Intelligent monitoring and power scheduling system of new energy storage power station

Abstract

This invention relates to the field of energy storage monitoring and scheduling technology, and discloses an intelligent monitoring and power scheduling system for a new energy storage power station. The system includes a topology mapping module for acquiring location data and establishing matrix parameters containing spatial heat dissipation weights; a state update module for updating the effective DC internal resistance based on the fluctuation variance of the charging and discharging current; an instruction extraction module for analyzing and judging total power instructions and distributing tasks; a thermoelectric assessment module for determining the estimated Joule heat by combining the trial current and the effective DC internal resistance, obtaining a thermoelectric penalty factor by combining the spatial heat dissipation weights, and generating a safety verification state; and a scheduling execution module for allocating comprehensive scheduling weights based on the safety verification state, the estimated Joule heat, and the thermoelectric penalty factor to generate the target active power. This invention uses heat dissipation efficiency as a benchmark to limit power allocation, and utilizes transient high current to drive the battery clusters across the peak internal resistance, achieving temperature equalization and reducing DC losses.

Description

Technical Field

[0001] This invention relates to the field of energy storage monitoring and dispatching technology, specifically to an intelligent monitoring and power dispatching system for a new energy storage power station. Background Technology

[0002] As the penetration rate of new energy sources in the power grid increases, the large-scale application and dispatch control of energy storage power stations are becoming increasingly important. Currently, conventional energy dispatch for energy storage power stations mainly relies on capacity balancing and power distribution based on the state of charge of battery clusters.

[0003] However, due to the inherent non-uniformity of airflow and resistance distribution within energy storage containers, battery clusters located at the air inlet and the end of the air duct exhibit differences in heat dissipation capacity. Traditional purely electrical dispatching methods do not take into account the heat dissipation characteristics of the physical space, leading to heat accumulation in battery clusters in heat-limited areas under long-term, high-load operation. This localized hotspot phenomenon frequently triggers the system's thermal limit protection, resulting in passive derating or shutdown, making it difficult to achieve temperature balance within the energy storage system.

[0004] Meanwhile, existing control systems typically employ a centralized processing strategy to handle dispatch commands issued by the power grid. This fails to analyze the time-frequency domain characteristics of the total power command to distinguish between steady-state long-cycle tasks and transient fast-response tasks, resulting in communication and execution delays in the control of transient power components. When responding to transient commands, if the battery cluster is in a high-resistance state at low ambient temperatures, traditional strategies usually adopt conservative current-limiting methods. These strategies fail to proactively leverage the nonlinear physical characteristics of the electrolyte, such as its negative temperature coefficient, to overcome the peak internal resistance region, increasing DC-side losses in the energy storage system's power conversion process.

[0005] Furthermore, existing technologies often employ online identification algorithms such as recursive least squares to update the dynamic DC internal resistance of battery clusters. In actual operation of energy storage power stations, the power demand of the power grid exhibits periods of stability or low-amplitude fluctuations. Under conditions of insufficient electrical excitation, continuously running online identification algorithms are prone to covariance matrix divergence, leading to distortion of the internal resistance identification results. The lack of a parameter update mechanism that isolates operating conditions reduces the stability of the algorithm's operation and affects the data reliability of input parameters in subsequent heat generation prediction and feedback scheduling stages. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an intelligent monitoring and power dispatching system for new energy storage power stations. This system solves the problems of existing dispatching technologies that mainly rely on the state of charge for centralized allocation, without comprehensively considering the differences in heat dissipation in physical space, the time-frequency characteristics of commands, and the divergence of parameter identification under weak excitation. As a result, the system suffers from frequent local hot spots, high DC losses, and poor reliability of online parameter identification.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] This invention provides an intelligent monitoring and power dispatching system for new energy storage power stations, comprising:

[0009] The topology mapping module is used to acquire the location data of the battery clusters and establish matrix parameters that include spatial heat dissipation weights.

[0010] The status update module is used to obtain the charging and discharging current of the battery cluster and calculate the fluctuation variance, and update the effective DC internal resistance based on the fluctuation variance;

[0011] The instruction extraction module is used to analyze and judge the acquired total power instruction to generate a scheduling instruction identifier, and to distribute tasks according to the scheduling instruction identifier in order to determine the corresponding scheduling level and scheduling instruction attributes.

[0012] The thermoelectric assessment module is used to set a trial current for the battery cluster, combine the trial current with the effective DC internal resistance to call the heat generation prediction formula to calculate the estimated Joule heat, combine the space heat dissipation weight to call the thermoelectric penalty formula to obtain the thermoelectric penalty factor, and combine the preset safety threshold to generate a safety verification state.

[0013] The scheduling execution module is used to allocate a comprehensive scheduling weight based on the security verification status, the scheduling instruction attribute, the estimated Joule heat, the thermoelectric penalty factor and the effective DC internal resistance, generate a target active power based on the comprehensive scheduling weight, and complete power scheduling in combination with the scheduling level.

[0014] Furthermore, the topology mapping module includes a parameter calibration unit and a static storage unit;

[0015] The parameter calibration unit is used to calculate the location data through thermodynamic analysis, calibrate the spatial heat dissipation weight of each battery cluster, and generate the matrix parameters containing the spatial heat dissipation weight.

[0016] The static storage unit is used to extract the matrix parameters and store the matrix parameters as static constants in the non-volatile memory of the energy management system and edge controller in the new energy storage power station, thereby quantifying the heat dissipation efficiency of the physical space and solidifying it, providing a physical boundary benchmark for system scheduling.

[0017] Furthermore, the state update module includes a feature acquisition unit;

[0018] The feature acquisition unit is used to synchronously acquire the real-time temperature, terminal voltage and charging / discharging current of each battery cluster through the underlying sensing device at a set sampling frequency; the feature acquisition unit continuously integrates the charging / discharging current in the time dimension and calculates the state of charge of the corresponding battery cluster in combination with the initial capacity of the corresponding battery cluster.

[0019] The sampling frequency is preset based on the hardware sampling capability of the underlying sensing device and the response requirements of the system control cycle.

[0020] Furthermore, the state update module also includes an adaptive update unit;

[0021] The adaptive update unit is used to calculate the fluctuation variance of the charging and discharging current within a set sliding time window;

[0022] When the fluctuation variance is greater than or equal to the system's preset effective excitation threshold, the adaptive update unit determines that the current state is a fully excited state of the battery cluster, calls the recursive least squares method to calculate the dynamic DC internal resistance of the corresponding battery cluster in real time by combining the state of charge and the real-time temperature, and uses it as the effective DC internal resistance.

[0023] When the fluctuation variance is less than the system's preset effective excitation threshold, the adaptive update unit determines that the current state is a constant current static state of the battery cluster, freezes the covariance matrix update of the recursive least squares method, extracts the static DC internal resistance under the current operating condition, and uses it as the effective DC internal resistance.

[0024] The effective excitation threshold is preset based on the boundary condition of avoiding the divergence of the covariance matrix due to insufficient external excitation. Then, the internal resistance parameter update logic is adaptively switched according to the electrical excitation intensity to obtain a high-confidence internal resistance parameter benchmark under different operating conditions.

[0025] Furthermore, the instruction extraction module includes a feature recognition unit;

[0026] The feature identification unit is used to perform frequency domain analysis and attribute discrimination on the time domain waveform of the total power command, and generate the scheduling command identifier;

[0027] When the command change rate of the total power command is less than or equal to the preset power change rate threshold, or when the command change rate is greater than the power change rate threshold and the time-domain waveform is not accompanied by high-frequency zero-crossing features, the feature identification unit will use the steady-state command identifier as the scheduling command identifier.

