Intelligent electric meter energy management method oriented to light storage and charging scene and intelligent electric meter

Abstract

The invention discloses an intelligent electric meter energy management method for an optical storage and charging scene and an intelligent electric meter, and relates to the technical field of intelligent electric meters, and the method comprises the steps: firstly, mirroring an original code stream and a compliance smooth fundamental wave through a one-way DMA channel by a legal metering core, and outputting a pure cross-core wake-up signal; performing offset addressing in the S-A-SOH sparse tensor tree in the Flash by the edge management core to obtain an anti-countercurrent power regulation total instruction; splitting into a low-frequency smooth component and a high-frequency compact component through a dynamic mask filter in combination with the temperature and the charge state of the battery cell; and finally, the low-frequency smooth component is issued to fixed energy storage, the high-frequency compact component is subjected to protocol compliance wave packet forming and then sent to the electric vehicle, and compensation and quit control are executed in combination with the direct-current bus capacitor voltage. According to the method, legal metering isolation, network disconnection adjustment, fixed energy storage constraint and vehicle end protocol constraint are considered, and the anti-countercurrent response continuity and adaptation capability are improved.

Description

Technical Field

[0001] This invention relates to the field of smart meter technology, specifically to a smart meter energy management method and a smart meter for photovoltaic, energy storage, and charging scenarios. Background Technology

[0002] Typically placed in industrial and commercial parks, factory parking lots, public charging stations, building rooftop photovoltaic power distribution systems, and microgrid scenarios with grid connection requirements, the site usually simultaneously sets up photovoltaic inverters, energy storage converters, energy storage battery systems, AC or DC charging equipment, grid connection points, power metering devices, and upper-level energy management devices. Currently, most engineering implementations use smart meters or meter-side sampling devices to collect grid connection point power, and external controllers generate or discharge or limit power based on grid connection, energy storage, and load conditions, thereby achieving self-consumption, peak shaving and valley filling, grid connection point power constraints, and charging load management.

[0003] Patent document CN108462211A proposes a method to improve the self-consumption rate of distributed photovoltaic (PV) systems. This document presents an operating system consisting of distributed PV power generation units, loads, a bus, a grid connection point, and an energy storage unit. The distributed PV power generation units are connected to the load and the grid connection point via the bus. The energy storage unit consists of a battery and an energy storage converter connected to the bus. A meter collects the real-time power at the grid connection point. The collected real-time power is compared with a power setpoint using a control error. The control error is then input to the energy storage converter via a digital PI controller to calculate the power setpoint for the energy storage converter. By adjusting the charging and discharging output of the energy storage unit, the power consumption at the grid connection point is kept near the setpoint. Furthermore, the document schedules the charging and discharging times of the energy storage unit according to a daytime charging and nighttime discharging cycle, representing a closed-loop regulation on the energy storage side based on the grid connection point power measurement.

[0004] However, the entry point of the aforementioned existing technologies is mainly based on the power quantity formed by the real-time power at the grid connection point. Furthermore, high-frequency disturbances, converter switching ripples, and transient changes caused by vehicle access or disconnection in the front-end sampled waveform have already undergone metering smoothing or power calculation before entering the control chain. In scenarios where photovoltaics, energy storage, and vehicle charging coexist, the system often only processes disturbances after they have been transmitted to the grid connection point power. This process involves data collection by the meter, comparison by the external controller, and correction of the energy storage output by calling the PI controller. The aforementioned process involves a relatively long control chain, including sampling, filtering, communication transmission, and control calculations.

[0005] Correspondingly, the controlled object is the output of the energy storage unit, and the constraint focus is only on maintaining a constant power at the grid connection point or increasing the self-consumption rate. There is no unified coordination on the acceptance boundary of the vehicle charging side, the charging communication cycle, and the internal tolerance boundary of the rapid power change on the energy storage side. It can be seen that for the photovoltaic-storage-charging coupled operation, the existing technology has the problems of delayed timing and separate boundary processing of each execution object. If this problem is not solved, it is easy to cause the grid connection point adjustment to lag and the control rhythm of the energy storage and charging sides to be inconsistent. The core technical problem caused by this is: how to identify grid connection disturbances earlier and complete the energy regulation coordinated with the energy storage side and the vehicle side while maintaining the normal operation of the metering link in the photovoltaic-storage-charging scenario with smart meters. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a smart meter energy management method and a smart meter for photovoltaic-storage-charging scenarios. The method involves an edge management core using an offset addressing mechanism within a SA-SOH sparse tensor tree in Flash memory to obtain a general command for anti-reverse current power regulation. This command is then combined with cell temperature and state of charge, and split into low-frequency smooth components and high-frequency dense components via a dynamic mask filter. Finally, the low-frequency smooth component is sent to the fixed energy storage system, while the high-frequency dense component is sent to the electric vehicle after protocol-compliant wave packet shaping. Compensation and exit control are then performed in conjunction with the DC bus capacitor voltage. This method balances legal metering isolation, grid disconnection regulation, fixed energy storage constraints, and vehicle-side protocol constraints, thus solving the technical problems described in the background section.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: Energy management methods for smart meters in photovoltaic, energy storage, and charging scenarios include, The legal metrology core mirrors the high-frequency raw code stream acquired by the ADC and the compliant smoothed fundamental wave processed by the legal FIR digital low-pass filter to the cross-core shared read-only FIFO through a one-way DMA channel, and outputs a pure cross-core wake-up signal based on the difference between the two. The edge management core responds to the pure cross-core wake-up signal, offsets and addresses in the SA-SOH sparse tensor tree in the local Flash to obtain the anti-reverse current target power adjustment general command; the anti-reverse current target power adjustment general command is input into the dynamic mask filter, and combined with the cell temperature and state of charge, low-frequency smooth component and high-frequency dense component are obtained. The low-frequency smooth component is sent to the fixed energy storage, and the high-frequency dense component is reconstructed by the protocol compliant wave packet shaper and sent to the electric vehicle. Compensation and exit control are performed in conjunction with the DC bus capacitor voltage.

[0008] Furthermore, the legal metrology core simultaneously mirrors and writes the high-frequency raw code stream and the compliant smoothed fundamental wave corresponding to the same sampling point into the cross-core shared read-only FIFO via a one-way DMA channel. Based on the difference between the two, a twin residual is formed. When the twin residual meets the trigger condition, a pure cross-core wake-up signal is output to the edge management core and the interrupt aggregation blanking window is started.

[0009] Furthermore, the legal metrology core determines the residual slope based on the twin residuals of adjacent sampling points, compares the residual slope with the hardware safety threshold, and switches to software SPI polling when the interrupt aggregation blanking window is locked for a timeout, so as to maintain the subsequent acquisition path of the pure cross-chip wake-up signal.

[0010] Furthermore, the edge management core responds to the pure cross-core wake-up signal, freezes the net power mirror value of the grid connection point, the fixed energy storage health status mirror value, the fixed energy storage adjustable margin mirror value, and the charging interface acceptance level mirror value, and encodes each frozen mirror value into a microgrid state vector for the transformer area, which serves as the addressing input for the SA-SOH sparse tensor tree.

[0011] Furthermore, the edge management core performs trunk addressing on the SA-SOH sparse tensor tree according to the scene index, disturbance index, and health index, and locates the sub-tables under the healthy leaf nodes by combining the margin index and the acceptance index. It prioritizes reading the leaf value code, and reads the backoff leaf value code when there is no corresponding leaf value code, so as to obtain the general command for anti-reverse flow target power adjustment.

[0012] Furthermore, the edge management core reads the cell body temperature, state of charge, and cell diffusion calibration parameters in the local parameter area to determine the digital cutoff frequency of the dynamic mask filter. Based on this, it filters and splits the anti-reverse current target power adjustment command to obtain a low-frequency smooth component and a high-frequency dense component.

[0013] Furthermore, the edge management core writes the low-frequency smooth component into the fixed energy storage power setpoint register, writes the high-frequency dense component into the protocol compliance wave packet shaper execution area, and synchronously writes the current digital cutoff frequency into the filter state area. It also restricts the alternating command from entering the fixed energy storage execution chain when the cell diffusion capability is lower than the preset boundary.

