heat penalty flag bit and send it to the cloud server, which is used to correct the next generated droop coefficient matrix; the present application takes into account the frequency stability and the thermal safety of energy storage.

Description

Technical Field

[0001] This invention relates to the field of microgrid operation control and power electronics technology, specifically to a real-time microgrid scheduling system and method based on big data analysis. Background Technology

[0002] With the development of distributed power sources, energy storage devices, and park-level microgrid technologies, microgrids are increasingly widely used in industrial parks, commercial parks, and other scenarios. In order to achieve stable operation of microgrids and coordinated control of source, load, and storage, higher requirements are placed on the ability to quickly dispatch and support frequencies under disturbance conditions.

[0003] In the operation and control of microgrids, droop control is widely used due to its simplicity and fast response. However, existing scheduling methods often use fixed droop parameters or rely on centralized optimization in the cloud to issue a single control value, which makes it difficult to simultaneously take into account the randomness of load fluctuations, real-time changes in frequency deviation, and thermal safety constraints of energy storage units. Especially in cases of sudden load increases or abrupt changes in distributed power output, if the control parameters are not adjusted in time, frequency recovery may be delayed. If the support strength of energy storage is increased indiscriminately, it may lead to problems such as excessively rapid battery temperature rise and increased thermal burden. In addition, when communication is interrupted for a short time, the topology changes, or measurement data is abnormal, existing solutions are also prone to insufficient local adjustment adaptability.

[0004] Therefore, analyzing existing historical load power data of microgrids, combining the temperature status and frequency disturbance characteristics of energy storage units, generating scheduling parameters that can be quickly called locally, and ensuring frequency stability while taking into account the thermal safety of energy storage and the operational reliability under abnormal conditions are crucial to ensuring the safe and efficient operation of microgrids. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time microgrid scheduling system and method based on big data analysis, and to solve the following technical problems:

[0006] It avoids the microgrid's dependence on high-frequency communication and achieves local millisecond-level fast disturbance response, while effectively balancing the frequency stability of the microgrid and the thermal safety of the energy storage unit.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A real-time dispatching method for microgrids based on big data analytics is applied to a microgrid system, which includes a local controller and a cloud server, energy storage unit, inverter module, and circuit breaker communicatively connected to the local controller. The method is executed by the local controller and includes:

[0009] The cloud server receives and stores the droop coefficient matrix, which is generated by the cloud server based on the historical load power records of the microgrid. The droop coefficient matrix contains a three-dimensional mapping relationship between the droop coefficient, the cell surface temperature and the frequency deviation.

[0010] The preset rated frequency of the microgrid, the real-time grid frequency, and the real-time cell surface temperature of the energy storage unit are obtained.

[0011] Based on the real-time power grid frequency, determine whether to trigger a local interpolation scheduling operation;

[0012] The local interpolation scheduling operation includes: calculating the frequency deviation between the real-time power grid frequency and the preset rated frequency; and performing interpolation calculation in the droop coefficient matrix based on the frequency deviation and the real-time cell surface temperature to generate a transient droop coefficient.

[0013] Based on the transient droop coefficient, a pulse width modulation signal is generated and the duty cycle of the pulse width modulation signal is adjusted and sent to the inverter module to change the active current amplitude injected into the microgrid by the inverter module.

[0014] After performing a single local interpolation scheduling operation, the frequency recovery time from the triggering of the local interpolation scheduling operation to the recovery of the real-time power grid frequency to the preset rated frequency is obtained, and the average temperature rise rate of the battery cells during the scheduling period corresponding to the frequency recovery time is calculated to determine whether a thermal penalty flag needs to be generated and sent to the cloud server to instruct the cloud server to correct the droop coefficient matrix generated next time.

[0015] Furthermore, the droop coefficient matrix issued by the cloud server is generated in the following way:

[0016] Apply a sliding window with a preset time width to the historical load power records of the microgrid, and calculate the variance of active power within the sliding window;

[0017] The variance is matched and compared with each preset variance threshold interval in the preset variance threshold set to determine the preset variance threshold interval in which the variance is located, so as to determine the fluctuation level corresponding to the sliding window.

[0018] Based on the preset droop coefficient corresponding to the fluctuation level, and combined with the preset battery internal resistance thermodynamic model that maps active power to predicted temperature rise increment, the corresponding droop coefficient matrix is ​​generated; wherein, the input parameters of the preset battery internal resistance thermodynamic model include the Joule heat power generated by the equivalent charge and discharge current, the time span of the sliding window, the equivalent thermal constant of the cell, the current cell surface temperature, the ambient temperature, and the equivalent thermal resistance of the system.

[0019] Furthermore, the step of determining whether to trigger a local interpolation scheduling operation based on the real-time power grid frequency includes:

[0020] Calculate the absolute value of the frequency deviation between the real-time power grid frequency and the preset rated frequency;

[0021] Determine whether the absolute value of the frequency deviation is greater than a preset frequency fluctuation threshold;

[0022] If the absolute value of the frequency deviation is greater than or equal to the frequency fluctuation threshold, the local interpolation scheduling operation is triggered; if the absolute value of the frequency deviation is less than the frequency fluctuation threshold, the local interpolation scheduling operation is not triggered.

[0023] Further, the step of generating the transient droop coefficient by performing interpolation calculations in the droop coefficient matrix based on the frequency deviation and the real-time cell surface temperature includes:

[0024] Extract at least 6 matrix node data adjacent to the current frequency deviation and the real-time cell surface temperature from the droop coefficient matrix;

[0025] The matrix node data is used to execute a surface fitting interpolation algorithm based on a quadratic polynomial to calculate the corresponding transient droop coefficient.

[0026] The transient droop coefficient is sent to the inverter module to reconstruct the sinusoidal modulation waveform.

[0027] Furthermore, the step of determining whether a hot penalty flag needs to be generated and sent to the cloud server includes:

[0028] Determine whether the frequency recovery time is less than a preset recovery time threshold, and determine whether the average temperature rise rate of the battery is greater than a preset temperature rise safety threshold;

[0029] If the frequency recovery time is less than or equal to the preset recovery time threshold and the average battery temperature rise rate is greater than or equal to the preset temperature rise safety threshold, then it is determined that the heat penalty flag needs to be generated and sent to the cloud server; otherwise, it is determined that the heat penalty flag does not need to be generated.

[0030] Furthermore, instructing the cloud server to correct the droop coefficient matrix generated next time means instructing the cloud server to perform the following correction operation:

[0031] When generating the droop coefficient matrix again, extract the target temperature range from the droop coefficient matrix where the temperature dimension is greater than or equal to the real-time cell surface temperature when the heat penalty flag is triggered.