[0028] When the command change rate is greater than the power change rate threshold and the time-domain waveform is accompanied by the high-frequency zero-crossing feature, the feature identification unit will use the transient command identifier as the scheduling command identifier.

[0029] The power change rate threshold is preset based on the rated active power output slope of the energy storage power station and the typical inertial response requirements of grid frequency adjustment.

[0030] Furthermore, the instruction extraction module also includes a task distribution unit;

[0031] The task distribution unit is used to distribute the total power command according to the scheduling instruction identifier;

[0032] If the scheduling instruction identifier is a steady-state instruction identifier, the task distribution unit retains the total power instruction in the centralized layer for global optimization, generates a steady-state instruction attribute as an attribute of the scheduling instruction, and determines the corresponding scheduling level;

[0033] If the scheduling instruction identifier is a transient instruction identifier, the task distribution unit extracts the power envelope parameters of the total power instruction and sends them to the edge controller to execute the waveform decomposition task, generates a transient instruction attribute as an attribute of the scheduling instruction, and determines the corresponding scheduling level. Thus, the control task is divided into the global layer and the edge layer by judging the instruction attribute, and the underlying processing path is optimized.

[0034] Furthermore, the thermoelectric evaluation module includes a feedforward prediction unit and a state quantization unit;

[0035] The feedforward estimation unit is used to set the trial current for each battery cluster based on the scheduling instruction attribute, and to calculate the estimated Joule heat of the corresponding battery cluster by multiplying the square of the trial current by the effective DC internal resistance using the heat generation estimation formula.

[0036] The state quantization unit is used to obtain the real-time temperature of the battery cluster and calculate the corresponding real-time temperature rise rate. Using the thermoelectric penalty formula, the two preset conversion constant coefficients of the system are multiplied by the real-time temperature and the real-time temperature rise rate respectively, and then summed and divided by the space heat dissipation weight to construct the thermoelectric penalty factor.

[0037] The conversion constant coefficient is predetermined based on the temperature rise experimental data of the battery cluster under different ambient temperatures through multiple linear regression analysis. This establishes a mapping relationship between electrical response and heat accumulation, and quantifies the deviation between physical heat dissipation capacity and real-time thermal state.

[0038] Furthermore, the thermoelectric evaluation module also includes a boundary constraint unit;

[0039] The boundary constraint unit is used to compare the real-time temperature with the safety threshold.

[0040] If the real-time temperature is greater than or equal to the product of the safety threshold and the set safety margin coefficient, the boundary constraint unit generates a safety verification state indicating failure, forcibly resets the comprehensive scheduling weight of the corresponding battery cluster to zero, and removes the corresponding battery cluster from the scheduling allocation stage of the current control cycle and triggers a re-loop when generating the safety verification state indicating failure.

[0041] If the real-time temperature is less than the product of the safety threshold and the safety margin coefficient, the boundary constraint unit generates the safety verification state indicating that the test is passed, retains the corresponding thermoelectric penalty factor, and releases the corresponding battery cluster.

[0042] The safety threshold is preset based on the upper limit of the long-term operating temperature allowed by the battery cell, and the safety margin coefficient is preset based on the sensor measurement error and the thermal gradient non-uniformity margin inside the container.

[0043] Furthermore, the scheduling execution module includes a branch optimization unit;

[0044] After the branch optimization unit extracts the safety verification state indicating that the test has passed, if the scheduling instruction attribute is a steady-state instruction attribute, it executes a steady-state optimization branch, uses the estimated Joule heat as the analysis input to construct the optimization function, and the optimization function approximates the actual heat generation distribution of each battery cluster with the corresponding spatial heat dissipation weight. Based on the positive correlation between the estimated Joule heat and the effective DC internal resistance, the spatial heat dissipation weight is used as a gain term, and the thermoelectric penalty factor and the effective DC internal resistance are used as attenuation terms for weighted calculation. The thermoelectric penalty factor is used as a negative feedback correction term to generate the comprehensive scheduling weight.

[0045] If the scheduling instruction attribute is a transient instruction attribute, the branch optimization unit executes a restricted pulse optimization branch and determines whether the real-time temperature is within the set negative temperature coefficient range.

[0046] If the battery cluster is in the negative temperature coefficient range and has a low spatial heat dissipation weight, the branch optimization unit ignores the current limiting constraint of the thermoelectric penalty factor, generates a negative temperature coefficient effect compensation term based on the deviation between the effective DC internal resistance and the preset nominal impedance value, and uses the negative temperature coefficient effect compensation term to generate the comprehensive scheduling weight, converting the estimated Joule heat as a reference benchmark into actual high current heat generation.

[0047] The negative temperature coefficient range is predetermined based on the physical properties of electrolyte ionic conductivity changing with temperature, so as to guide the battery cluster across the inefficient operating zone through differentiated instructions while ensuring safety.

[0048] Furthermore, the scheduling execution module also includes an instruction normalization unit and a closed-loop drive unit;

[0049] The instruction normalization unit is used to divide the comprehensive scheduling weight of a single battery cluster by the sum of the comprehensive scheduling weights of all battery clusters in the station that are in the release state, in order to calculate the allocation ratio coefficient, and multiply the absolute value of the obtained total power instruction by the allocation ratio coefficient to generate the target active power of each energy storage converter.

[0050] The closed-loop drive unit is used to perform power control in conjunction with the scheduling level. When the scheduling level corresponds to a steady-state scheduling command, it drives the corresponding energy storage converter by uniformly issuing control messages through the energy management system.

[0051] When the scheduling level corresponds to a transient scheduling command, the closed-loop drive unit directly drives the corresponding energy storage converter in a local closed loop through the edge controller to complete the power scheduling.

[0052] This invention provides an intelligent monitoring and power dispatching system for new energy storage power stations. It has the following beneficial effects:

[0053] 1. This invention establishes spatial heat dissipation weights by acquiring battery cluster location data, constructs thermoelectric penalty factors by combining real-time temperature and temperature rise rate, and uses them as negative feedback correction terms in the steady-state optimization branch to limit the allocation of comprehensive scheduling weights. This makes the heat generation distribution inside the power station match the heat dissipation capacity boundary of the physical space, slows down the local temperature rise trend, and, together with the safety verification state reset weight mechanism, effectively avoids the problem of shutdown due to thermal limit reduction caused by local hot spots, thus achieving temperature balance inside the energy storage system.

[0054] 2. This invention generates an identifier by extracting the frequency domain characteristics of the total power command and dividing the scheduling hierarchy. The transient command is then sent to the edge controller for execution. At the same time, the negative temperature coefficient range is identified in the constrained pulse optimization branch, and a negative temperature coefficient effect compensation term is actively allocated. The transient high current heat generation promotes the electrolyte temperature rise, drives the internal resistance parameter to decrease, and enables the specified battery cluster to quickly cross the internal resistance peak region, reducing DC loss at the power conversion level. Furthermore, the tracking delay for transient power components is shortened through edge-side control.

[0055] 3. This invention determines the excitation state of the battery cluster by calculating the fluctuation variance of the charging and discharging current. When the excitation is sufficient, the recursive least squares method is called to update the dynamic DC internal resistance in real time. When the constant current is at rest, the static DC internal resistance is extracted. This solves the problem of covariance matrix divergence caused by stable or low-amplitude fluctuations of the power grid signal, maintains the stability of the online identification process, and ensures the data reliability of the input parameters in the subsequent heat generation prediction stage. Attached Figure Description

[0056] Figure 1This is a diagram illustrating the architecture of an intelligent monitoring and power dispatching system for a new energy storage power station according to an embodiment of the present invention.