[0014] Furthermore, the edge management core reads the handshake phase field, current allowable range, minimum dwell time and corresponding message sequence number returned by the charging interface controller, reconstructs the high-frequency dense components into a stepped current request sequence ordered by the current transmission cycle, and writes it into the current request field according to the message cycle.

[0015] Furthermore, within the current shaping window, the edge management core determines the remaining energy based on the target energy corresponding to the high-frequency dense component and the actual delivered energy corresponding to the step current request sequence, and maps the remaining energy to the DC bus capacitor voltage rise amount to update subsequent shaping cycles or step currents.

[0016] Furthermore, when the DC bus voltage corresponding to the DC bus capacitor voltage rise reaches the overvoltage exit condition, the edge management core stops integrating the new high-frequency dense component, maintains the transmission path of the low-frequency smooth component, clears the unsent step current request sequence in the current shaping window, and issues a derating curtailment command to the photovoltaic inverter.

[0017] This invention also provides a smart meter for photovoltaic energy storage and charging scenarios, including a legal metering core, an edge management core, a one-way DMA channel, a cross-core shared read-only FIFO, local Flash, a dynamic mask filter, and a protocol compliant wave packet shaper. The legal metering core mirrors the high-frequency raw code stream acquired by the ADC and the compliant smoothed fundamental frequency processed by the legal FIR digital low-pass filter to the cross-core shared read-only FIFO through the one-way DMA channel, and outputs a pure cross-core wake-up signal to the edge management core based on the difference between the two. The edge management core obtains the anti-reverse current target power adjustment general command by offset addressing in the SA-SOH sparse tensor tree in the local Flash. After passing through the dynamic mask filter, it obtains the low-frequency smooth component and the high-frequency dense component. The low-frequency smooth component is sent to the fixed energy storage, and the high-frequency dense component is reconstructed by the protocol compliant wave packet shaper and sent to the electric vehicle. It also performs compensation and exit control in combination with the DC bus capacitor voltage.

[0018] (III) Beneficial Effects This invention provides a smart meter energy management method and a smart meter for photovoltaic, energy storage, and charging scenarios, which have the following beneficial effects: The legal metering core synchronously mirrors the high-frequency raw code stream and the compliant smooth fundamental frequency to the cross-core shared read-only FIFO via a one-way DMA channel, forming a pure cross-core wake-up signal. Disturbances at the grid connection point reach the edge management core before the legal metering link is rewritten, making it suitable for regulating integrated photovoltaic, energy storage, and charging cabinets. The edge management core uses offset addressing in the SA-SOH sparse tensor tree in its local Flash memory, converting online solving into local lookup. This allows the dual-core smart meter to continue outputting the anti-reverse current target power regulation command even in the event of grid outages, weak grid conditions, or load fluctuations, maintaining control chain connectivity.

[0019] The dynamic mask filter combines cell temperature and state of charge to decompose the total command for anti-reverse current target power regulation into low-frequency smooth components and high-frequency dense components, so that stationary energy storage is only responsible for the portion that can be borne, avoiding the direct entry of rapidly changing power into the battery side, and improving the boundary matching relationship when stationary energy storage participates in regulation.

[0020] Low-frequency smooth components flow into stationary energy storage, while high-frequency dense components are sent to the electric vehicle after passing through a protocol-compliant wave packet shaper. Stationary energy storage constraints and vehicle-side protocol constraints are processed separately according to different execution chains, eliminating control conflicts caused by mutual constraints between the two types of equipment. The simultaneous action of protocol-compliant wave packet shaping and DC bus capacitor voltage compensation ensures that the high-frequency dense components are released at a pace acceptable to the vehicle. The DC bus capacitor acts as a short-term buffer during lags, maintaining the order between photovoltaic, stationary energy storage, and electric vehicles.

[0021] A unidirectional DMA channel, a pure cross-core wake-up signal, a SA-SOH sparse tensor tree, a dynamic mask filter, a protocol-compliant wave packet shaper, and DC bus capacitor voltage compensation together form a collaborative chain, enabling smart meters to simultaneously support legal metering isolation, fixed energy storage constraints, and electric vehicle power constraints, thereby improving the scenario adaptability of photovoltaic energy storage and charging systems. Attached Figure Description

[0022] Figure 1 This is a system architecture diagram of the present invention for the photovoltaic storage and charging scenario; Figure 2 This is a diagram of the internal hardware structure of the dual-core smart meter of the present invention; Figure 3 This is a flowchart illustrating the overall process of the energy management method of the present invention. Figure 4 This is a diagram of the homogeneous sampling and cross-chip wake-up mechanism in step one of the present invention; Figure 5 This is the state freezing and sparse tensor tree addressing diagram for step two of the present invention; Figure 6 This is the microscopic solid-phase diffusion constraint flow diagram for step three of this invention; Figure 7 This is the protocol compliance wave packet shaping and DC bus compensation diagram for step four of the present invention. Detailed Implementation

[0023] 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.

[0024] Please see Figures 1-7 This invention provides a smart meter energy management method for photovoltaic, energy storage, and charging scenarios, comprising: For photovoltaic (PV) and solar-storage-charging (PSC) scenarios, the execution entity is a dual-core smart meter. The legal metering core is responsible for power sampling, metering, and compliant output, while the edge management core is responsible for anti-reverse current regulation, energy storage power diversion, and vehicle-side power shaping. Because power fluctuations at the grid connection point first appear at the analog front end in the form of high-frequency ripple, peak harmonics, and phase abrupt changes, and conventional metering links must first perform low-pass smoothing before entering the business thread, the control side often only becomes aware of the actual dangerous disturbance after it has already formed.

[0025] Step one does not modify the measurement caliber of the legal metrology core. By opening a read-only bypass in the underlying physical isolation domain, the filtered information is converted into a priori wake-up signal for the edge management core, providing the earliest trigger moment and a cleaner entry point for sparse tensor space addressing in step two. Step one, without changing the compliance boundaries of the legal metrology link, extracts high-frequency signs of grid-connected power mutations from the original analog sampling in advance and converts them into a pure cross-core wake-up signal.

[0026] In integrated photovoltaic-storage-charging distribution cabinets, the switching actions of photovoltaic inverters, the commutation actions of fixed energy storage bidirectional converters, and the bus absorption actions of electric vehicle charging modules are not synchronized. These actions superimposed at the grid connection point initially manifest as sudden steepening of voltage and current waveform edges, instantaneous rise in harmonic clusters, and localized distortion near zero crossing. After continuous sampling by the analog-to-digital converter at the core front end of the legal metering system, if data is only allowed to travel along the existing legal filtering links, these earliest appearing transient components, which best represent dangerous trends, will be actively flattened by a finite impulse response digital low-pass filter, and subsequently reflected in the metering register as a slower change in mean. At this point, the edge management core takes action based on the regular polling results, but it has usually missed the optimal time slot for pre-occupying the anti-reverse power in step two. Based on this engineering contradiction, a closed-loop system of same-source sampling, dual-track mirroring, read-only listening, and hard interrupt wake-up is adopted: the same analog-to-digital converter sampling result is sent to the legal metrology core on one track to complete the compliant metrology, while the other track mirrors the unsmoothed original high-frequency code stream in a unidirectional manner, and then constructs a twin residual by subtracting it from the compliant smoothed fundamental wave, thereby turning the filtered part directly into preamble information that the edge management core can use.

[0027] To ensure that there is no reverse write path between the legal metrology core and the edge management core, the motherboard of the dual-core smart meter arranges the unidirectional DMA channel, cross-core shared read-only FIFO, unshielded hardware interrupt line and dedicated general-purpose input / output pins in a physical topology on the side of the legal metrology core that only outputs and does not return.