[0032] The droop coefficient mapping value corresponding to the target temperature range in the droop coefficient matrix is ​​reduced to limit the subsequent active power output amplitude of the energy storage unit.

[0033] Furthermore, it also includes:

[0034] Monitor the switching status of circuit breakers installed in the microgrid to obtain the topology status;

[0035] Based on the transition of the switch state, determine whether the topology state has undergone physical reconstruction;

[0036] If the topology state undergoes physical reconstruction and is within a preset transient transition period, it will be forcibly switched to constant voltage and constant frequency islanding control mode; otherwise, the local interpolation scheduling operation will be maintained.

[0037] The microgrid real-time dispatch system based on big data analysis includes a cloud server, a local controller, and an energy storage unit and an inverter module that are communicatively connected to the local controller.

[0038] The cloud server is used to generate a droop coefficient matrix based on the historical load power records of the microgrid and send it to the local controller; and to receive a hot penalty flag sent by the local controller to correct the droop coefficient matrix generated next time.

[0039] The local controller includes a processor and a memory. The memory is used to store the droop coefficient matrix and a computer program. When the processor executes the computer program, it implements the steps of the microgrid real-time scheduling method based on big data analysis.

[0040] The beneficial effects of this invention are:

[0041] 1. This invention generates a three-dimensional droop coefficient matrix containing temperature and frequency deviations in the cloud and sends it to the local controller, enabling the local controller to achieve interpolation scheduling based on real-time frequency and cell surface temperature without high-frequency communication, and quickly respond to frequency disturbances; it generates a thermal penalty flag bit by combining recovery time and temperature rise rate to correct the cloud matrix, ensuring the frequency recovery speed while taking into account the thermal safety of the energy storage unit, and avoiding the problem of increased thermal burden caused by excessive battery temperature rise.

[0042] 2. This invention introduces a sliding window analysis based on the variance of active power in historical loads to determine the load fluctuation level and combines it with the battery internal resistance thermodynamic model to predict the temperature rise. This mechanism breaks the limitations of fixed droop empirical parameters, enabling the generated droop coefficient matrix to truly reflect the time-varying characteristics of load fluctuations and the energy storage heating constraints, thereby improving the adaptability of local scheduling parameters to actual field conditions from the source.

[0043] 3. This invention compares the absolute value of the deviation between the real-time grid frequency and the rated frequency with a preset frequency fluctuation threshold, and triggers local interpolation scheduling only when the deviation exceeds the threshold. This effectively filters out normal load disturbances and measurement noise, and avoids the inverter from frequently starting ineffective regulation due to slight natural fluctuations. While achieving rapid response, it significantly reduces the risk of ineffective microcirculation and thermal stress accumulation in the energy storage unit.

[0044] 4. This invention extracts adjacent node data in the matrix and uses a surface fitting interpolation algorithm of quadratic polynomial to solve the transient droop coefficient, avoiding step-like changes in control parameters when the state crosses intervals; this mechanism enables the energy storage support strength to change smoothly and continuously with the surface temperature and frequency deviation of the cell, ensuring the continuity of inverter control, effectively reducing current surges, and improving power quality during the frequency recovery period.

[0045] 5. This invention uses a dual judgment based on frequency recovery time and average battery temperature rise rate, generating a thermal penalty flag only when the frequency recovery is fast and the temperature rise rate exceeds the limit. This mechanism incorporates grid stability and battery thermal safety into the same evaluation closed loop, accurately identifies over-limit active power dispatching behavior that exchanges high thermal load increments for effective support, eliminates misjudgments of abnormal and inefficient equipment, and provides a reliable basis for cloud-based adaptive correction control strategies.

[0046] 6. After receiving the heat penalty flag bit in the cloud, the present invention only reduces the droop coefficient mapping value in the target temperature range above the trigger flag bit. This interval correction avoids the damage to the normal support capability under low temperature healthy state caused by the global down-adjustment, so that the energy storage can automatically switch to the mild power output mode when facing the same disturbance in the subsequent high temperature scenario, and realize the active power output convergence and battery life protection under high temperature conditions. Attached Figure Description

[0047] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 This is a flowchart illustrating the real-time microgrid scheduling method based on big data analysis in an embodiment of this application.

[0049] Figure 2 This is a schematic diagram of the structure of a microgrid real-time dispatching system based on big data analysis provided in an embodiment of this application. Detailed Implementation

[0050] 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.

[0051] Please see Figure 1 A real-time dispatching method for microgrids based on big data analytics is applied to a microgrid system. The microgrid system includes a local controller and a cloud server, energy storage units, inverter modules, and circuit breakers connected to the local controller. The method is executed by the local controller and includes:

[0052] The system receives and stores the droop coefficient matrix sent by the cloud server. The droop coefficient matrix is ​​generated by the cloud server based on the historical load power records of the microgrid. The droop coefficient matrix contains a three-dimensional mapping relationship between the droop coefficient, the cell surface temperature and the frequency deviation. The system also obtains the preset rated frequency of the microgrid, the real-time grid frequency and the real-time cell surface temperature of the energy storage unit.

[0053] Based on the real-time power grid frequency, determine whether to trigger a local interpolation scheduling operation; the local interpolation scheduling operation includes: calculating the frequency deviation between the real-time power grid frequency and the preset rated frequency;

[0054] Based on the frequency deviation and real-time cell surface temperature, interpolation calculation is performed in the droop coefficient matrix to generate transient droop coefficient; based on the transient droop coefficient, a pulse width modulation signal is generated and the duty cycle of the pulse width modulation signal is adjusted and sent to the inverter module to change the active current amplitude injected into the microgrid by the inverter module.

[0055] After performing a single local interpolation scheduling operation, the frequency recovery time from the triggering of the local interpolation scheduling operation to the real-time grid frequency recovering to the preset rated frequency is obtained, and the average temperature rise rate of the battery cells during the scheduling period corresponding to the frequency recovery time is calculated to determine whether a thermal penalty flag needs to be generated and sent to the cloud server to instruct the cloud server to correct the droop coefficient matrix generated next time.

[0056] This embodiment provides a real-time scheduling mechanism for microgrids based on big data analysis. Specifically, a typical application scenario is set as an industrial park-type microgrid, which includes a photovoltaic array, a 1MWh lithium iron phosphate energy storage system, several injection molding machines and refrigeration units. During normal grid-connected operation, the cloud server performs slow-cycle optimization, and the local controller is responsible for millisecond-level disturbance response.