[0057] Figure 2 This is a flowchart of an intelligent monitoring and power dispatching method for a new energy storage power station according to an embodiment of the present invention;

[0058] Figure 3 This is an architecture diagram of the topology mapping module in an embodiment of the present invention;

[0059] Figure 4 This is an architecture diagram of the state update module in an embodiment of the present invention;

[0060] Figure 5 This is an architecture diagram of the instruction extraction module in an embodiment of the present invention;

[0061] Figure 6 This is an architectural diagram of the thermoelectric evaluation module in an embodiment of the present invention;

[0062] Figure 7 This is an architecture diagram of the scheduling and execution module in an embodiment of the present invention;

[0063] Figure 8 This is a diagram showing the target active power allocation ratio of battery clusters in three typical states in an application embodiment of the present invention.

[0064] Figure 9 This is a comparison chart of the highest real-time temperature evolution of the control group and the experimental group under continuous operating conditions in an application embodiment of the present invention. Detailed Implementation

[0065] The technical solutions in 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.

[0066] Please see the appendix Figure 1 This invention provides an intelligent monitoring and power dispatching system for a new energy storage power station, including a topology mapping module, a status update module, an instruction extraction module, a thermoelectric assessment module, and a dispatching execution module.

[0067] The system typically operates within the control architecture of a new energy storage power station, which integrates an energy management system, an edge controller, and sensing devices that connect to the battery clusters.

[0068] Before a new energy storage power station is put into operation, the topology mapping module uses the acquired location data of each battery cluster in the new energy storage power station to process the location data through thermodynamic analysis, calculate and calibrate matrix parameters that include spatial heat dissipation weights. These matrix parameters are stored as static constants in the non-volatile memory of the energy management system and edge controller, thereby pre-constructing a spatial physical boundary benchmark.

[0069] During system operation, the status update module synchronously collects data using sensing devices. The real-time temperature, terminal voltage, and charge / discharge current of each battery cluster are used to estimate the state of charge (SOC) in conjunction with the initial capacity of the battery cluster.

[0070] The state update module calculates the variance of the charging and discharging current fluctuation and compares it with a preset effective excitation threshold. When the variance is greater than or equal to the effective excitation threshold, the state update module determines that the battery cluster is in a fully electrically excited state. It then uses a recursive least squares algorithm combined with the state of charge and real-time temperature to calculate and update the dynamic DC internal resistance of the battery cluster under the current state of charge and real-time temperature, and uses this as the effective DC internal resistance. When the variance is less than the effective excitation threshold, the state update module determines that the battery cluster is in a constant current static state. It updates the covariance matrix using the recursive least squares algorithm and extracts the static DC internal resistance under the current operating condition, using this as the effective DC internal resistance.

[0071] The instruction extraction module is responsible for receiving the total power instruction from the external power grid dispatch instructions, extracting the instruction change rate by analyzing the time-domain waveform of the total power instruction, and performing attribute discrimination.

[0072] When the command change rate of the total power command is less than or equal to the preset power change rate threshold, or when the command change rate of the total power command is greater than the power change rate threshold and the time-domain waveform of the total power command is not accompanied by high-frequency zero-crossing characteristics, the command extraction module generates a steady-state command identifier as a scheduling command identifier.

[0073] When the rate of change of the total power command exceeds the power change rate threshold and the time-domain waveform of the total power command exhibits high-frequency zero-crossing characteristics, the command extraction module generates a transient command identifier as the scheduling command identifier. Based on the scheduling command identifier, the command extraction module generates corresponding scheduling command attributes (steady-state command attributes or transient command attributes), issues specific waveform decomposition tasks, and determines the corresponding scheduling level.

[0074] After obtaining the steady-state or transient command attributes, the thermoelectric evaluation module performs the following steps: Each battery cluster is given a trial current, and feedforward calculations are performed using the heat generation prediction formula in conjunction with the effective DC internal resistance to determine the predicted Joule heat.

[0075] Simultaneously, the thermoelectric assessment module calculates the temperature rise rate based on real-time temperature and, in conjunction with the space heat dissipation weight, calls the thermoelectric penalty formula to obtain the thermoelectric penalty factor. Subsequently, the thermoelectric assessment module compares the real-time temperature with internally preset safety thresholds and safety margin coefficients. If the real-time temperature is greater than or equal to the product of the safety threshold and the safety margin coefficient, the thermoelectric assessment module generates a safety verification status indicating failure, forcibly resets the comprehensive scheduling weight of the battery cluster to zero to block its output permissions, and triggers power redistribution logic to reallocate the power vacancy corresponding to the restricted battery cluster to other battery clusters in a permitted state, awaiting re-verification in the next control cycle. If the real-time temperature is less than the product of the safety threshold and the safety margin coefficient, the thermoelectric assessment module generates a safety verification status indicating success, allowing the battery cluster to proceed.

[0076] After extracting the safety verification status indicating that the signal has passed, the scheduling execution module assigns a comprehensive scheduling weight to each battery cluster based on the scheduling instruction attributes, thermoelectric penalty factor, estimated Joule heat, spatial heat dissipation weight, and effective DC internal resistance.

[0077] The scheduling execution module normalizes the comprehensive scheduling weights and calculates the allocation ratio coefficient. Finally, the scheduling execution module multiplies the absolute value of the total power command by the allocation ratio coefficient to determine the target active power, and according to the scheduling level, sends control messages through the energy management system or uses the edge controller to drive the energy storage converter in a local closed loop, ultimately completing the power dispatch.

[0078] See appendix Figure 2 This invention also provides an intelligent monitoring and power dispatching method for new energy storage power stations, comprising the following steps:

[0079] S1, the topology mapping module acquires the location data of the battery clusters before the new energy storage power station is put into operation. It processes the location data through thermodynamic analysis, calculates and calibrates the matrix parameters containing spatial heat dissipation weights, and stores the matrix parameters as static constants in the non-volatile memory of the energy management system and edge controller, and pre-constructs the spatial physical boundary benchmark.

[0080] S2, the status update module uses sensing devices to synchronously collect the data of the first... The system calculates the real-time temperature, terminal voltage, and charging / discharging current of each battery cluster, and estimates the state of charge by combining the initial capacity of the battery cluster. It then calculates the fluctuation variance of the charging / discharging current and compares it with the preset effective excitation threshold to determine whether the battery cluster is in a state of sufficient electrical excitation or a state of constant current quiescence. Finally, it calls the recursive least squares algorithm to update the dynamic DC internal resistance or extract the static DC internal resistance to determine the effective DC internal resistance.

[0081] S3, the instruction extraction module receives the total power instruction from the external power grid dispatch instruction, analyzes the time-domain waveform of the total power instruction to extract the instruction change rate and perform attribute discrimination, generates a steady-state instruction identifier or a transient instruction identifier as the dispatch instruction identifier, determines the dispatch instruction attribute (steady-state instruction attribute or transient instruction attribute) accordingly, issues waveform decomposition tasks and divides the dispatch levels.