[0028] For example, the legal metrology core uses a metrology microcontroller with dual-buffered sampling registers, while the edge management core uses a low-power microcontroller with an independently running control program. The two do not share executable memory or writable registers; they only transfer mirrored results through a hardware-gated read-only data window. The cross-chip shared read-only FIFO preferably uses a hardware first-in-first-out array, but can also be implemented using a read-only latch array or a read-only dual-port storage window, as long as a unidirectional read-write relationship is maintained. Each time the analog-to-digital converter in the legal metrology core completes a frame sampling, it writes the current sampling frame to the original sampling buffer and simultaneously writes the output processed by a finite impulse response digital low-pass filter to the fundamental frequency sampling buffer. Then, a unidirectional DMA channel transfers the data from both buffers to the cross-chip shared read-only FIFO with a fixed word width. The edge management core does not participate in the data transfer and cannot request retransmissions; it only reads the difference result already existing in the cross-chip shared read-only FIFO upon receiving a non-masked hardware interrupt.

[0029] Even if the edge management core program malfunctions, resets, or shuts down, it will not contaminate the sampling link and measurement results of the legal metrology core. The front-end ADC of the legal metrology core synchronously samples the AC voltage and AC current waveforms at the grid connection point and outputs the original high-frequency code stream arranged according to the sampling timing sequence. The original high-frequency code stream This is a discrete sequence of analog waveforms after quantization. A French-made FIR digital low-pass filter performs compliant smoothing on the same sampled sequence, outputting a compliant smoothed fundamental frequency. The edge management core does not participate in the above sampling and filtering; it only receives the mirrored results.

[0030] During residual construction, instead of performing complex spectral decomposition on the entire waveform, a twin residual that can be stably implemented in the underlying hardware is constructed based on the goal of whether the edge management core needs to be woken up immediately. Let the original high-frequency bitstream be... Compliance smoothing fundamental frequency is The legal metrology core focuses on the original high-frequency code stream of the same sampling point. With compliance smoothing fundamental wave By subtracting, we obtain the twin residuals. : ;

[0031] Among them, the original high-frequency code stream Indicates that the ADC is in the first stage. The discrete sampled values ​​output from each sampling point, without legal filtering, are used to preserve information on high-frequency spikes, edge abrupt changes, and local distortions; they are obtained from the raw digital output after synchronous sampling by the ADC; the fundamental frequency is compliantly smoothed. This represents the discrete sampled values ​​of the same sampling point after processing by a legally compliant FIR digital low-pass filter, used as a smoothing reference under compliant metrology standards; it is obtained by the legal metrology core calling the legally compliant filtering link. The result of simultaneous processing; Twin residuals This represents the differential components filtered out by the legal filtering link, used to characterize the degree of deviation between the original disturbance and the compliant fundamental frequency; subsequently, the legal metrology core uses the twin residuals of adjacent sampling points... Calculate the residual slope : ; Among them, residual slope This represents the rate of change of the twin residuals between adjacent sampling points, used to distinguish between slowly varying perturbations and abrupt perturbations; sampling interval This represents the time interval between two consecutive ADC samples, used to convert the residual difference into a rate of change; it is obtained by reading the ADC sampling clock configuration value from the legal metrology core; the current snapshot residual... Residual from the previous shot All data originates from the internal buffer of the legal metrology core before being written to the cross-core shared read-only FIFO.

[0032] The advantage of this definition is that the implementation path is extremely short. It does not require floating-point operations or the introduction of additional models, so it is easy to embed in the interrupt pre-level hardware.

[0033] The key to same-source sampling is not to sample once more, but to ensure that the original high-frequency code stream and the compliant smooth fundamental frequency come from the same sampling time, the same range, and the same front-end gain link, thus eliminating spurious differences caused by asynchronous sampling from the source.

[0034] To this end, the legal metrology core first freezes the current frame of the analog-to-digital converter in each sampling cycle. Then, one data path directly sends the frozen frame to the original sampling buffer, while another data path sends the same frozen frame to a finite impulse response digital low-pass filter. The filter coefficients are preferably stored in read-only memory to prevent the edge management core from changing the metrological smoothing caliber through parameter downloads. Subsequently, the unidirectional DMA channel moves data in the order of the original sampling buffer first, followed by the fundamental frequency sampling buffer, forming fixed-format data pairs within the cross-core shared read-only FIFO. Whenever the edge management core is woken up, it reads only time-aligned data pairs, rather than two misaligned waveform segments.

[0035] For example, in an outdoor integrated cabinet, the photovoltaic inverter experiences a power surge due to cloud shadow movement, while the stationary energy storage maintains its original power. The electric vehicle is just entering the constant current to constant voltage transition phase, at which point the current edge at the grid connection point will suddenly become steeper. The legal metrology core does not change the original metering process; it simply pushes the two expressions sampled in the same frame into the cross-chip shared read-only FIFO simultaneously. The edge management core then reads not a vague indication of a potential power increase, but rather paired information showing a significant rise in the original high-frequency code stream within the same frame, while the compliant smooth fundamental frequency remains relatively flat, thus providing the prerequisite for proceeding to step two.

[0036] When using this method, ensure that the difference calculation is performed on the same source data and do not regard the sampling time series offset as a disturbance; ensure that the measurement link with legal metrology as the core operates independently and that there is no compliance risk caused by bypassing; and provide the same short-path data for subsequent residual slope calculation.

[0037] To reduce the impact of high-frequency isolated cusps on the judgment, the legal metrology core preferably performs median preservation on three consecutive residual slope samples before comparing the results with a safety threshold. Comparison. Among them, the safety threshold. The residual slope threshold required to trigger an unmasked hardware interrupt is used to distinguish between normal switching ripple and leading disturbances that could cause the grid connection point power to exceed the limit rapidly. It is preferable to pre-write the one-time configuration area during the factory calibration stage based on the front-end range of the target meter position, the installation scenario, and the inverter switching frequency band.

[0038] In practice, residual slope determination does not rely on operating system threads but is embedded in the fast processing near the sampling end of the legal metrology core. As soon as the residual slope exceeds the safety threshold, the unmasked hardware interrupt line is immediately set, and the edge management core can be woken up without waiting for serial port polling or shared variable polling. If the residual slope does not exceed the safety threshold, the mirror link continues to work but does not interrupt the billing reporting or heartbeat tasks currently being executed by the edge management core.

[0039] When in use, continuous waveform differences are compressed into a single determinable event, which is easy to implement within small hardware resources. The triggering condition falls on the rate of change rather than the absolute amplitude, which is closer to the leading characteristics of network disturbances and provides a clear and sparse basis for suspending low-priority threads in the subsequent edge management core.

[0040] Rapid wake-up alone is insufficient to guarantee field availability. In photovoltaic-storage-charging scenarios, electromagnetic interference from inverter clusters, contactor jitter in charging modules, and crosstalk from nearby cables can all cause multiple consecutive threshold crossings within a very short period. If each threshold crossing unconditionally triggers the edge management core, the edge management core will fall into a state of continuously responding to unmasked hardware interrupts and will be unable to return to the main process.

[0041] Therefore, interrupt aggregation blanking window is started simultaneously with setting the non-maskable hardware interrupt. And within that time window, newly added threshold-crossing events are merged and maintained. Interruption aggregation blanking window. The preferred setting is 10 to 20ms. Its purpose is not to delay the initial wake-up, but rather to mask repeated interrupts after the initial wake-up, allowing the edge management core to handle a cluster of related disturbances with a single context switch. This is done within the interrupt aggregation blanking window. If threshold overflows occur continuously, the hardware gating logic only refreshes the sticky flag indicating a subsequent disturbance, without repeatedly raising the interrupt line. Interrupt aggregation blanking window. If the adhesion flag still exists after the event ends, the next threshold re-setting of the non-maskable hardware interrupt is allowed.

[0042] For example, during peak evening charging hours in an underground parking lot, multiple vehicles begin switching charging phases almost simultaneously, resulting in dense spikes superimposed on the input side of the charging module and the output side of the photovoltaic inverter. Without an interruption aggregation blanking window, the edge management core would be interrupted again before one process is completed, causing the spatial addressing in step two to fail to enter stably. With the addition of the interruption aggregation blanking window, the edge management core is woken up once, and then only receives information about the continued sticking of disturbance clusters during the window period. It decides whether to accept the next wake-up call after completing a power decision.