[0057] The cloud server does not directly send a fixed power value at a certain moment to the field equipment, but instead sends a set of droop coefficient matrices that can be used locally;

[0058] This matrix can be understood as a control map bound to the physical state. One dimension corresponds to the frequency deviation range, and the other dimension corresponds to the cell surface temperature range. The appropriate droop coefficient is given by the corresponding position.

[0059] To illustrate with a simplified application example, if a temperature range is set in the matrix... , , and frequency deviation range , , Then the matrix nodes correspond and What is represented is not abstract data, but the allowable support strength when the battery is at a moderate temperature and with a large frequency deviation;

[0060] The reason for this design is that frequency deviation reflects the real-time power imbalance of the microgrid, while cell surface temperature reflects the current thermal load bearing capacity of the energy storage unit. Both together determine whether energy storage is suitable for high-intensity active power output.

[0061] After receiving the matrix, the local controller stores it in local memory, such as SRAM or the cache of the industrial controller. During operation, the local controller continuously obtains the real-time grid frequency through the phasor measurement unit or phase-locked loop and obtains the real-time cell surface temperature through the battery management system.

[0062] If there is a sudden drop in photovoltaic output or a concentrated start-up of injection molding machines on site, the grid frequency will deviate from the rated value, and the local controller will determine whether to trigger local interpolation scheduling operation accordingly.

[0063] After being triggered, instead of re-initiating a calculation request to the cloud, it directly finds adjacent nodes in the local droop coefficient matrix based on the current frequency deviation and cell surface temperature, and generates a transient droop coefficient that adapts to the current state.

[0064] After the transient droop coefficient is written into the inverter control link, the digital control chip of the inverter module reconstructs the sinusoidal modulation waveform according to the updated transient droop coefficient, generates a pulse width modulation signal and adjusts the duty cycle of the pulse width modulation signal, so that the inverter module injects active current with the corresponding duty cycle into the AC bus to suppress frequency offset.

[0065] Furthermore, after a single local scheduling action is completed, the local controller counts the recovery time from the triggering of scheduling to the frequency returning to near the rated value, and calculates the average temperature rise rate by combining the battery temperature change during this period.

[0066] The specific calculation method for the average temperature rise rate is as follows: subtract the real-time cell surface temperature at the moment the frequency recovery time ends from the real-time cell surface temperature at the moment the local interpolation scheduling operation is triggered to obtain the temperature rise difference, and then divide the temperature rise difference by the frequency recovery time.

[0067] The physical significance of this step is that if the frequency recovers quickly but is accompanied by a significant temperature rise, it indicates that although the energy storage unit is effectively supported, the thermal burden is too large.

[0068] If the frequency recovery time is greater than the preset time threshold and the temperature rise rate is lower than the preset safety threshold, it indicates that the current support strength has not reached the rated output power limit; based on this, it can be decided whether to generate a thermal penalty flag and send it back to the cloud to constrain the subsequent matrix generation process.

[0069] Regarding the anomaly handling mechanism, if the cloud matrix has not been updated successfully, the local cache is empty, the communication link is interrupted for a short time, or the temperature sensor shows an abnormal value, the local controller can call the preset backup control parameter group and continue to maintain operation by using a fixed droop coefficient or the most recent valid matrix to avoid the inverter module from going out of control due to data loss.

[0070] If the frequency sensor data jitter is within the measurement noise range, it will not enter a drastic adjustment, but will only maintain normal steady-state control; during the afternoon production peak in the industrial park, multiple injection molding machines entered the heating and mold-locking stages one after another, and the bus load suddenly increased, while at the same time, cloud cover caused the photovoltaic output to decrease.

[0071] The local controller detected a drop in the grid connection frequency and read from the BMS that the real-time cell surface temperature of the energy storage unit was in the medium temperature range. Therefore, it directly called the transient droop coefficient suitable for the cell surface temperature and frequency deviation from the locally stored matrix and immediately adjusted the duty cycle of the pulse width modulation signal to improve the active power support of the inverter module.

[0072] After the frequency is restored, the local controller finds that although the support was fast, the battery temperature rises quickly. Therefore, it uploads the thermal penalty flag to the cloud in the periodic status monitoring message.

[0073] The purpose of this step is to transform the slow big data analysis results in the cloud into control rule parameters that can be executed locally in real time, so that the microgrid can balance frequency stability and energy storage thermal safety without relying on high-frequency communication.

[0074] In a preferred embodiment of the present invention, the droop coefficient matrix issued by the cloud server is generated in the following manner: applying a sliding window with a preset time width to the historical load power records of the microgrid, and calculating the variance of active power within the sliding window; matching and comparing the variance with each preset variance threshold interval in the preset variance threshold set, determining the preset variance threshold interval in which the variance is located, and thus determining the fluctuation level corresponding to the sliding window; using the preset basic droop coefficient corresponding to the fluctuation level as a benchmark, and combining it with a preset battery internal resistance thermodynamic model that maps active power to predicted temperature rise increment, generating the corresponding droop coefficient matrix; wherein, the input parameters of the preset battery internal resistance thermodynamic model include the Joule heat power generated by the equivalent charge and discharge current, the time span of the sliding window, the equivalent thermal constant of the cell, the current cell surface temperature, the ambient temperature, and the equivalent thermal resistance of the system.

[0075] This embodiment provides a cloud-based generation mechanism for the droop coefficient matrix; specifically, in the aforementioned industrial park microgrid scenario, relying solely on manual experience to generate a fixed matrix easily overlooks the time-varying characteristics of the park's load.

[0076] For example, the load of the refrigeration unit is relatively stable at night, while the injection molding machine and air compressor start and stop frequently during the day. The two have significantly different requirements for energy storage support strategies. Therefore, this embodiment introduces a matrix generation process based on historical load fluctuation levels and a preset battery internal resistance thermodynamic model.

[0077] The cloud server reads the active power records of the park over a period of time in the past, at the second or minute level, from the historical database and observes them in segments using a sliding window of preset width; here, variance is not a simple mathematical quantity, but an engineering indicator that reflects whether the load fluctuation is stable within a certain period of time.

[0078] If the variance of the load curve within a certain window is less than the preset lower limit threshold for fluctuation, it indicates that the load composition is stable during that period and energy storage does not need to perform high-gain regulation; if the load frequently increases and decreases sharply within a certain window, it indicates that there is a significant impact load on site, and the subsequent local droop control should be prepared with higher disturbance support.