[0082] S4, the thermoelectric assessment module is based on the scheduling instruction attribute as the first Each battery cluster is given a trial current. The estimated Joule heat is calculated using the heat generation prediction formula in combination with the effective DC internal resistance. At the same time, the temperature rise rate is calculated using the real-time temperature and the thermoelectric penalty factor is generated by the thermoelectric penalty formula in combination with the space heat dissipation weight. Finally, the real-time temperature is compared with the safety threshold and safety margin coefficient to generate a safety verification status.

[0083] S5. After extracting the safety verification status indicating that the signal has passed, the scheduling execution module allocates a comprehensive scheduling weight to the battery cluster based on the comprehensive scheduling instruction attributes, thermoelectric penalty factor, estimated Joule heat, spatial heat dissipation weight, and effective DC internal resistance. After normalization processing, the allocation ratio coefficient is calculated, and the target active power is generated by combining the absolute value of the total power instruction. Based on the scheduling level, the energy management system or edge controller drives the energy storage converter to complete the power scheduling.

[0084] See appendix Figure 3 In this embodiment, the heat dissipation characteristics of the physical space are quantified and solidified to provide a static benchmark for subsequent thermoelectric coupling scheduling. The specific implementation details are as follows:

[0085] S11, the parameter calibration unit acquires the position data of each battery cluster in the new energy storage power station. This position data comes from the three-dimensional spatial layout coordinate system determined during the prefabrication and installation of the energy storage container. The parameter calibration unit processes the position data using thermodynamic analysis when using fluid dynamics simulation.

[0086] When using fluid dynamics simulation and thermodynamic analysis with the finite volume method, the parameter calibration unit performs mesh discretization of the airflow domain inside the energy storage container, and sets the flow boundary conditions of the air inlet according to the geometric gap of the heat dissipation channel between each battery cluster, the distribution of resistance elements, and the preset fan speed.

[0087] The preset fan speed is determined in advance based on the rated heating power inside the container and the target heat exchange temperature difference, with a range of 2000rpm to 4500rpm.

[0088] The parameter calibration unit solves the momentum, energy, and turbulence closure equations numerically to obtain the local velocity distribution vectors at the locations of each battery cluster within the container. Based on mean normalization logic, it maps the ratio of the local velocity at a specific location to the average velocity across the entire station as a preliminary weight to a preset weight range, thereby calibrating the spatial heat dissipation weight for each battery cluster. .

[0089] The space's heat dissipation weight The dimensionless scalar value used to characterize the heat dissipation efficiency of a physical location is set between 0.6 and 1.4. Battery clusters located in areas directly impacted by air inlets or in areas with strong convection are assigned a higher space heat dissipation weight due to their lower thermal resistance. Battery clusters located downstream of the flow field or with dead zones are assigned a lower space heat dissipation weight due to the reduced effective heat exchange temperature difference. .

[0090] For the selection of the computational domain, mesh quality verification, and specific selection of the turbulence model involved in the process of fluid dynamics simulation modeling, those skilled in the art can use existing numerical simulation analysis methods or commercial simulation software to achieve these. The modeling and setting logic is a well-known technology in this field and will not be elaborated here.

[0091] When performing thermodynamic analysis using heat distribution testing, temperature acquisition nodes are arranged on the surface of each battery cluster to monitor the temperature rise process under preset charge / discharge rate conditions ranging from 0.2C to 1.0C. The time constant required for the battery cluster temperature to rise to 63.2% of the steady-state temperature rise is analyzed to reflect the relative heat dissipation rate at each location. The faster the heat dissipation rate, the smaller the time constant corresponding to the physical location. Based on this, a spatial heat dissipation weight is generated by mapping using an inverse proportional function. The analysis process must ensure the overall heat dissipation weight of the entire station space. The arithmetic mean is within the preset deviation range of 1.0±0.05, thereby avoiding the deviation of the total power command in subsequent scheduling. After calibration, a matrix parameter containing the heat dissipation weight of all battery clusters is formed.

[0092] S12, the static storage unit extracts the matrix parameters generated during calibration, which include the heat dissipation weights of the entire battery cluster space, and writes them as static constants into the non-volatile memory of the energy management system and edge controller in the new energy storage power station; the non-volatile memory adopts EEPROM or Flash Memory, and stores the matrix parameters into the preset address space (such as the storage sector starting at 0x0800C000) defined according to the controller memory mapping table through the SPI protocol.

[0093] The stored static constant serves as the physical boundary benchmark in the subsequent thermoelectric coupling scheduling process, enabling the system to directly retrieve the spatial heat dissipation characteristics in real-time power scheduling, thus avoiding the real-time performance degradation caused by repeatedly executing high-load thermodynamic simulation calculations within the control cycle.

[0094] See appendix Figure 4 In this embodiment, high-frequency dynamic capture of multi-dimensional electrical signals during the charging and discharging process of the battery cluster is performed, and the identification logic is adaptively switched based on the electrical excitation intensity to ensure that high-confidence internal resistance parameters can be obtained under different operating conditions. The specific implementation details are as follows:

[0095] S21, the feature acquisition unit synchronously acquires the first characteristic through sensing devices distributed at the bottom layer of each branch of the new energy storage power station at a preset sampling frequency of 1Hz to 20Hz. The real-time temperature, terminal voltage, and charging / discharging current of each battery cluster are collected. The feature acquisition unit uses the ampere-hour integration method to continuously accumulate and calculate the charging / discharging current. That is, the charge increment is determined by multiplying the charging / discharging current collected at the current moment with the sampling period. Combined with the pre-stored initial capacity of the battery cluster (the value ranges from 50Ah to 300Ah), the current state of charge of the battery cluster is calculated by accumulating the ratio of the initial state of charge to the charge change.

[0096] For the initial SOC calibration and current sensor bias error compensation during the ampere-hour integration process, those skilled in the art can use open-circuit voltage curves or Kalman filtering algorithms to achieve this. The implementation logic is well-known in the field and will not be elaborated here.

[0097] S22, the adaptive update unit extracts the acquired charging and discharging current, real-time temperature, and state of charge, and calculates the fluctuation variance of the charging and discharging current within a sliding time window of 10s to 120s, and then compares this fluctuation variance with 0.01A. 2 up to 0.1A 2 The preset effective threshold for excitation is compared;

[0098] If the fluctuation variance is greater than or equal to the effective excitation threshold, it is determined that the battery cluster is in a state of sufficient electrical excitation. The recursive least squares algorithm with a forgetting factor (range from 0.95 to 0.999) is called to perform matrix iteration, so as to separate the dynamic response after the coupling of the ohmic internal resistance and polarization internal resistance inside the battery cluster. The dynamic DC internal resistance of the battery cluster under the current state of charge and real-time temperature is updated in real time and used as the effective DC internal resistance.

[0099] If the fluctuation variance is less than the effective excitation threshold, it is determined that the battery cluster is currently in a constant current static state. To avoid the covariance matrix diverging due to insufficient external excitation, the system forcibly freezes the recursive least squares update process and stops online parameter identification. Instead, it queries a pre-stored two-dimensional lookup table with real-time temperature as the row index and state of charge as the column index to extract the static DC internal resistance of the battery cluster under the current operating conditions as the effective DC internal resistance.

[0100] The static DC internal resistance data is established based on the prior Hybrid Pulse Power Characterization (HPPC) test, which ensures that a continuous and accurate internal resistance reference can still be output in the control cycle where dynamic excitation is lacking, thereby supporting the subsequent calculation of estimated Joule heat.