[0043] Considering that extreme electromagnetic interference may still cause the interrupt line to lock for a long time, a degraded holding path is further set up. When the hardware timer detects a lockout timeout, the hardware logic gate first removes the interrupt line, and then changes the edge management core reading method to polling the common software serial peripheral interface. The response speed is reduced, but the legal metering core retains the original metering link. The edge management core can continue to perform basic energy management at a slower pace without falling into an interrupt storm.

[0044] In use, to avoid continuous threshold exceedances that could drag the edge management core into deadlock, the value of the initial fast wake-up is preserved, the most critical preamble time is not sacrificed, and a fallback path is provided even in extreme situations, ensuring that the overall system performance is degraded rather than failed. A cross-core wake-up signal is output through the concatenation of homogeneous sampling, dual-track mirroring, twin residual determination, and interrupt aggregation blanking window. This signal carries a clear determination of whether a rapid disturbance has occurred and proceeds to step two, allowing the edge management core to avoid high-cost parsing of every frame of measurement data, and instead only suspend low-priority tasks and perform sparse tensor space addressing when control slots are truly needed.

[0045] Step 2: Under the condition that the available storage and computing power of the edge management core are limited, the pure cross-core wake-up signal delivered in Step 1 is concatenated with the local cache state to form the microgrid state vector of the substation area, and the anti-reverse current target power adjustment general command is directly obtained by offset addressing. .

[0046] In optical storage and charging cabinets, the edge management core is typically still an auxiliary controller on the metering side, with less random access memory capacity, floating-point computing power, and thread concurrency margin than a dedicated gateway. If online rolling optimization, matrix inversion, or iterative search paths are used, the pre-emptive time slots just acquired in step one will be consumed again during the solution phase. Based on this constraint, step two does not design the edge management core as an online solver, but rather as an offline knowledge indexer: the cloud first compresses the high-dimensional policy space into a static dictionary, and after receiving a clean cross-core wake-up signal, the edge management core only performs four actions: state freezing, interval encoding, offset addressing, and quantization inverse solution. In this way, step one is responsible for detecting disturbances as early as possible, and step two is responsible for sending the disturbances to the decision-making point as quickly as possible, without burdening the edge management core with heavy computational loads.

[0047] After the legal metrology core sends a clean cross-core wake-up signal in step one, the unmasked hardware interrupt service routine of the edge management core first suspends the billing reporting thread, heartbeat maintenance thread, and remote parameter receiving thread, retaining only the clock reference, watchdog timer feeding, and local communication receive buffer to continue running. Subsequently, the edge management core reads the data pairs associated with the current disturbance from the cross-core shared read-only FIFO and simultaneously freezes the photovoltaic inverter power mirror value, fixed energy storage management unit message mirror value, charging interface handshake mirror value, and grid connection point net power register mirror value that have been cached in the most recent scheduling cycle. This freezing does not shut down external devices; instead, it copies the current values ​​of each mirror register to the interrupt context buffer at once, ensuring that the entire addressing process is based on data slices from the same moment. After freezing, the edge management core converts continuous values ​​into discrete indices according to the interval table pre-written in Flash, then directly extracts the leaf value encoding according to the SA-SOH sparse tensor tree's level offset table, and finally recovers the anti-reverse current target power adjustment general instruction based on the quantization ratio. .

[0048] The engineering intent of state freezing is to transform a set of changing quantities in the field into a set of tags that can be addressed at once. To this end, after entering an unmasked hardware interrupt, the edge management core first locks the current read pointer of the cross-chip shared read-only FIFO, then copies the mirror value of the grid-connected point net power register, the state of charge and health fields in the fixed energy storage management unit, and the acceptance level field in the charging interface handshake message.

[0049] To eliminate online floating-point comparisons for addressing paths, all these frozen quantities are mapped to a finite number of discrete intervals. Preferably, the net power of the grid connection point is encoded as a scene index. The clean cross-core wake-up signal output in step one is encoded as a perturbation index. The health status of stationary energy storage is coded as a health index. The current available release or absorption capacity of stationary energy storage is encoded as a margin index. The current power levels that the charging port can accept are coded as acceptance indices. .

[0050] The above five indices together constitute the state vector of the transformer substation microgrid: ; Among them, the state vector of the microgrid in the transformer area The discrete state set used in this round of decision-making serves as the sole input for subsequent offset addressing, and its value range is determined by the combination space of five indices; Scene Index The segment containing the net power at the grid connection point serves to distinguish between the reverse power transmission edge, the balance zone, and the receiving zone. Its value range is a preset set of segment numbers for the scenario. Disturbance index. The disturbance level corresponding to the pure cross-core wake-up signal is used to explicitly pass the trigger information from step one to step two. The value range is the preset disturbance level number set. Health Index : Fixed energy storage current health level, its function is to limit the risk boundary of subsequent power output, and its value range is a preset set of health level numbers; margin index : Fixed energy storage current adjustable margin level, its function is to reflect the absorption or release space that the energy storage side can still undertake, and the value range is the set of preset margin level numbers; Acceptance Index The charging interface currently accepts certain power levels. Its purpose is to prevent commands that significantly exceed the vehicle's receiving capabilities from being directly sent to subsequent steps. The value range is a preset set of acceptable power levels. Preset set of acceptance level numbers Allowable current range returned by the charging interface controller during the current handshake phase Minimum stay time Vehicle-side voltage representative value and preset step size The acceptable power range is obtained by discretely dividing it into intervals and then writing it into the interval table or update package.

[0051] Microgrid state vector At least include: Scene Index Disturbance Index Health Index Margin Index and Acceptance Index Among them, the SA-SOH sparse tensor tree is used to complete the scene indexing. Disturbance Index and health index Three-level backbone addressing, margin index With Acceptance Index Used to locate the sub-tables under the corresponding healthy leaf nodes.

[0052] As a supplement: The cloud side is based on a discrete state set. With action set Construct the teacher action value function For each discrete state The set of feasible actions that satisfy grid connection constraints, stationary energy storage constraints, and vehicle acceptance constraints. Within the range, select the action with the largest action value as the original action value: ; Among them, discrete state Indicates index by scene Disturbance Index Health Index Margin Index and Acceptance Index A combination of states; obtained by enumerating states in an offline sample set. Set of actionable actions. Indicates the state The set of candidate actions that satisfy the equipment boundary and constraint conditions; it is obtained by a cloud-based constraint filtering module based on the grid-connected power boundary, fixed energy storage power boundary, and vehicle acceptance boundary. Teacher action value function. This represents the state-action value function obtained after offline training in the cloud, used to evaluate actions. In state The resulting profits.

[0053] raw action value Representing state The corresponding optimal anti-reverse flow target power adjustment command is used as the blade value reference before local quantization.

[0054] Teacher action value function The state in For a set of discrete states Any state element in the equation satisfies Discrete state For a set of discrete states The Middle A state element can be written as State vector of the microgrid in the transformer area The five-element state vector, after being frozen and encoded on-site, can be written as: .

[0055] For example, in a scenario where rooftop solar power and parking spaces share a common access point, as the clouds disperse, the output of the solar inverter increases, and the vehicle has just entered the constant current phase. The clean cross-core wake-up signal given in step one already indicates a rapid upward trend in the grid connection point. After receiving an unmasked hardware interrupt, the edge management core immediately copies the current net power of the grid connection point, the fixed energy storage state of charge, the fixed energy storage health status, and the vehicle-side acceptance level to the interrupt buffer, and then writes a scenario index based on the interval table in Flash. Disturbance Index Health Index Margin Index and Acceptance Index Subsequent addressing deals not with the drifting original values, but with structurally stable discrete labels.

[0056] When in use, state freezing eliminates the decision-making gaps caused by asynchronous changes in multiple registers, and discrete coding combines field quantities from different sources and with different dimensions into an index system, allowing the edge management core to proceed to the next step without floating-point iteration.