[0079] To facilitate understanding, a simplified application example can be used for illustration: Suppose the cloud divides the historical load window into... , , Three segments, among which Corresponding to the constant operation of the cold storage at night, To accommodate the continuous start-up of injection molding machines during the early morning shift, This corresponds to the afternoon fluctuations in photovoltaic power generation combined with production disruptions;

[0080] After evaluating the load dispersion of these three windows respectively, they can be mapped to low fluctuation, medium fluctuation and high fluctuation levels; the cloud takes the preset basic droop coefficient corresponding to each level as the starting point, and then superimposes the constraints of the preset battery internal resistance thermodynamic model; this thermal model is used to reflect the temperature rise trend of the battery under a certain output intensity in the matrix generation.

[0081] The specific calculation and transfer rules are as follows: divide the historical average active power corresponding to each fluctuation level by the rated bus voltage of the energy storage system, and convert it into the equivalent charging and discharging current;

[0082] The equivalent charge and discharge current is input into the preset battery internal resistance thermodynamic model. Based on the current temperature conditions, the Joule heat power generated by the equivalent charge and discharge current is calculated by calling the dynamic internal resistance value of the cell.

[0083] To avoid closed calculations in the flow of model data, the mapping process of the preset battery internal resistance thermodynamic model is realized through rigorous structured calculation: multiply the Joule heat power by the time span of the sliding window, divide it by the equivalent thermal capacity constant of the cell, and subtract the heat dissipation loss value calculated by dividing the difference between the current cell surface temperature and the ambient temperature by the equivalent thermal resistance of the system. Finally, the predicted temperature rise increment on the corresponding sliding window time scale is calculated.

[0084] The specific calculation formula is as follows:

[0085]

[0086] in, To predict the increase in temperature, Joule heat power, The time span of the sliding window. Let be the equivalent heat capacity constant of the battery cell. This is the current surface temperature of the battery cell. For ambient temperature, The system's equivalent thermal resistance; equivalent heat capacity constant. Equivalent thermal resistance of the system These are known parameters pre-set based on the factory calibration data or historical operation test data of the energy storage unit;

[0087] Through the data flow logic of the above non-complex formula, the expected temperature rise trend under a specific fluctuation level can be clearly obtained; since the battery internal resistance changes with temperature and operating state, the greater the current, the more obvious the internal resistance heats up. Therefore, the same droop coefficient should not be used for the same frequency support strength under low temperature and high temperature conditions.

[0088] The resulting matrix is ​​not a single fixed set of values, but is organized according to fluctuation level, temperature range, and frequency deviation range; for example, for high fluctuation levels, a more aggressive support slope can be allowed in the low temperature range.

[0089] In the high-temperature range, even when facing the same frequency deviation, if the predicted temperature rise increment is superimposed on the current cell surface temperature approaching the safety limit, the droop coefficient mapping value should be appropriately reduced to avoid the energy storage from continuing to bear the drastic adjustment when the thermal state is too high.

[0090] Regarding the anomaly handling mechanism, if historical data is missing, sampling timestamps are misaligned, or the amount of data within a certain window is insufficient to reflect the actual load fluctuations, the cloud server can skip the abnormal window and use the statistical results of adjacent valid windows to make up the difference.

[0091] If the battery internal resistance parameters required by the thermal model have not been updated, the factory calibration parameters or the most recent calibration parameters corresponding to the battery model can be called to avoid complete matrix mismatch.

[0092] Based on a month of continuous operation data from the park, the cloud platform found that the load fluctuation of the injection molding machines during the 8:00 to 10:00 AM start-up period was significantly higher than that during the nighttime freezing and preservation period. Therefore, when generating the droop coefficient matrix for the next day, the morning shift was marked as a higher fluctuation level, and a coefficient mapping more suitable for rapid support was set in the lower cell surface temperature range. For the afternoon high-temperature period, the coefficient mapping value in the high-temperature range was appropriately reduced based on the battery temperature rise prediction results.

[0093] The purpose of this step is to ensure that the control matrix issued by the cloud accurately reflects historical operating patterns and energy storage thermal constraints, thereby improving the adaptability of local scheduling parameters from the source.

[0094] In a preferred embodiment of the present invention, the step of determining whether to trigger a local interpolation scheduling operation based on the real-time power grid frequency includes: calculating the absolute value of the frequency deviation between the real-time power grid frequency and the preset rated frequency; determining whether the absolute value of the frequency deviation is greater than a preset frequency fluctuation threshold; if the absolute value of the frequency deviation is greater than or equal to the frequency fluctuation threshold, then triggering a local interpolation scheduling operation; if the absolute value of the frequency deviation is less than the frequency fluctuation threshold, then not triggering a local interpolation scheduling operation.

[0095] This embodiment provides a triggering determination mechanism for local interpolation scheduling. Specifically, in the aforementioned scheme, if the matrix is ​​immediately invoked and the inverter control parameters are rewritten as soon as any slight change in frequency occurs, although the response seems positive, it will cause the inverter to adjust frequently, and the energy storage unit will also continuously respond to measurement noise and slight natural fluctuations, which will increase ineffective microcirculation and thermal stress. Therefore, this embodiment introduces a frequency fluctuation threshold as a triggering threshold.

[0096] The local controller first calculates the absolute value of the deviation of the real-time grid frequency from the rated frequency, and then compares it with the preset frequency fluctuation threshold; in engineering terms, the threshold represents the minimum disturbance intensity that requires energy storage intervention.

[0097] Deviations below this threshold are usually caused by normal load disturbances, measurement noise, or short-term phase-locked loop jitter, and are insufficient to indicate that there is a substantial power imbalance in the microgrid.

[0098] The specific method for setting the preset frequency fluctuation threshold is as follows: obtain the allowable steady-state frequency deviation limit in the power grid standard specifications of the microgrid area, and combine it with the maximum measured white noise amplitude in the steady-state operation history data of the system phase-locked loop for weighted summation, and use the summation result as the preset frequency fluctuation threshold;

[0099] Only when the deviation continuously or momentarily exceeds the threshold does it indicate that there is a considerable disturbance in the active power balance on the AC bus, and only then is it necessary for the local controller to initiate interpolation scheduling.

[0100] A simplified application example can be used to illustrate this: If the local controller continuously acquires the frequency status of three sampling periods as close to the rated value, slightly low, and significantly low, then the first two types of states are only monitored, while the third type of state is entered into the matrix calling process; this design can avoid mixing the slight drift that should be monitored with the actual disturbance that should be intervened in.

[0101] In the fault-tolerant control mechanism, if the frequency deviation is in the threshold boundary critical region and high-frequency over-limit oscillation occurs, the local controller can set a short-term hold strategy or a minimum trigger duration to prevent repeated entry and exit from the scheduling state.