[0101] See appendix Figure 5 In this embodiment, by decoupling time-frequency domain features, mutually exclusive power components in the grid demand are separated, thereby achieving parallel processing of long-cycle economic dispatch and millisecond-level power support. Specific implementation details are as follows:

[0102] S31, the feature identification unit obtains the total power instruction in the power grid dispatching instruction. In actual application, the feature identification unit establishes a time series model by collecting the digital instruction sequence issued by the communication gateway.

[0103] The feature identification unit obtains the frequency domain energy distribution characteristics by performing a fast Fourier transform on the time domain waveform of the total power command, and calculates the command change rate of the total power command over time. The specific calculation method of the command change rate is to obtain the absolute value of the first derivative of the total power command with respect to time, which is used to quantify the severity of the command waveform.

[0104] The feature identification unit compares the calculated command change rate with a preset power change rate threshold. This power change rate threshold is set based on the rated active power output slope of the energy storage power station and the typical inertial response requirements of grid frequency adjustment, and its value is usually set between 3% / s and 15% / s of the rated power of the energy storage power station.

[0105] When the rate of change of the total power command is less than or equal to the power change rate threshold, the feature identification unit generates a steady-state command identifier as the scheduling command identifier. When the rate of change of the command is greater than the power change rate threshold and the time-domain waveform of the total power command is not accompanied by high-frequency zero-crossing characteristics, it indicates that although the total power command is in a rapid ramp-up or descent phase, there is no frequent reversal of energy flow. At this time, the feature identification unit still generates a steady-state command identifier as the scheduling command identifier.

[0106] The identification logic for high-frequency zero-crossing features is as follows: within a preset zero-power neighborhood (e.g., the range of rated power ±1%), count the number of total power command symbol switching times and calculate the switching frequency per unit time.

[0107] The preset frequency limit is determined based on the hardware switching frequency response range of the energy storage converter, for example, 5Hz can be selected.

[0108] If the command change rate is greater than the power change rate threshold and is accompanied by high-frequency zero-crossing characteristics, the feature identification unit determines that the total power command contains a high-dynamic transient component and generates a transient command identifier as a scheduling command identifier.

[0109] S32, the task distribution unit extracts the scheduling instruction identifier and the corresponding total power instruction generated by the feature identification unit.

[0110] When the scheduling instruction identifier is a steady-state instruction identifier, the task distribution unit determines that the total power instruction is a steady-state scheduling instruction. The steady-state scheduling instruction corresponds to long-cycle tasks such as peak shaving or planned output, which have high requirements for overall economic efficiency.

[0111] The task distribution unit retains the steady-state scheduling instructions for execution in the energy management system, generates steady-state instruction attributes, and determines the corresponding scheduling level as the system global level, so as to maximize the power plant's energy utilization efficiency through global resource optimization.

[0112] When the scheduling instruction is identified as a transient instruction, the task distribution unit determines that the total power instruction is a transient scheduling instruction. Transient scheduling instructions usually correspond to fast-response tasks such as frequency modulation or voltage support. To avoid long-distance communication delays, they need to be executed at the edge.

[0113] The task distribution unit generates transient instruction attributes and extracts the power envelope parameters from the transient scheduling instructions before issuing them.

[0114] The power envelope parameter specifically includes the amplitude envelope of the transient waveform, the response time constant, and the phase compensation coefficient.

[0115] The specific waveform decomposition task is devolved to the edge controller for local execution. The corresponding scheduling level is the edge control level. Through the devolution mechanism, the edge controller can directly run algorithms such as discrete wavelet transform locally to map transient components to the current control loop of the converter in real time.

[0116] The specific algorithm implementation of the Fast Fourier Transform and the discretization difference process for power envelope extraction can be achieved by those skilled in the art using conventional mathematical tools in the field of digital signal processing. The details of the algorithm are well-known in the field and will not be elaborated here.

[0117] See appendix Figure 6In this embodiment, by constructing an electrothermal coupling feedforward model, future electrical excitation is transformed into a perceptible heat accumulation trend, thereby pre-adjusting the allocation weights of the power branches before the temperature rise occurs. The specific implementation details are as follows:

[0118] S41, the feedforward prediction unit extracts the scheduling instruction attributes (steady-state instruction attributes or transient instruction attributes) generated by the task distribution unit, and sets a trial current for the i-th battery cluster based on these scheduling instruction attributes. .

[0119] During the implementation process, this tentative current The method for determining it is as follows:

[0120] The total power command is initially allocated according to the rated capacity ratio of each battery cluster in the entire station to obtain the step-by-step power demand. Then, this step-by-step power demand is divided by the first power demand. The real-time terminal voltage of each battery cluster is obtained. This trial current... The value ranges from 0.1C to 2C.

[0121] If the scheduling command attribute is a steady-state command attribute, then the trial current... Take the arithmetic mean of the initial current allocations; if the dispatch command attribute is transient, then the trial current... Take the peak current from the power envelope parameter.

[0122] The feedforward prediction unit extracts the effective DC internal resistance generated by the adaptive update unit. and combined with trial current The estimated Joule heat was obtained by using the heat production prediction formula. The formula for predicting heat production is:

[0123] ;

[0124] In the formula: For the first Estimated Joule heat of each battery cluster; For the first The trial current of each battery cluster; For the first The effective DC internal resistance of each battery cluster is determined. By mapping the electrical response in advance to the thermal accumulation increment, data support is provided for subsequent preventative scheduling.

[0125] S42, the state quantization unit extracts the real-time temperature acquired by the feature acquisition unit. And calculate its real-time temperature rise rate. Real-time temperature rise rate The temperature is obtained by performing a first-order difference calculation on the real-time temperature within a sliding time window.

[0126] The state quantization unit extracts the corresponding space heat dissipation weights from the matrix parameters stored in the static storage unit. Based on real-time temperature Real-time temperature rise rate and space heat dissipation weight Using the thermoelectric penalty formula, a thermoelectric penalty factor is constructed to characterize the deviation between the real-time thermal state of the battery cluster and its heat dissipation capacity at its physical location. The thermoelectric penalty formula is:

[0127] ;

[0128] In the formula: For the first Thermoelectric penalty factor for each battery cluster; and All are system-preset conversion constants; For the first Real-time temperature of each battery cluster; For the first Real-time temperature rise rate of each battery cluster; For the first Spatial heat dissipation weight of each battery cluster.

[0129] Conversion constant coefficients and It was determined through multiple linear regression analysis based on experimental data of battery cluster temperature rise under different ambient temperatures, and is used to balance the contribution of absolute temperature and temperature rise tendency to scheduling constraints. and The values ​​range from 0.1 to 1.5.

[0130] Through this thermoelectric penalty factor The system can quantify the degree of derating in scheduling priority for battery clusters that are in a position of poor heat dissipation or whose temperature rises too quickly.

[0131] S43, the boundary constraint element will display the real-time temperature. Compare with the internally preset safety threshold.

[0132] If real-time temperature If the product of the safety threshold and the internally preset safety margin coefficient is greater than or equal to the value of the battery cluster, the battery cluster is determined to have entered the thermal limit protection state. At this time, the boundary constraint unit generates a safety verification state indicating that the test has failed, forcibly resets the comprehensive scheduling weight of the battery cluster to zero, and blocks its power output permission in the current control cycle. Subsequently, the system triggers the power redistribution logic to reallocate the power vacancy corresponding to the restricted battery cluster to other battery clusters in the released state.

[0133] If real-time temperature If the value is less than the product of the safety threshold and the safety margin coefficient, the boundary constraint element generates a safety verification state indicating that the test is passed, and retains the corresponding thermoelectric penalty factor. .