[0057] After discrete encoding, the edge management core will not perform linear traversal. Instead, it will directly address the hierarchy offset table of the SA-SOH sparse tensor tree. The sparse tensor tree is placed in the local noRFLash or on-chip EFLash. The running area and the update area are mirrored. In the middle of the running area, the scene offset table, perturbation offset table, health offset table, margin offset table, receiving leaf value table and backoff leaf value table are placed in sequence.

[0058] The rollback leaf value table is generated synchronously by the cloud during the offline strategy compilation phase. For uncovered state combinations or state combinations with missing fields, the cloud selects the lower-risk rollback action from conservative candidate actions that meet grid connection constraints, fixed energy storage constraints, and vehicle acceptance constraints. Then according to the scaling factor and zero point Quantization to backtrack leaf value encoding Write the value to the rollback leaf value table and send it to the local Flash along with the update package.

[0059] The offset table of each layer reflects the starting offset for entering the next layer of effective child node blocks. That is, states not covered by the training policy do not occupy storage space. The edge management core obtains the microgrid state vector of the transformer area. Then, the leaf value address can be directly calculated using the offset formula: ; Among them, leaf value address The current state combination's offset position in the receiving leaf value table directly points to the quantized leaf value code to be read, and its value range is the effective address range of the runtime leaf value table; scene offset. : Indexed by scene The selected first-level starting offset is used to locate the perturbation subtree corresponding to the current scene. Its value range is the set of valid offset values ​​in the scene offset table; the perturbation offset is... : Index of a given scene Indexed by perturbation under certain conditions The selected second-level starting offset is used to locate the healthy subtree corresponding to the current perturbation. Its value range is the set of valid offset values ​​in the perturbation offset table. Health offset : Index of a given scene and perturbation index Health Index under Conditions The selected third-level starting offset is used to locate the margin quantum tree corresponding to the current health state. Its value range is the set of valid offset values ​​in the health offset table; margin offset. : Using margin indexes given the first three levels of indexes The selected fourth layer starting offset is used to locate the receiving leaf value segment under the current margin state, and the value range is the set of valid offset values ​​in the margin offset table. Acceptance Index Here, the final leaf displacement is used to select the leaf value code corresponding to the vehicle acceptance level from the current acceptance leaf value segment. The value range is the preset acceptance level number set. Each offset is folded to the starting position of the next level during offline compilation, so the edge management core only performs table entry reading and integer addition on-site. If a certain state combination does not have a corresponding receiving leaf value segment in the runtime area, the edge management core immediately reads the fallback leaf value table entry recorded in the previous level and prioritizes the command on the side with lower risk.

[0060] When used, sparse storage removes unnecessary state combinations from storage, allowing limited Flash memory to hold only valid strategies; offset addressing converts multidimensional state access into sequential list reading and integer accumulation; and back off leaf value tables enable uncovered combinations to also have penetrable output.

[0061] The following content is used to explain quantized leaf value encoding. The cloud-based offline generation process differs from the previous description of the edge management core reading quantized leaf value encoding from local Flash. And execute the local process of dequantization recovery.

[0062] After obtaining the leaf value address Subsequently, the edge management core reads the quantized leaf value code from the receiving leaf value table. Based on the proportional factor and zero point stored in the same table header, the reverse solution is performed to recover the general command for anti-reverse current target power regulation. : ; Among them, the general command for anti-reverse current target power adjustment The net power adjustment target that should be handled in step three in this round is to uniformly accept the results obtained in step two and enter the subsequent frequency domain splitting. The value range is determined by the compilation boundary and scaling factor of the receiving leaf value table. Scale factor : Scaling factors used in quantization leaf value encoding to recover power quantities. Their function is to restore discrete codes to physical power quantities. They are a preset set of scaling factors with positive values. Quantization leaf value encoding Leaf value address The leaf value it points to is a signed integer, which carries the decision result compressed into the local Flash after offline training. Its value range is the INT8 encoding range. midnight The zero-power reference point for quantized leaf value encoding is used to correct the offset between the signed integer leaf value and the actual zero power point. Its value range is a preset set of integer zero points. In the preferred embodiment, the edge management core runs a trimmed operating system. The SA-SoH sparse tensor tree is compiled using an offline toolchain and written to the local Flash runtime area. Both the quantization leaf value table and the rollback leaf value table have page-level verification fields. The actual operation process is as follows: After the clean cross-core wake-up signal is prompted in step one, the edge management core freezes its state and generates the substation microgrid state vector. The leaf value address is obtained through the offset formula. Read the quantized leaf value encoding The reverse solution yields the overall command for adjusting the target power to prevent backflow. Along with scene index and health index Write to the input buffer from step three.

[0063] As a supplement: The cloud-based system constructs a state-action value table based on historical and simulated operation samples, and indexes the scenarios. Disturbance Index Health Index Margin Index and Acceptance Index The resulting discrete state combinations are used to screen candidate actions that satisfy grid connection constraints, fixed energy storage constraints, and vehicle acceptance constraints. The candidate action that maximizes the target benefit is then selected as the original action value for that discrete state. The cloud side is then adjusted according to a scaling factor. and zero point For the original action value Perform INT8 quantization to obtain the leaf value code. : ; Among them, the original action value This represents the unquantized anti-reverse current target power adjustment total command corresponding to the discrete state combination. Its purpose is to serve as the baseline action value before quantization. It is obtained from the offline strategy screening results on the cloud side.

[0064] Scale factor This is a scaling factor for the power value within the Int8 encoding range, used to calculate the quantization step size; it is written to the update packet via offline distribution of action values ​​on the cloud side. (Zero point) This indicates the corresponding position of zero power within the INT8 encoding range, used for offset conversion between signed and unsigned power storage formats, and written to the update package via offline settings on the cloud side. Leaf value encoding. This indicates the INT8 action code written to the local Flash leaf node, used for local dequantization recovery. Saturation operator. This indicates a truncation operation that cuts the quantization result within the Int8 encoding range to prevent encoding overflow.

[0065] The edge management core reads leaf value codes from local Flash. Then, based on the scaling factor in the same update package and zero point Perform inverse quantization recovery to obtain the overall command for anti-reverse current target power regulation. : .

[0066] In a parallel approach, if a mirror field is missing, a conservative index is filled into the interval encoding during interval encoding, and the backtracking leaf value table outputs a more conservative anti-reverse current target power adjustment general instruction. If page-level verification in the running zone fails, switch to the most recently verified mirror page in the update zone.

[0067] In use, quantization reverse decoding restores the offline compression strategy to power commands that can be directly consumed in subsequent steps; dual-zone mirroring and page-level verification ensure that there are still continuous and usable addressing paths when the local Flash is updated or encounters anomalies; conservative indexes and rollback leaf value tables include missing fields in the normal processing range, ensuring that the output of step two can always be passed to step three, and that the preamble time slots obtained in step one are not consumed again in the online solution stage, i.e., the total power regulation command for anti-reverse current target. This establishes a single entry point for the frequency domain shunting under electrochemical constraint in step three.

[0068] Step 3: Issue the general command for adjusting the target power to prevent backflow. Mapped to the frequency boundary that the microscopic solid-phase diffusion of stationary energy storage can withstand, and then split into low-frequency smooth components that can be directly injected into stationary energy storage. High-frequency dense components that need to be transferred to subsequent processing stages .

[0069] The propagation speed of power commands within a photovoltaic energy storage and charging cabinet is determined by the switching cycles of the control chip, isolation drive, and bidirectional converter, while the material migration speed within a fixed energy storage system is determined by the electrode particle size, diffusion channels, and temperature field. These two types of speeds are inherently on different levels. If only a fixed empirical cutoff frequency is used to filter and prevent backflow, the target power regulation command will be affected. This means that the same set of filter parameters will mismatch across all operating conditions, including both high-temperature, high-charge and low-temperature, low-charge states. Therefore, step three adopts a path of first determining the material boundary and then shaping the digital filter to adjust the digital cutoff frequency. Anchored to the current microscopic solid-phase diffusion time constant of fixed energy storage This makes the filter's response capability directly determined by the material at the moment of fixed energy storage.