[0102] If the frequency sampling link is abnormal, such as PMU communication packet loss or PLL lockout, the trigger judgment based on the sampled value will be suspended, and the energy storage output will be maintained by the backup control mode until the frequency measurement is restored to reliability.

[0103] During the afternoon production phase in the industrial park, the refrigeration unit continued to operate, causing the load to change slowly. However, this slow change did not exceed the preset threshold, so the local controller did not start matrix interpolation and only maintained the current droop parameters. A high-power injection molding machine suddenly switched on, causing the frequency deviation to quickly exceed the threshold. Based on this, the local controller determined it to be a valid disturbance and immediately triggered local interpolation scheduling.

[0104] The purpose of this step is to distinguish between grid disturbances that truly require energy storage and normal minor fluctuations, thereby achieving a balance between rapid response and suppressing ineffective actions.

[0105] In a preferred embodiment of the present invention, the step of performing interpolation calculation in the droop coefficient matrix based on the frequency deviation and the real-time cell surface temperature to generate the transient droop coefficient includes: extracting at least 6 matrix node data adjacent to the current frequency deviation and the real-time cell surface temperature in the droop coefficient matrix.

[0106] The matrix node data is used to execute a surface fitting interpolation algorithm based on a quadratic polynomial to calculate the corresponding transient droop coefficient; the transient droop coefficient is then sent to the inverter module to reconstruct the sinusoidal modulation waveform.

[0107] This embodiment provides a local interpolation generation mechanism for transient droop coefficients; specifically, in the previous embodiment, although it was possible to determine when scheduling was triggered, if the value of the nearest single node in the matrix was directly used, control abrupt changes could easily occur when the state crosses intervals.

[0108] For example, when the cell temperature moves from the edge of one temperature zone to an adjacent temperature zone, if the coefficients switch abruptly, the inverter output will experience discontinuous adjustments, affecting power quality. Therefore, this embodiment introduces a surface fitting interpolation algorithm based on adjacent matrix node data and quadratic polynomials.

[0109] The local controller extracts data from at least six matrix nodes adjacent to the current frequency deviation and the current real-time cell surface temperature from the matrix.

[0110] To construct a locally continuous control surface, the local controller expands the search range; to illustrate with a simplified application example, if the current real-time cell surface temperature is located at... and Between, the frequency deviation is located and In the meantime, the local controller prioritizes extracting broader neighborhood node data covering the current state from the droop coefficient matrix, for example, selecting data including temperature. , , Frequency deviation , , Composition A total of 9 matrix nodes;

[0111] Using these 9 known matrix node data as sample inputs, a surface fitting interpolation algorithm based on quadratic polynomials is executed through methods such as least squares to calculate the corresponding transient droop coefficients;

[0112] To clearly demonstrate the computational flow of the surface fitting interpolation algorithm without relying on closed logic descriptions, this embodiment decomposes its data flow as follows: frequency deviation and real-time cell surface temperature are taken as the first and second independent feature variables, respectively, and a quadratic polynomial equation with a total of 6 fitting basis functions is constructed, including a constant term, a linear term and a quadratic term of the first feature variable, a linear term and a quadratic term of the second feature variable, and a cross-product term of the two variables.

[0113] The specific expression for the quadratic polynomial equation is:

[0114]

[0115] in, To fit the target value, i.e., the transient droop coefficient, The first characteristic variable is frequency deviation. The second characteristic variable is the real-time cell surface temperature. to The coefficients of the polynomial to be solved;

[0116] After substituting the coordinates and corresponding coefficients of the nine matrix node data extracted above into the equation, the least squares method is used to solve for the six polynomial coefficients that minimize the sum of squares of the local surface residuals.

[0117] By substituting the exact coordinates of the current real-time frequency deviation and the real-time cell surface temperature into a quadratic polynomial with definite coefficients, the transient sag coefficient of smooth transition can be accurately calculated; its essence is to make the energy storage support strength change continuously with the cell surface temperature and frequency deviation, rather than change in a step.

[0118] The engineering significance of using a surface fitting interpolation algorithm based on quadratic polynomials is that the relationship between battery thermal state and frequency support requirements is usually not a simple linear one; as the temperature rises, the changes in battery internal resistance, allowable current, and heat accumulation trend may all exhibit bending characteristics; when the frequency deviation increases, the support strength may not increase proportionally; through local fitting, the actual stress and heat generation patterns of the equipment can be more closely matched.

[0119] After generating the transient droop coefficient, the local controller sends it to the inverter module to reconstruct the sinusoidal modulation waveform, thereby adjusting the duty cycle of the pulse width modulation signal accordingly, which in turn changes the amplitude of the active current injected into the microgrid by the inverter module.

[0120] Regarding the anomaly handling mechanism, if the current state is near the matrix boundary and it is impossible to obtain enough complete neighborhood nodes to form an overdetermined or well-determined equation, for example, if there are fewer than 6 effective nodes, then boundary interpolation, reduced-order bilinear interpolation, or extrapolation of the nearest effective interval can be used.

[0121] If some node data is corrupted or verification fails, the abnormal nodes are removed and the same reduced-order interpolation is used; if the current cell surface temperature or frequency deviation exceeds the matrix's predefined range, it can be clamped to the safety control area corresponding to the boundary value to prevent abnormal output caused by out-of-range calls.

[0122] During a sudden load surge in the industrial park, the local controller detected that the real-time cell surface temperature was approaching the medium-high temperature boundary, while the frequency deviation was in the moderate downward range. At this time, if the fixed coefficient in the high temperature zone is switched directly, the energy storage support may suddenly weaken; if the coefficient in the medium temperature zone is still used, the thermal burden on the battery may increase.

[0123] Therefore, the local controller calls the data of the nine adjacent matrix nodes to perform local surface fitting and outputs a smooth transient droop coefficient between the two temperature zones. The inverter module then reconstructs the sinusoidal modulation waveform according to this coefficient to achieve continuous and abrupt active power support.

[0124] The purpose of this step is to ensure that the local scheduling parameters change smoothly with the physical conditions on site, thereby achieving inverter control continuity, reducing current surges, and improving power quality during frequency recovery.

[0125] In a preferred embodiment of the present invention, the step of determining whether a thermal penalty flag needs to be generated and sent to the cloud server includes: determining whether the frequency recovery time is less than a preset recovery time threshold and whether the average battery temperature rise rate is greater than a preset temperature rise safety threshold; if the frequency recovery time is less than or equal to the preset recovery time threshold and the average battery temperature rise rate is greater than or equal to the preset temperature rise safety threshold, then it is determined that a thermal penalty flag needs to be generated and sent to the cloud server; otherwise, it is determined that a thermal penalty flag does not need to be generated.