[0134] The safety threshold is determined based on the upper limit of the long-term operating temperature allowed for a single battery cell, and its value ranges from 45°C to 60°C; the safety margin coefficient is determined based on the sensor measurement error and the margin for thermal gradient non-uniformity within the container, and its value ranges from 0.85 to 0.95.

[0135] For the discretization and differential processing of real-time temperature rise rate, those skilled in the art can use digital filtering algorithms and basic differential operations to achieve it. The processing logic is a well-known technology in this field and will not be described in detail here.

[0136] See appendix Figure 7 In this embodiment, by introducing electrothermal coupling feedback into the weighting model and utilizing the nonlinear characteristics of battery electrochemical properties (such as the negative temperature coefficient effect), the battery cluster is guided across the inefficient operating region through differentiated power commands while ensuring safety, thereby improving the overall system response quality. Specific implementation details are as follows:

[0137] In step S51, after extracting the safety verification status indicating that the signal has passed, the branch optimization unit assigns a comprehensive scheduling weight to each battery cluster based on the comprehensive scheduling instruction attributes, thermoelectric penalty factor, estimated Joule heat, spatial heat dissipation weight, and effective DC internal resistance. This allocation process analyzes the heat accumulation risk based on the estimated Joule heat, establishes the optimization target, derives the actual execution weight calculation formula, and finally generates the comprehensive scheduling weight by combining the weights of each operating condition branch.

[0138] Under steady-state conditions, the distribution logic aims to achieve a balanced distribution of heat load across the spatial dimension. The branch optimization unit uses the feedforward calculation of the estimated Joule heat... As a baseline input for assessing the risk of heat accumulation, an optimization function is constructed. The control objective of this optimization function is to suppress abnormal branches that exceed the estimated Joule heat limit, so that the actual heat generation distribution of each battery cluster approximates its corresponding spatial heat dissipation weight. .

[0139] Based on the positive correlation between predicted Joule heat and effective DC internal resistance, the space heat dissipation weight is adjusted. As a gain term placed in the molecule, the thermoelectric penalty factor is also included. With effective DC internal resistance As a decay term, it is included in the denominator for weighted calculation, using the thermoelectric penalty factor. As a negative feedback correction term, the comprehensive scheduling weight is generated, along with the thermoelectric penalty factor. The higher the battery cluster, the lower the overall scheduling weight, which in turn limits the target active power it can acquire, causing the actual heat to adaptively converge towards the physical heat dissipation boundary.

[0140] Under transient conditions, this invention utilizes the negative temperature coefficient (NTC) effect to achieve active scheduling of peak internal resistance.

[0141] In practice, the branch optimization unit determines whether the current real-time temperature of the battery cluster is within the negative temperature coefficient range; if the current real-time temperature of the battery cluster is within the negative temperature coefficient range (excluding the two endpoints of the range), and the effective DC internal resistance is higher than the preset nominal impedance value; And correspondingly lower space heat dissipation weight The branch optimization unit ignores the thermoelectric penalty factor. The current-limiting constraint actively generates a negative temperature coefficient effect compensation term based on the deviation between the effective DC internal resistance and the preset nominal impedance value, and uses the negative temperature coefficient effect compensation term to generate the comprehensive dispatch weight. This intervention mechanism amplifies the target active power command and uses the estimated Joule heat as a reference benchmark. This is converted into actual high current heat generation, which drives the battery cluster to quickly cross the peak internal resistance region.

[0142] The negative temperature coefficient range is determined based on the physical properties of electrolyte ionic conductivity changing with temperature, and its value is usually between 20℃ and 40℃; the nominal impedance value is determined by searching a pre-stored impedance reference table based on the current state of charge and real-time temperature.

[0143] The actively generated temperature rise utilizes the NTC effect to drive a rapid decrease in internal resistance, thereby crossing the peak internal resistance region. Combining the negative feedback weighting mechanism used to constrain the estimated Joule heat under steady-state conditions, and the compensation term generation mechanism triggered to convert the estimated Joule heat under transient conditions, the final formula for calculating the combined scheduling weight is as follows:

[0144] ;

[0145] In the formula: For the first The overall scheduling weight of each battery cluster; For the first Spatial heat dissipation weight of each battery cluster; For the first Thermoelectric penalty factor for each battery cluster; For the first The effective DC internal resistance of each battery cluster; The preset steady-state weighting coefficient for the system has a value range between 0.5 and 1.5; The transient compensation coefficient is preset for the system, and its value ranges from 0 to 1.0; This is a compensation term for the negative temperature coefficient effect.

[0146] When the system is determined to be in the negative temperature coefficient range and active crossover scheduling is performed, the value of the negative temperature coefficient effect compensation term is determined based on the positive deviation gradient between the effective DC internal resistance and the nominal impedance value, and the value range is between 0.1 and 0.5. Otherwise, the value of this term is 0.

[0147] S52, the instruction normalization unit performs normalization processing on the comprehensive scheduling weight of all battery clusters in the station that are in the release state. Specifically, it calculates the ratio of the comprehensive scheduling weight of a single battery cluster to the sum of the total weights of the entire station, thereby determining the allocation ratio coefficient.

[0148] To ensure the integrity of the algorithm logic, if all battery clusters fail the security check, resulting in a total weight of zero, the instruction unification unit will trigger a system-level shutdown protection message and block the total power instruction.

[0149] Under normal scheduling, the command unification unit multiplies the absolute value of the total power command by the allocation ratio coefficient to determine the target active power to be undertaken by each battery cluster.

[0150] The formula for calculating the allocation ratio coefficient is:

[0151] ;

[0152] In the formula: For the first Allocation ratio coefficient of each battery cluster; For the first The overall scheduling weight of each battery cluster; Indicates the first The overall scheduling weight of each battery cluster; for The sum of the overall scheduling weights of each battery cluster currently in the release state.

[0153] The formula for calculating the target active power is:

[0154] ;

[0155] In the formula: For the first The target active power of each battery cluster; This is the absolute value of the total power command.

[0156] S53, the closed-loop drive unit executes control outputs based on the scheduling level determined during the instruction extraction phase. If the scheduling level is the system-wide level, the closed-loop drive unit encapsulates the target active power into a control message and sends it to the energy management system via an industrial Ethernet bus (such as Modbus TCP or IEC 61850 protocol). After receiving the message, the energy management system issues power adjustment commands to the corresponding branch energy storage converters.

[0157] If the scheduling level is the edge control level, the closed-loop drive unit directly utilizes the real-time task kernel of the edge controller to drive the current control loop of the energy storage converter through local hardwiring or fieldbus (such as the CANopen protocol).

[0158] During the drive process, the edge controller performs closed-loop regulation of the battery cluster current with a preset control cycle between 1ms and 20ms. This control cycle is set according to the switching frequency of the converter IGBT and the bandwidth of the communication link to ensure zero steady-state error tracking of transient power components.

[0159] The control methods and frequency modulation logic of energy storage converters can be implemented by those skilled in the art using existing power electronic control strategies. The specific modulation process is a well-known technology in the field and will not be described in detail here.

[0160] This invention constructs a complete closed loop from topology perception to cross-level control and drive through the coordinated cooperation of the above stages.

[0161] The technical solution and practical application effects of the present invention will be described below with reference to specific application embodiments.