[0070] The edge management core receives the general command for anti-backflow target power adjustment. Then, the cell body temperature is first read from the fixed energy storage management unit. State of charge The system retrieves the particle characteristic length, reference diffusion factor, and activation energy parameters corresponding to the current cell serial identifier from the local parameter area.

[0071] In a preferred embodiment, the cell body temperature The state of charge is collected by a thermistor installed at the base of the battery cell tab. Based on the coulomb metering and open-circuit voltage correction results within the fixed energy storage management unit, the edge management core reads these fields via an isolated serial bus and freezes them into a single-step input at the same scheduling point. Subsequently, the edge management core first calculates the effective diffusion factor, then the microscopic solid-phase diffusion time constant, subsequently generates the digital cutoff frequency, and finally applies the anti-reverse current target power regulation command. Frequency domain tearing is performed by feeding the data into a dynamic mask filter.

[0072] Edge management core in each control cycle Based on the current digital cutoff frequency Update dynamic mask filter coefficients : ; Subsequently, based on the filter coefficients Calculate low-frequency smoothing components : ; Then, calculate the high-frequency compact component based on the difference between the total command and the low-frequency smooth component. : ; Control cycle This is the time interval between two scheduling operations of the edge management core, which can be used for discretization filtering; the edge management core scheduling clock is the edge management core scheduling clock. Filter coefficients This refers to the low-pass hold ratio during the current control cycle, which is the dynamic cutoff frequency. Weighting coefficients for current and historical instructions; edge management core obtained through digital cutoff frequency; low-frequency smoothing component. This refers to the power component within the current tolerable boundary of stationary energy storage, which is then distributed to stationary energy storage. High-frequency dense component. It is the power component that is outside the current acceptable boundary of fixed energy storage, and is processed by step four.

[0073] when or At this time, the dynamic mask filter enters a protective state, allowing only the baseline component to enter the fixed energy storage execution chain.

[0074] Step three first addresses the physical upper limit of how quickly the fixed energy storage can respond, without processing the power command itself. To this end, the edge management core solves for the effective diffusion factor based on the Arrhenius diffusion model stored in its local parameter area. : ; Among them, effective diffusion factor Current cell materials at cell body temperature With state of charge The equivalent diffusion capacity under the combined effect serves as the time constant for subsequent microscopic solid-phase diffusion. The input value is within the diffusion capability range greater than 0; it is obtained by the edge management core based on the sampled value and calibration parameters.

[0075] Basic diffusion factor The material diffusion factor of a fixed energy storage cell refers to the material diffusion factor based on the material state. Its value range is obtained from the cell model calibration and is written into the local parameter area. The parameter is obtained through diffusion tests or equivalent model fitting during cell selection and is also written into the parameter area. Activation energy parameter. The material parameters that allow diffusion to cross the potential barrier are used to characterize the effect of temperature changes on diffusion capacity. The range of values ​​is obtained by calibration of the battery cell system. The parameters are obtained by cell calibration and written into the parameter area.

[0076] gas constant Thermodynamic constants, whose function is to unify the activation energy parameter. With the temperature of the battery cell body The dimensional relationship of the values ​​is a fixed constant within a fixed range; the cell body temperature The temperature of the cell or module currently being regulated is a representative point temperature. Its function is to reflect the strength of particle thermal motion, and its value range is limited by the temperature sampling link of the fixed energy storage management unit. State of charge correction function State of charge The correction factor for diffusion capability is used to incorporate the diffusion differences corresponding to different lithium insertion depths into the effective diffusion factor. The value range is a dimensionless interval greater than 0; preferably, a piecewise interpolation table is established based on the pulse test or AC impedance test results of multiple state-of-charge points, and the value is retrieved or interpolated in the edge management core according to the table. The state-of-charge correction function... The piecewise linear form is as follows: ; Among them, the state of charge correction function It represents the correction coefficient of the state of charge to the effective diffusion factor, used to characterize the difference in diffusion capacity of fixed energy storage under different state of charge ranges; it is obtained by the edge management core looking up and substituting the current state of charge and pre-stored segment parameters into a table for calculation.

[0077] State of charge This is the current charge level of the stationary energy storage, used as the input variable for the state of charge correction function. It is obtained from the state of charge output by the stationary energy storage management unit. (Segment boundary) It is the dividing point of the state of charge interval, used to divide the continuous state of charge into multiple fitting intervals; it is obtained by the parameter storage area obtained during cell calibration based on the diffusion test or impedance test results of different states of charge.

[0078] linear coefficients The slope of the correction function within each state of charge interval characterizes the influence of state of charge changes on the diffusion capability correction coefficient within that interval. It is obtained by calibration within the corresponding interval, with the intercept parameter being the intercept of the correction function within each state of charge interval. This intercept, along with the corresponding slope, determines the linear correction relationship within the interval. Mark the number of segments within the corresponding interval. The total number of intervals for each state of charge correction function represents the discrete accuracy of that state of charge correction function; it is obtained by determining the edge management core computing resources under calibration conditions.

[0079] Preferably, the state of charge When using normalized caliber representation, , The state of charge When expressed as a percentage, , The endpoints of adjacent intervals are arranged consecutively, satisfying... This ensures that the state of charge correction function provides complete coverage of the entire range of states of charge.

[0080] In a preferred embodiment, the state of charge correction function This is achieved using a piecewise cubic interpolation table, with interpolation nodes pre-written into the local parameter area. In parallel implementations, the state-of-charge correction function... Implemented using a piecewise linear list to reduce the number of multiply-accumulate operations in the edge management core. For example, when photovoltaic power increases in the afternoon while the vehicle is still receiving power, the edge management core first reads the temperature of the battery cell itself. With state of charge Then, the reference diffusion factor corresponding to this batch of cells is retrieved from the parameter area. and activation energy parameters Finally, the effective diffusion factor was obtained. .

[0081] In practice, the current material response capability of fixed energy storage is specifically expressed as the effective diffusion factor. Subsequent filtering no longer relies on fixed empirical values; the temperature sampling chain, the state of charge chain, and the local parameter region form a closed input, enabling the same type of battery cell to obtain different boundaries at different times.

[0082] After obtaining the effective diffusion factor Next, step three continues by converting the material capabilities into time and frequency scales usable by the filter. The edge management core first calculates the microscopic solid-phase diffusion time constant based on the particle diffusion approximation relationship. Then, it isomorphically converted to a digital cutoff frequency. : ; Among them, the microscopic solid-phase diffusion time constant The characteristic time required for lithium ions to diffuse from the surface of electrode particles to the interior of the particles defines the time boundary for a fixed energy storage device to smoothly absorb power changes; its value is within a time interval greater than 0. Particle characteristic length. : Effective diffusion path length of electrode particles involved in diffusion modeling; its role is to incorporate material geometry into the time constant solution; its value range is determined by the cell electrode formulation and the particle size calibration results after compaction; Effective diffusion factor Continuing here as input, its function is to determine the feature length of the same particle. The speed of down-diffusion completion; digital cutoff frequency Dynamic mask filters currently allow penetration into the highest frequency boundary of fixed energy storage; their function is to reduce the diffusion time constant of the microscopic solid phase. Directly translated into frequency thresholds in digital signal processing, the value range is a frequency interval greater than 0 and limited by the sampling beat frequency; pi : A constant, whose function is to perform conversions between the time domain and the frequency domain; In a preferred embodiment, the particle characteristic length Given from the cell model parameter table. For cylindrical cells, the characteristic length of the cell is... Take the effective diffusion radius of the active particles; for square aluminum-cased cells, the characteristic length of the particles. The diffusion radius of the active particles is still used to avoid mistakenly substituting the shell size into the model. When the outside temperature drops at night in winter and the parking space continues to receive power, the cell body temperature... The decline led to an effective diffusion factor Contraction, microscopic solid-phase diffusion time constant Consequently, the digital cutoff frequency is lengthened. It descends automatically.

[0083] When in use, the digital cutoff frequency It is no longer a code constant, but is directly given by the current material state of the fixed energy storage; particle characteristic length. The introduction of this feature solidifies the physical differences between different cell systems into the local parameter area.