[0126] This embodiment provides a mechanism for generating a thermal penalty flag. Specifically, in the aforementioned scheme, the local controller is already able to quickly recover the frequency, but if the success of frequency recovery is used as the evaluation criterion, it may obscure the thermal costs of the energy storage unit.

[0127] In other words, although a certain scheduling action achieves the goal of grid stability, it may cause the battery to suffer excessive thermal shock in a short period of time; therefore, this embodiment introduces a joint determination of frequency recovery time and average temperature rise rate.

[0128] After each local interpolation scheduling, the local controller calculates the time taken from the disturbance trigger to the frequency recovering to near the rated value, and evaluates the average temperature rise rate in conjunction with the temperature changes during this period. The joint determination here emphasizes an engineering balance: if the recovery time meets the requirements, it means that the support is effective on the grid side.

[0129] If the temperature rise rate exceeds the safety threshold at the same time, it means that this effectiveness is achieved at a high thermal cost and should not be repeated with the same aggressive degree in the future.

[0130] A thermal penalty flag is generated only when both conditions of rapid recovery and rapid temperature rise are met. Conversely, if recovery is slow but temperature rise is high, it usually indicates that the equipment is in an abnormally inefficient state, and the battery health or power device status should be checked first, rather than simply applying a penalty control strategy. If recovery is fast but temperature rise is not high, it indicates that the current matrix is ​​appropriate and no penalty is needed.

[0131] A simplified application example can be described as: three consecutive disturbance events , , Post-event evaluations will be conducted separately, if If the recovery is timely and the temperature rise is stable, no marker will be generated; if... If the recovery is timely and the temperature rise is significant, a marker is generated; if If the symptoms include slow recovery and excessively high temperature rise, then the system will be placed under abnormal monitoring rather than direct punishment.

[0132] Therefore, this flag does not simply indicate that the temperature is high, but rather that the current scheduling method is not thermally safe but is still frequently used on the frequency side;

[0133] In the fault-tolerant control mechanism, if the temperature sensor experiences a communication interruption during a single scheduling period, the recovery time cannot be reliably calculated, or the disturbance has not completely ended, the current evaluation can be marked as an invalid sample and no flag bit will be generated. If the frequency recovery process is affected by external faults, such as switching of the upstream power grid or circuit breaker reconfiguration, the event can be removed from the thermal penalty judgment to prevent topology anomalies from being mistaken for control problems.

[0134] When multiple injection molding machine start-up and shutdown disturbances occurred in the industrial park in the afternoon, the local controller found that in one of the events, the energy storage inverter quickly pulled the frequency back to the stable range, but the surface temperature of the battery cells rose significantly faster than usual within one minute, indicating that this rapid support was accompanied by a high thermal load.

[0135] Therefore, the local controller marks the event as requiring hot punishment and sends the corresponding flag to the cloud server in subsequent communication cycles;

[0136] The purpose of this step is to incorporate grid stability and battery thermal safety into the same evaluation closed loop, thereby providing a basis for adaptive correction of the subsequent control matrix.

[0137] In a preferred embodiment of the present invention, instructing the cloud server to correct the droop coefficient matrix generated next time is to instruct the cloud server to perform the following correction operation: when generating the droop coefficient matrix next time, extract the target temperature range in the droop coefficient matrix where the temperature dimension is greater than or equal to the real-time cell surface temperature when the thermal penalty flag is triggered.

[0138] Reduce the droop coefficient mapping value corresponding to the target temperature range in the droop coefficient matrix to limit the subsequent active power output amplitude of the energy storage unit.

[0139] This embodiment provides a matrix correction mechanism based on a hot penalty flag. Specifically, in the previous embodiment, although the local system can identify events that recover quickly but have excessive hot load, if the flag is only recorded and not applied to the next round of matrix generation, the closed loop cannot be truly formed. Therefore, this embodiment further specifies the correction method after the cloud receives the flag.

[0140] When the cloud server generates the droop coefficient matrix for the next time, it identifies the real-time cell surface temperature corresponding to the trigger flag and determines the target temperature range accordingly. The correction here does not uniformly reduce the entire matrix, but only applies targeted constraints to the temperature dimension at and above that temperature.

[0141] The physical basis is that thermal risks have obvious state dependence. The problem is not necessarily in all operating conditions, but is more likely to be concentrated in the local area where more aggressive support is still performed under high temperature conditions. Therefore, the cloud only reduces the droop coefficient mapping value within the target temperature range, so that the same frequency disturbance corresponds to a more conservative active support strength in subsequent high temperature scenarios.

[0142] A simplified application example can be used to illustrate this: assuming the matrix temperature dimension is divided into... , , , One of the heat punishment incidents occurred If the interval is specified, the cloud will retain it during the next round of matrix correction. and The interval remains unchanged, only for and The relevant droop coefficients in the range are adjusted downwards; in this way, strong support can still be maintained during low-temperature periods, while automatically switching to power reduction and limited output mode during high-temperature periods.

[0143] The reason why this interval-based correction is better than the global reduction is that the global reduction will impair the normal support capability of energy storage in a low-temperature healthy state; while the local temperature zone correction takes into account both frequency stability and thermal safety; in addition, the cloud can apply different magnitudes of correction to different temperature zones according to the distribution of multiple consecutive flag bits to achieve targeted thermal constraints.

[0144] In the fault-tolerant control mechanism, if a reported flag lacks a corresponding temperature context, or the temperature value is not within the range defined by the matrix, the cloud can merge it into the nearest high-temperature safe zone for processing.

[0145] If a large number of flag bits are received in a short period of time, it indicates that there may be abnormal heat dissipation, battery aging, or cooling system failure on site. In this case, in addition to correcting the matrix, an operation and maintenance alarm can also be triggered, but it will not affect the matrix downsizing process of this solution.

[0146] During the continuous high temperatures in the industrial park, the energy storage system experienced multiple events in the afternoon where the frequency recovered quickly but the temperature rose too fast, and the cell temperature was in a relatively high range when the events were triggered.

[0147] After receiving these flag bits, the cloud maintains the support slope of the low-temperature range in the morning basically unchanged when generating the matrix next time, while reducing the droop coefficient mapping value corresponding to the high-temperature range in the afternoon, so that the energy storage will no longer output overly aggressive active power support at high temperatures.

[0148] The purpose of this step is to feed back the thermal risks identified locally as cloud-based parameter correction actions, thereby achieving power output convergence and battery life protection under high-temperature operating conditions.