[0162] In a new energy storage power station with a rated power of 1000kW, the system is set to receive a transient frequency regulation command from the grid, and the absolute value of the total power command is... Three representative battery clusters (cluster 1, cluster 2, and cluster 3) from within the energy storage container were selected for calculation. The system's preset parameters are as follows:

[0163] The safety threshold is 50℃, the safety margin coefficient is 0.9, and the conversion constant coefficient is... , Steady-state weighting coefficient Transient compensation coefficient .

[0164] During the topology mapping and state feature update phase, the system obtains the following real-time state parameters of the battery cluster:

[0165] Battery cluster 1 is located at the air inlet, which is a key factor in heat dissipation. Current real-time temperature rate of temperature rise Effective DC internal resistance .

[0166] Battery cluster 2 is located at the end of the air duct, with a high spatial heat dissipation weight. Current real-time temperature rate of temperature rise Effective DC internal resistance .

[0167] Battery cluster 3 is located in the central area, with a high spatial heat dissipation weight. Current real-time temperature rate of temperature rise Its effective DC internal resistance is relatively high due to being in a specific electrochemical polarization state. .

[0168] During the thermoelectric evaluation phase, the boundary constraint unit performs a safety verification comparison. The real-time temperatures of the three battery clusters (25℃, 42℃, and 22℃) are all less than the product of the safety threshold and the safety margin coefficient (i.e., 45℃). All of them generate a safety verification status indicating that they have passed and are released.

[0169] The state quantization unit constructs the thermoelectric penalty factor for each battery cluster based on the thermoelectric penalty formula. .

[0170] Thermoelectric penalty factor of battery cluster 1 ;

[0171] Thermoelectric penalty factor of battery cluster 2 ;

[0172] Electricity penalty factor of battery cluster 3 .

[0173] During the scheduling execution phase, the branch optimization unit determines that the current scheduling instruction is transient and executes the restricted pulse optimization branch.

[0174] The system determines whether each battery cluster is within the negative temperature coefficient range of 20℃ to 40℃ and has an abnormally high internal resistance. Battery cluster 3 currently has a temperature of 22℃ and an internal resistance of 3.5mΩ, which is higher than the preset nominal impedance value. The system triggers active cross-schedule scheduling for battery cluster 3, assigning it a negative temperature coefficient effect compensation term. The negative temperature coefficient effect compensation terms for both battery cluster 1 and battery cluster 2 are 0.

[0175] The branch optimization unit calculates the weight based on the formula for calculating the overall scheduling weight:

[0176] Comprehensive scheduling weight of battery cluster 1 ;

[0177] Comprehensive scheduling weight of battery cluster 2 ;

[0178] Comprehensive scheduling weight of battery cluster 3 .

[0179] The instruction normalization unit normalizes the overall scheduling weights and calculates the sum of the overall scheduling weights of the three battery clusters. .

[0180] Based on the calculation formulas for the allocation ratio coefficient and the target active power, we obtain:

[0181] Allocation ratio coefficient of battery cluster 1 Target active power ;

[0182] Allocation ratio coefficient of battery cluster 2 Target active power ;

[0183] Allocation ratio coefficient of battery cluster 3 Target active power .

[0184] Battery cluster 2, which faces the risk of triggering the thermal limit, is significantly derated to allow its actual heat to converge toward the heat dissipation boundary. Battery cluster 1, which has good heat dissipation, bears a larger reference load. Battery cluster 3, which is in a high internal resistance state, receives the maximum power tilt, amplifies the target active power command, converts the estimated Joule heat into actual high current heat generation, and uses the negative temperature coefficient effect to reduce its internal resistance, thus taking into account both overall safety and system operating efficiency.

[0185] To verify the effectiveness of the technical solution of the present invention, a corresponding full-scale energy storage experimental platform was built for experimental verification and effect comparison.

[0186] The experiment set up two control groups. The control group adopted the capacity-balanced equal-distribution scheduling strategy, while the experimental group adopted the intelligent monitoring power scheduling method provided by the present invention. Under the condition of constant ambient temperature and continuous operation of full-power rated charge and discharge cycle, the data of the system running continuously for 120 minutes were recorded.

[0187] Experimental data showed that after 45 minutes of continuous operation, the battery cluster 2 at the end of the air duct in the control group had poor local heat dissipation, and its real-time temperature reached the limit of 45°C, triggering the forced derating protection. This caused the overall output power of the energy storage power station to drop by 15%, and the maximum temperature difference between the battery clusters to expand to 12.5°C.

[0188] Because the experimental group pre-quantified the spatial heat dissipation weights of each location and combined them with negative feedback penalties based on the temperature rise rate, the heat load was dynamically transferred in space. After 120 minutes of operation, the maximum temperature of all battery clusters stabilized at around 38℃, and the maximum temperature difference between battery clusters throughout the station was controlled within 3.2℃. At the same time, by performing active cross-schedule scheduling on battery cluster 3, the total DC power loss of the experimental group system was reduced by 6.8% compared with the control group.

[0189] See appendix Figure 8 , Figure 8 The CCP presented three independent data sequences, corresponding to battery cluster 1, battery cluster 2, and battery cluster 3, respectively.

[0190] Among them, the column filled with intersecting diagonal lines represents battery cluster 1 with good heat dissipation, and its distribution ratio is stably maintained above 40%.

[0191] The columns filled with dense lattice represent battery cluster 2 under thermal compression, with its distribution ratio suppressed to below 10%.

[0192] The column formed by the hollow blank area represents battery cluster 3, which is performing active internal resistance crossing scheduling and has obtained the highest transient gain allocation within a specific time sequence.

[0193] See appendix Figure 9 , Figure 9 It contains two evolution curves: a thin solid line marked with hollow circles represents the evolution of the highest real-time temperature of the control group, and a thick dashed line marked with solid triangles represents the evolution of the highest real-time temperature of the experimental group.

[0194] The curve of the control group showed a rapid upward trend in the early stage of operation and touched the 45°C horizontal dotted line marked with the safe derating limit in the figure at about 45 minutes.

[0195] The curve of the experimental group remained on a gradual upward trend and eventually converged to a stable period. Throughout the process, it maintained a sufficient margin of distance from the safety derating limit, which shows the effectiveness of this scheme in curbing local hot spots and extending the high-load operating life of the system.

Claims

1. An intelligent monitoring and power dispatching system for a new energy storage power station, characterized in that, include: The topology mapping module is used to obtain the location data of battery clusters in new energy storage power stations and establish matrix parameters that include spatial heat dissipation weights. The status update module is used to obtain the charging and discharging current of the battery cluster and calculate the fluctuation variance, and update the effective DC internal resistance based on the fluctuation variance; The instruction extraction module is used to analyze and judge the acquired total power instruction to generate a scheduling instruction identifier, and to distribute tasks according to the scheduling instruction identifier in order to determine the corresponding scheduling level and scheduling instruction attributes. The thermoelectric assessment module is used to set a trial current for the battery cluster, combine the trial current with the effective DC internal resistance to call the heat generation prediction formula to calculate the estimated Joule heat, combine the space heat dissipation weight to call the thermoelectric penalty formula to obtain the thermoelectric penalty factor, and combine the preset safety threshold to generate a safety verification state. The scheduling execution module is used to allocate a comprehensive scheduling weight based on the security verification status, the scheduling instruction attribute, the estimated Joule heat, the thermoelectric penalty factor and the effective DC internal resistance, generate a target active power based on the comprehensive scheduling weight, and complete power scheduling in combination with the scheduling level.

2. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 1, characterized in that, The topology mapping module includes a parameter calibration unit and a static storage unit; The parameter calibration unit is used to calculate the location data through thermodynamic analysis, calibrate the spatial heat dissipation weight of each battery cluster, and generate the matrix parameters containing the spatial heat dissipation weight. The static storage unit is used to extract the matrix parameters and store the matrix parameters as static constants in the non-volatile memory of the energy management system and edge controller in the new energy storage power station.

3. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 1, characterized in that, The status update module includes a feature acquisition unit; The feature acquisition unit is used to synchronously acquire the real-time temperature, terminal voltage and charging / discharging current of each battery cluster at a set sampling frequency through the underlying sensing device. The feature acquisition unit continuously integrates the charging and discharging current over time and calculates the state of charge of the corresponding battery cluster by combining the initial capacity of the corresponding battery cluster. The sampling frequency is preset based on the hardware sampling capability of the underlying sensing device and the response requirements of the system control cycle.

4. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 3, characterized in that, The state update module also includes an adaptive update unit; The adaptive update unit is used to calculate the fluctuation variance of the charging and discharging current within a set sliding time window; When the fluctuation variance is greater than or equal to the system's preset effective excitation threshold, the adaptive update unit determines that the current state is a fully excited state of the battery cluster, calls the recursive least squares method to calculate the dynamic DC internal resistance of the corresponding battery cluster in real time by combining the state of charge and the real-time temperature, and uses it as the effective DC internal resistance. When the fluctuation variance is less than the system's preset effective excitation threshold, the adaptive update unit determines that the current state is a constant current static state of the battery cluster, freezes the covariance matrix update of the recursive least squares method, extracts the static DC internal resistance under the current operating condition, and uses it as the effective DC internal resistance. The effective threshold for excitation is preset based on the boundary conditions to avoid the covariance matrix divergence due to insufficient external excitation.

5. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 1, characterized in that, The instruction extraction module includes a feature recognition unit; The feature identification unit is used to perform frequency domain analysis and attribute discrimination on the time domain waveform of the total power command, and generate the scheduling command identifier; When the command change rate of the total power command is less than or equal to the preset power change rate threshold, or when the command change rate is greater than the power change rate threshold and the time-domain waveform is not accompanied by high-frequency zero-crossing features, the feature identification unit will use the steady-state command identifier as the scheduling command identifier. When the command change rate is greater than the power change rate threshold and the time-domain waveform is accompanied by the high-frequency zero-crossing feature, the feature identification unit will use the transient command identifier as the scheduling command identifier. The power change rate threshold is preset based on the rated active power output slope of the energy storage power station and the typical inertial response requirements of grid frequency adjustment.

6. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 5, characterized in that, The instruction extraction module also includes a task distribution unit; The task distribution unit is used to distribute the total power command according to the scheduling instruction identifier; If the scheduling instruction identifier is a steady-state instruction identifier, the task distribution unit retains the total power instruction in the centralized layer for global optimization, generates a steady-state instruction attribute as an attribute of the scheduling instruction, and determines the corresponding scheduling level; If the scheduling instruction identifier is a transient instruction identifier, the task distribution unit extracts the power envelope parameters of the total power instruction and sends them out to the edge controller to execute the waveform decomposition task, generate the transient instruction attribute as the attribute of the scheduling instruction, and determine the corresponding scheduling level.

7. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 6, characterized in that, The thermoelectric evaluation module includes a feedforward prediction unit and a state quantization unit; The feedforward estimation unit is used to set the trial current for each battery cluster based on the scheduling instruction attribute, and to calculate the estimated Joule heat of the corresponding battery cluster by multiplying the square of the trial current by the effective DC internal resistance using the heat generation estimation formula. The state quantization unit is used to obtain the real-time temperature of the battery cluster and calculate the corresponding real-time temperature rise rate. Using the thermoelectric penalty formula, the two preset conversion constant coefficients of the system are multiplied by the real-time temperature and the real-time temperature rise rate respectively, and then summed and divided by the space heat dissipation weight to construct the thermoelectric penalty factor. The conversion constant coefficient is predetermined based on the temperature rise experimental data of the battery cluster under different ambient temperatures through multiple linear regression analysis.

8. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 7, characterized in that, The thermoelectric evaluation module also includes a boundary constraint unit; The boundary constraint unit is used to compare the real-time temperature with the safety threshold. If the real-time temperature is greater than or equal to the product of the safety threshold and the set safety margin coefficient, the boundary constraint unit generates a safety verification state indicating failure, forcibly resets the comprehensive scheduling weight of the corresponding battery cluster to zero, and removes the corresponding battery cluster from the scheduling allocation stage of the current control cycle and triggers a re-loop when generating the safety verification state indicating failure. If the real-time temperature is less than the product of the safety threshold and the safety margin coefficient, the boundary constraint unit generates the safety verification state indicating that the test is passed, retains the corresponding thermoelectric penalty factor, and releases the corresponding battery cluster. The safety threshold is preset based on the upper limit of the long-term operating temperature allowed by the battery cell, and the safety margin coefficient is preset based on the sensor measurement error and the thermal gradient non-uniformity margin inside the container.

9. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 8, characterized in that, The scheduling execution module includes a branch optimization unit; After the branch optimization unit extracts the safety verification state indicating that the test has passed, if the scheduling instruction attribute is a steady-state instruction attribute, it executes a steady-state optimization branch, uses the estimated Joule heat as the analysis input to construct an optimization function, and the optimization function approximates the actual heat generation distribution of each battery cluster with the corresponding spatial heat dissipation weight. Based on the positive correlation between the estimated Joule heat and the effective DC internal resistance, the spatial heat dissipation weight is used as a gain term, and the thermoelectric penalty factor and the effective DC internal resistance are used as attenuation terms for weighted calculation. The thermoelectric penalty factor is used as a negative feedback correction term to generate the comprehensive scheduling weight. If the scheduling instruction attribute is a transient instruction attribute, the branch optimization unit executes a restricted pulse optimization branch and determines whether the real-time temperature is within the set negative temperature coefficient range. If the battery cluster is in the negative temperature coefficient range and has a low spatial heat dissipation weight, the branch optimization unit ignores the current limiting constraint of the thermoelectric penalty factor, generates a negative temperature coefficient effect compensation term based on the deviation between the effective DC internal resistance and the preset nominal impedance value, and uses the negative temperature coefficient effect compensation term to generate the comprehensive scheduling weight, converting the estimated Joule heat as a reference benchmark into actual high current heat generation. The negative temperature coefficient range is predetermined based on the physical properties of the electrolyte ionic conductivity changing with temperature.

10. The intelligent monitoring and power dispatching system for a new energy storage power station according to claim 9, characterized in that, The scheduling execution module also includes an instruction normalization unit and a closed-loop drive unit; The instruction normalization unit is used to divide the comprehensive scheduling weight of a single battery cluster by the sum of the comprehensive scheduling weights of all battery clusters in the station that are in the release state, in order to calculate the allocation ratio coefficient, and multiply the absolute value of the obtained total power instruction by the allocation ratio coefficient to generate the target active power of each energy storage converter. The closed-loop drive unit is used to perform power control in conjunction with the scheduling level. When the scheduling level corresponds to a steady-state scheduling command, it drives the corresponding energy storage converter by uniformly issuing control messages through the energy management system. When the scheduling level corresponds to a transient scheduling command, the closed-loop drive unit directly drives the corresponding energy storage converter in a local closed loop through the edge controller to complete the power scheduling.

Images