[0084] Obtaining the digital cutoff frequency Subsequently, the edge management core updates the current coefficients of the dynamic mask filter accordingly and adjusts the overall command for anti-backflow target power. Split into low-frequency smooth components With high-frequency dense components : ; Among them, high-frequency dense components : No. Among the scheduling points, the current digital cutoff frequency is exceeded. The portion of power that cannot be directly integrated into fixed energy storage within the permissible range serves to be reserved for protocol compliance wave packet shaping and physical compensation in step four. Its value range is subject to the overall command for backflow prevention target power regulation. With low-frequency smoothing components The range of differences is limited; Anti-backflow target power adjustment general command : No. The total power command that the scheduling point enters in step three serves as the complete input before splitting, and its value range is limited by the output boundary of step two. Low-frequency smoothing component : No. The power portion of the scheduling points that can be injected into the fixed energy storage is retained by the dynamic mask filter. Its function is to directly send it to the fixed energy storage for execution, and its value range is affected by the digital cutoff frequency. The previous filtering state and the current tolerance limit of fixed energy storage are both limiting factors. Dispatch Point : The current discrete scheduling sequence number, which identifies the control timing position of the split operation. The value range is determined by the continuous scheduling count of the edge management core.

[0085] In a preferred embodiment, the dynamic mask filter employs a fixed-point first-order recursive structure, with the filter coefficients determined by the digital cutoff frequency. The low-frequency smoothing component is obtained by looking up a table together with the current scheduling frame width. High-frequency dense components updated recursively The difference is given directly. The edge management core smooths the low-frequency components. Write to the fixed energy storage power setpoint register, and simultaneously add the high-frequency dense component. and current digital cutoff frequency Write to the input buffer in step four. If the cell body temperature... The effective diffusion factor continued to decline, causing it to decrease. When the signal approaches the lower threshold set in the local parameter area, the edge management core directly switches the dynamic mask filter to DC pass-through protection mode: low-frequency smoothing component. Only the slow baseline component and the high-frequency dense component are retained. In the fourth step of the full transfer, alternating commands are no longer allowed to impact stationary energy storage. In the parallel implementation, the cell body temperature... Take the module's representative temperature; state of charge A conservative value is taken within the charged state window; the dynamic mask filter is implemented using a second-order structure after bilinear transformation.

[0086] For example, when the photovoltaic power declines in the evening and the vehicle charger is still increasing its power absorption, the general command for anti-reverse current target power adjustment given in step two... There will be a shift in direction. At this point, the edge management core will not directly push the entire instruction to the fixed energy storage, but will instead write the portion that can be absorbed by the current material boundary as a low-frequency smooth component. The energy is transferred to fixed energy storage, and the remaining portion is written as high-frequency compact components. Pass it to step four.

[0087] In use, the dynamic mask filter breaks down the issue of whether execution is possible into determining which part of the fixed energy storage should be executed first and which part should be handled in subsequent stages; degradation protection incorporates extreme cold or diffusion capacity contraction scenarios into the normal control chain; and parallel implementation methods preserve equivalent implementation paths for different cell packages and filter structures. Step three directly presses the material-scale boundary into the digital filter chain, preventing the fixed energy storage from undergoing power transitions exceeding its internal diffusion capacity under the dominance of power electronic bandwidth.

[0088] Step 4: High-frequency dense components The energy is reconstructed into a low-frequency stepped wave packet that conforms to the vehicle-side communication constraints, and the energy accumulation formed during the reconstruction process is transferred to the DC bus capacitor absorption path of the photovoltaic inverter, thereby completing the energy management closed loop without disrupting the vehicle's power supply continuity.

[0089] Step three has broken down the anti-reverse current target power adjustment command into two parts, one of which is the low-frequency smoothing component. Provided by stationary energy storage, high-frequency dense components This will be addressed later. The real challenge at this point is no longer whether there is surplus power, but whether the surplus power can be delivered in a way that the vehicle is willing to accept. The electric vehicle's power receiving link consists of an onboard battery management system, a charging interface controller, DC contactors, and a charger. Its acceptance logic is not a continuous function, but a discrete protocol process with stage fields, minimum dwell time, and message update cycles. High-frequency dense components. Even if the total energy is correct, the vehicle will still reject the signal if the rising edge is too fast or the step duration is too short. Therefore, step four adopts a processing approach of first rewriting the power configuration according to the protocol, and then having the hardware absorb the rewriting delay.

[0090] The edge management core outputs a high-frequency dense component in step three. Then, first read the handshake phase fields, current request allowable range, minimum dwell time field, message sequence number, and loop check status from the charging interface controller; simultaneously, read the initial DC bus voltage from the photovoltaic inverter. DC bus capacitor Calibration value, overvoltage threshold And the inverter derating enable state. Subsequently, the protocol compliant wave packet shaper shapes the high-frequency compact components within a shaping window. The system performs area equivalent integration to generate step current segments that meet the vehicle-side update cycle, and writes them one by one into the current request field according to the transmission cycle. If there is still residual energy that has not been transmitted in time within this window, this part is not queued in the software, but is temporarily retained in the DC bus capacitor in the form of a DC bus voltage rise. The edge management core then determines whether to continue shaping, reduce the step size, or directly trigger the photovoltaic inverter to derating and curtail the light based on the new bus voltage state.

[0091] The protocol-compliant wave packet shaper takes only high-frequency compact components as input. Its output object is a set of step pairs ordered by the sending sequence. .

[0092] Among them, current amplitude The dwell time is limited by the allowable current window corresponding to the current vehicle handshake phase. It is limited by the minimum dwell time field and the message transmission cycle. After entering the shaping window, the edge management core first sorts out the high-frequency dense components. The energy within this window is summed, and then the allowable current step size for this window is calculated based on the estimated current voltage received by the vehicle. The shaper follows the principle of area equivalence, meaning that within the same shaper window, the total energy delivered to the vehicle by the stepped wave packet is close to the high-frequency dense component. The corresponding total target energy: ; Among them, the number of steps The number of steps generated within the current shaping window determines the number of discrete segments; its value ranges from a set of positive integers. Step current. : No. The current amplitude of each step segment is directly written to the current request field, and its value range is limited by the vehicle's current allowed current window; dwell time. : No. The duration of each step segment is maintained to ensure that the on-board battery management system can completely sample the current step. Its value range is limited by both the minimum dwell time and the transmission cycle. The vehicle-side voltage... The vehicle's receiving voltage value within the shaping window is used for power conversion; its function is to convert high-frequency dense components... The corresponding energy is converted into the current step area, and the range of values ​​is determined by the feedback or estimation results from the charging interface controller; window start point : No. The starting time of each shaping window defines the starting position of the current integration interval; the width of the shaping window... The duration of a single shaping operation defines the time boundary for the equivalent area calculation, and its value is determined by the message tick and the edge management core scheduling cycle; high-frequency dense components. The continuous power components to be reconstructed within the shaping window serve to provide the target energy source for this window, and their value range is limited by the output boundary of step three. Preferably, the shaping window width Align with the charging interface controller's one-round message update cycle; when the vehicle-side handshake field declares a minimum current step size of 1A and a minimum dwell time of 100ms, the edge management core uses this threshold as the minimum step granularity within the current window; when the vehicle-side declares a wider step size or dwell boundary, it shapes according to the declared boundary.

[0093] When in use, high-frequency dense components The object is rewritten into a discrete object that can be recognized and accepted by the vehicle side; the principle of area equivalence ensures that the original target energy is retained after rewriting; after the shaping window is aligned with the message beat, the action chain between the edge management core, the charging interface controller and the vehicle battery management system remains consistent.

[0094] Protocol-compliant wave packet shaping introduces time hysteresis due to the high-frequency compact components. The original intention was to send the packet out in a shorter time, but the stepped wave packet must be released segment by segment according to the dwell time.

[0095] Step four avoids accumulating this delayed energy in the software buffer queue for an extended period; instead, it temporarily transfers it to the energy storage path of the DC bus capacitor.