[0149] In a preferred embodiment of the present invention, the method further includes: monitoring the switching state of circuit breakers installed in the microgrid to obtain the topology state; determining whether the topology state has undergone physical reconstruction based on the switching state transition; if the topology state has undergone physical reconstruction and is within a preset transient transition period, then forcibly switching to constant voltage and constant frequency islanding control mode; otherwise, maintaining local interpolation scheduling operation.

[0150] This embodiment provides an anomaly fallback mechanism under topology reconfiguration; specifically, in the aforementioned scheme, the local droop coefficient matrix is ​​generated based on historical load and existing network structure; when the microgrid topology is basically stable, this matrix can meet the preset error control requirements;

[0151] However, if the main feeder inside the park is disconnected, the tie switch is switched, or a branch becomes an isolated island, the impedance distribution and power flow path corresponding to the original matrix may no longer hold true. If interpolation scheduling is still performed as usual, it may cause the control target to become disconnected from the actual network state. Therefore, this embodiment introduces a topology awareness and forced handover mechanism.

[0152] The local controller monitors the status of key circuit breakers and tie switches in real time to obtain the current microgrid topology status. In engineering, a change in switch status usually means a change in the physical connection relationship of the network, such as a change from a ring network to a radial power supply, or a switch of some load branches from grid-connected power supply to local island power supply.

[0153] Once the local controller detects such a transition, it determines that the topology may be undergoing physical reconfiguration. Since the voltage support point, equivalent impedance, and power distribution path are all in the process of being re-established in the short period after the topology reconfiguration, it is more important to ensure the stability of the bus voltage and frequency reference than to perform fine interpolation scheduling during this stage. Therefore, the system is forced to switch to constant voltage and constant frequency islanding control mode. After the preset transient transition period ends, if the network stabilizes again, it will resume local interpolation scheduling operation.

[0154] A simplified application example can be used to illustrate this: If the switch state sequence is... Switch to ,and When the main feeder is operated in segments, the local controller identifies it as a topology reconfiguration event. During the transition period, all dynamic droop actions based on the historical matrix are suspended, and the inverter switches to constant voltage and constant frequency mode to establish the bus reference, ensuring that the local load is not powered and avoiding simultaneous frequency and voltage drift.

[0155] In the fault-tolerant control mechanism, if a change in switch state is detected but the duration is extremely short, it may be due to contact bounce or transient communication sampling. In this case, the confirmation time window can be increased to avoid misjudgment.

[0156] If switch status data is missing, cross-judgment can be made by combining auxiliary means such as bus voltage change and power flow direction change; if frequency and voltage fluctuations are still found after the transition period ends, the constant voltage and constant frequency mode will continue to be maintained and the updated matrix will be sent from the cloud based on the new topology.

[0157] During a fault repair in the industrial park, maintenance personnel disconnected the interconnecting circuit breaker, temporarily separating the two busbars that were originally connected to the grid.

[0158] At this point, the local controller detects a change in the circuit breaker status and identifies it as a physical reconfiguration event. It immediately suspends the interpolation scheduling executed according to the old matrix and instead puts the energy storage inverter into constant voltage and constant frequency islanding control, prioritizing support for the critical injection molding production line and refrigeration load that remain on this bus section. After the topology stabilizes, the matrix-driven dynamic scheduling is restored.

[0159] The purpose of this step is to provide a verifiable and safe degradation path for the system in the event of sudden changes in network structure, thereby achieving risk isolation of the mismatch matrix and ensuring the stability of islanded operation.

[0160] Please see Figure 2 The microgrid real-time dispatch system based on big data analytics includes a cloud server, a local controller, and energy storage units and inverter modules that communicate with the local controller. The cloud server is used to generate a droop coefficient matrix based on the historical load power records of the microgrid and send it to the local controller; and to receive a heat penalty flag sent by the local controller to correct the droop coefficient matrix generated next time. The local controller includes a processor and a memory. The memory is used to store the droop coefficient matrix and a computer program. When the processor executes the computer program, it implements the steps of the microgrid real-time dispatch method based on big data analytics.

[0161] This embodiment provides a real-time microgrid scheduling system based on big data analysis; specifically, the system is deployed in the aforementioned industrial park-type microgrid and includes a cloud server, a local controller, an energy storage unit, an inverter module, a phasor measurement unit, a circuit breaker status acquisition unit, and a battery management system interface.

[0162] The cloud server can be deployed in the park's energy management center or remote operation and maintenance center. Its processor is used to execute historical load analysis, fluctuation level identification, thermal model calling, and droop coefficient matrix generation programs.

[0163] Its memory is used to store historical load databases, matrix templates, heat penalty records, and operating parameters under different topology conditions; the cloud server sends the droop coefficient matrix to the local controller through industrial Ethernet, private network, or other reliable communication links, and receives the heat penalty flag, frequency recovery record, and necessary operating summary information returned locally;

[0164] The local controller can be an industrial PLC, edge controller, or embedded industrial computer. Its processor performs real-time frequency acquisition, trigger judgment, matrix interpolation, PWM parameter update, hot penalty reporting, and topology switching logic. Its memory is used to store the current valid matrix, the most recent valid matrix, the backup control parameters, and event logs.

[0165] The local controller receives the cell surface temperature reported by the BMS via the CAN bus or other fieldbus, obtains the real-time grid frequency via the PMU or PLL interface, and sends the updated droop parameters or PWM control quantities to the inverter control module via the digital interface or analog interface.

[0166] In terms of hardware coordination, the inverter module includes a power conversion unit and a digital control unit. The digital control unit reconstructs the modulation waveform based on the transient droop coefficient provided by the local controller, while the power conversion unit performs the corresponding current injection action through IGBTs or other power devices.

[0167] The circuit breaker status acquisition unit is responsible for uploading the status of the main feeder, tie switch and key branch circuit breakers in the park to the local controller to support topology reconfiguration judgment;

[0168] In the fault-tolerant control mechanism, if the cloud server is unavailable, the local controller can still call the cached matrix and backup parameters to run independently; if the local controller fails, the inverter can revert to the preset basic droop mode or constant voltage and constant frequency mode.

[0169] If communication link latency increases, the system will still maintain the structure of slow updates in the cloud and fast response locally, and will not interrupt on-site adjustments due to remote delays.

[0170] In the complete operation chain of the industrial park, the cloud server generates a new matrix every day based on the historical load of the most recent 30 days and the heat penalty record of the previous day, and distributes it to the local controller via a private network;

[0171] During peak daytime production periods, the local controller receives real-time data from the PMU and BMS, and immediately performs interpolation scheduling and controls the inverter output when load changes suddenly.

[0172] If a support event shows that the thermal load in the high-temperature zone is too high, the local controller sends back a flag bit so that the cloud can correct the high-temperature zone control mapping in the next round of matrix generation; if the switching of the park's segmented switches causes topology reconfiguration, the local controller immediately switches the inverter to constant voltage and constant frequency islanded control.

[0173] The purpose of this step is to support the aforementioned implementation methods with a systematic hardware and software collaborative structure, so that cloud-based strategy generation, local real-time scheduling, thermal feedback correction, and abnormal topology fallback can operate in a closed loop within the same microgrid architecture.

[0174] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A real-time dispatching method for microgrids based on big data analysis, applied to a microgrid system, wherein the microgrid system includes a local controller and a cloud server, energy storage unit, inverter module, and circuit breaker communicatively connected to the local controller, characterized in that, The method is executed by the local controller and includes: The cloud server receives and stores the droop coefficient matrix, which is generated by the cloud server based on the historical load power records of the microgrid. The droop coefficient matrix contains a three-dimensional mapping relationship between the droop coefficient, the cell surface temperature and the frequency deviation. The system acquires the preset rated frequency of the microgrid, the real-time grid frequency, and the real-time cell surface temperature of the energy storage unit; based on the real-time grid frequency, it determines whether to trigger a local interpolation scheduling operation. The local interpolation scheduling operation includes: calculating the frequency deviation between the real-time power grid frequency and the preset rated frequency; and performing interpolation calculation in the droop coefficient matrix based on the frequency deviation and the real-time cell surface temperature to generate a transient droop coefficient. Based on the transient droop coefficient, a pulse width modulation signal is generated and the duty cycle of the pulse width modulation signal is adjusted and sent to the inverter module to change the active current amplitude injected into the microgrid by the inverter module. After performing a single local interpolation scheduling operation, the frequency recovery time from the triggering of the local interpolation scheduling operation to the recovery of the real-time power grid frequency to the preset rated frequency is obtained, and the average temperature rise rate of the battery cells during the scheduling period corresponding to the frequency recovery time is calculated to determine whether a thermal penalty flag needs to be generated and sent to the cloud server to instruct the cloud server to correct the droop coefficient matrix generated next time.

2. The microgrid real-time scheduling method based on big data analysis according to claim 1, characterized in that, The droop coefficient matrix issued by the cloud server is generated in the following way: Apply a sliding window with a preset time width to the historical load power records of the microgrid, and calculate the variance of active power within the sliding window; The variance is matched and compared with each preset variance threshold interval in the preset variance threshold set to determine the preset variance threshold interval in which the variance is located, so as to determine the fluctuation level corresponding to the sliding window. Based on the preset droop coefficient corresponding to the fluctuation level, and combined with the preset battery internal resistance thermodynamic model that maps active power to predicted temperature rise increment, the corresponding droop coefficient matrix is ​​generated; wherein, the input parameters of the preset battery internal resistance thermodynamic model include the Joule heat power generated by the equivalent charge and discharge current, the time span of the sliding window, the equivalent thermal constant of the cell, the current cell surface temperature, the ambient temperature, and the equivalent thermal resistance of the system.

3. The microgrid real-time scheduling method based on big data analysis according to claim 1, characterized in that, The steps for determining whether to trigger a local interpolation scheduling operation based on the real-time power grid frequency include: Calculate the absolute value of the frequency deviation between the real-time power grid frequency and the preset rated frequency; Determine whether the absolute value of the frequency deviation is greater than a preset frequency fluctuation threshold; If the absolute value of the frequency deviation is greater than or equal to the frequency fluctuation threshold, the local interpolation scheduling operation is triggered; if the absolute value of the frequency deviation is less than the frequency fluctuation threshold, the local interpolation scheduling operation is not triggered.

4. The microgrid real-time scheduling method based on big data analysis according to claim 1, characterized in that, Based on the frequency deviation and the real-time cell surface temperature, the step of performing interpolation calculations in the droop coefficient matrix to generate the transient droop coefficient includes: Extract at least 6 matrix node data adjacent to the current frequency deviation and the real-time cell surface temperature from the droop coefficient matrix; The matrix node data is used to execute a surface fitting interpolation algorithm based on a quadratic polynomial to calculate the corresponding transient droop coefficient. The transient droop coefficient is sent to the inverter module to reconstruct the sinusoidal modulation waveform.

5. The microgrid real-time scheduling method based on big data analysis according to claim 1, characterized in that, The steps for determining whether a hot penalty flag needs to be generated and sent to the cloud server include: Determine whether the frequency recovery time is less than a preset recovery time threshold, and determine whether the average temperature rise rate of the battery is greater than a preset temperature rise safety threshold. If the frequency recovery time is less than or equal to the preset recovery time threshold and the average battery temperature rise rate is greater than or equal to the preset temperature rise safety threshold, then it is determined that the heat penalty flag needs to be generated and sent to the cloud server; otherwise, it is determined that the heat penalty flag does not need to be generated.

6. The microgrid real-time scheduling method based on big data analysis according to claim 5, characterized in that, Instructing the cloud server to correct the droop coefficient matrix generated next time means instructing the cloud server to perform the following correction operation: When generating the droop coefficient matrix again, extract the target temperature range from the droop coefficient matrix where the temperature dimension is greater than or equal to the real-time cell surface temperature when the heat penalty flag is triggered. The droop coefficient mapping value corresponding to the target temperature range in the droop coefficient matrix is ​​reduced to limit the subsequent active power output amplitude of the energy storage unit.

7. The microgrid real-time scheduling method based on big data analysis according to claim 1, characterized in that, Also includes: Monitor the switching status of circuit breakers installed in the microgrid to obtain the topology status; Based on the transition of the switch state, determine whether the topology state has undergone physical reconstruction; If the topology state undergoes physical reconstruction and is within a preset transient transition period, it will be forcibly switched to constant voltage and constant frequency islanding control mode; otherwise, the local interpolation scheduling operation will be maintained.

8. A real-time dispatching system for microgrids based on big data analytics, characterized in that: It includes a cloud server, a local controller, and an energy storage unit and an inverter module that are communicatively connected to the local controller; The cloud server is used to generate a droop coefficient matrix based on the historical load power records of the microgrid and send it to the local controller. And receive the hot penalty flag sent by the local controller to correct the droop coefficient matrix generated next time; The local controller includes a processor and a memory, the memory being used to store the droop coefficient matrix and a computer program, and the processor executing the computer program to implement the steps of the microgrid real-time scheduling method based on big data analysis as described in any one of claims 1 to 7.

Images