[0096] The edge management core calculates the remaining energy of each shaping window at the end of that window. Based on this, the new DC bus voltage can be estimated: ; Among them, the remaining energy : No. The portion of energy within the shaped window that has not yet been delivered to the vehicle via the stepped wave packet serves to measure the short-time voltage accumulation that needs to be handled by the DC bus capacitor; its value range is within the non-negative energy range. Vehicle-side voltage received. : No. The vehicle receiving voltage corresponding to each step segment is used to calculate the actual electrical energy delivered to the vehicle in this segment. The value range is determined by the charging interface controller's feedback or estimation results, obtained from the voltage field returned by the charging interface controller or the cached value from the edge management core based on the most recent valid message; step current. Indicates the first The current amplitude of each step segment is used to write the current request field at the vehicle end; it is obtained by generating the protocol compliance wave packet shaper.

[0097] Duration of stay on the steps Indicates the first The duration of each step segment is used to meet the sampling stability requirements of the vehicle battery management system; it is obtained by determining the minimum dwell time and transmission cycle. Other symbols... , , , The meaning remains consistent with the above; The edge management core is then combined with the DC bus capacitor. and DC bus initial voltage Estimate the voltage rise after the remaining energy enters the DC bus capacitor: ; Among them, the bus voltage after the rise : No. The estimated voltage after the DC bus capacitor absorbs the remaining energy after each shaping window ends serves as the basis for determining whether to continue shaping. Its value must not be lower than the initial DC bus voltage. The voltage range; its acquisition method is the voltage sampling link feedback value of the photovoltaic inverter; DC bus initial voltage The DC bus voltage measured before the start of the current shaping window serves as the voltage reference for this window, and its value range is determined by the inverter voltage detection link. DC bus capacitor The equivalent capacitance used to absorb short-term energy accumulation; its function is to convert excess energy... It is converted into a calculable voltage rise, the range of which is determined by the inverter model calibration value or the external buffer unit calibration value; it is obtained from the inverter model parameter table or the calibration value of the external buffer capacitor unit.

[0098] Preferably, the DC bus capacitor DC bus capacitor bank from inside the photovoltaic inverter; in parallel implementation, DC bus capacitor It can also be provided by an independent buffer capacitor unit connected in parallel to the DC bus.

[0099] In use, the time difference caused by protocol rewriting is transferred to the hardware capacitor path instead of the software blocking path; the edge management core directly calculates the current compensation margin based on the voltage rise; the photovoltaic inverter and the charging interface controller participate in the same closed loop.

[0100] When the DC bus voltage rises too much, the protocol compliance packet shaping can no longer continue at its original pace; otherwise, short-term voltage buildup will translate into hardware overvoltage risk, requiring the calculation of the overvoltage margin ratio. : ; Among them, the overpressure margin ratio : No. The relative margin between the current bus voltage and the overvoltage threshold after the end of each shaping window serves as a unified criterion for continuing, narrowing, or exiting shaping, and its value range is a real number interval less than or equal to 1. Overvoltage threshold The overvoltage protection red line of the inverter hardware defines the absolute upper limit of the allowable rise in DC bus voltage. Its value range is given by the inverter protection parameters; the bus voltage after rise... The meaning remains consistent with the above; when the overpressure margin ratio When the value falls below a preset exit threshold, the edge management core stops integrating new high-frequency dense components. And issued a curtailment instruction to the photovoltaic inverters.

[0101] Preferably, when the overpressure margin ratio When within the safe range, the edge management core continues to send subsequent stepped wave packets according to the current window cycle; when the overvoltage margin ratio Upon entering the compression zone, the edge management core reduces the step current of the next window. Or increase the number of steps. When the overpressure margin is greater than When approaching the exit range, the edge management core immediately stops integrating the new high-frequency dense component. It also issues a derating and curtailment instruction to the photovoltaic inverters while maintaining the low-frequency smoothing components that have already been transmitted. constant.

[0102] In the preferred implementation, the edge management core, charging interface controller, and photovoltaic inverter share a single window time base; in a parallel implementation, if the system is configured with dual-gun charging interfaces, the edge management core generates a set of step pairs for each charging interface. The remaining energy from multiple interfaces that were not sent out is merged and mapped to the same DC bus capacitor absorption path; if a charging interface returns to a rejection state, the step queue corresponding to that interface is immediately cleared, and the remaining high-frequency dense components are... The process directly switches to the inverter derating and curtailment path.

[0103] When in use, the shaping process has a clear stopping boundary and will not push energy to the DC bus indefinitely; the derating curtailment path is not confused with the vehicle power receiving path and the stationary energy storage path; dual-gun and rejection scenarios are both included in the equivalent implementation methods under the same terminology system.

[0104] This second embodiment provides a smart meter that is not simply a metering terminal. Instead, while maintaining the independence of the legal metering link, it introduces the field status of the photovoltaic side, the fixed energy storage side, and the electric vehicle charging side into the same in-meter control closed loop, so that the aforementioned steps one to four are all completed by the same smart meter body.

[0105] The smart meter in this embodiment includes a meter casing, a metering terminal block, a sampling front-end board, a dual-core main control board, an isolated power supply board, a communication interface board, and a parameter memory. The inside of the meter casing is divided into a metering chamber and a control chamber by an insulating partition. The sampling front-end board and the legal metering core are arranged in the metering chamber, while the edge management core, local Flash memory, communication interface board, and an isolated transceiver for connecting to external devices are arranged in the control chamber. This compartmentalized structure aims to spatially separate the legal metering link from the control link, while shortening the analog signal path from the sampling front-end to the legal metering core, reducing high-frequency interference intrusion into the meter. The metering terminal block is preferably a three-phase four-wire connection structure, connecting to the voltage sampling circuit and current sampling circuit of the grid connection point respectively. As a parallel solution, a single-phase connection structure can also be used, as long as the legal metering core can still output the digital sampling results corresponding to the original high-frequency AC analog waveform.

[0106] The sampling front-end board is equipped with a voltage divider network, a current transformer or shunt sampler, an analog conditioning circuit, and an analog-to-digital converter (ADC). The output of the ADC is directly connected to the legal metrology core. The legal metrology core internally includes at least a raw sampling buffer, a finite impulse response digital low-pass filter link, a unidirectional DMA channel control register, a cross-core shared read-only FIFO write entry, a hardware timer, and an unmasked hardware interrupt output pin. The edge management core is connected to at least the cross-core shared read-only FIFO read exit, local Flash memory, dynamic mask filter coefficient area, protocol compliance wave packet shaper execution area, fixed energy storage communication port, charging interface communication port, and photovoltaic inverter communication port.

[0107] In terms of structural connections, the legal metrology core and the edge management core do not share writable registers or program storage space. They communicate solely through a cross-core shared read-only FIFO driven by a unidirectional DMA channel, transmitting mirrored data and trigger events via a dedicated unshielded hardware interrupt line. The cross-core shared read-only FIFO is preferably implemented using a dual-port memory array with write protection gating, with its write end fixedly connected to the legal metrology core and its read end fixedly connected to the edge management core. As a parallel alternative, the cross-core shared read-only FIFO can also be implemented using a latch array with address incrementing logic, as long as the edge management core cannot write data back to the legal metrology core. This structural arrangement corresponds to the twin residual eavesdropping and immune triggering of the underlying physical isolation domain in step one.

[0108] The specific operation process is as follows: When waveform edge disturbances occur at the grid connection point due to sudden changes in photovoltaic inverter output, commutation of fixed energy storage, or changes in vehicle power supply, the sampling front-end board first sends the AC voltage and current waveforms to the analog-to-digital converter. After receiving the sampling results, the legal metrology core sends one path along the original metering link to the finite impulse response digital low-pass filter link to form a compliant smooth fundamental waveform; the other path retains the high-frequency raw code stream. The unidirectional DMA channel writes the high-frequency raw code stream and the compliant smooth fundamental waveform into the cross-core shared read-only FIFO at the same sampling point. The legal metrology core internally calculates the difference between the two to obtain the twin residuals and further calculates the residual slope. When the residual slope exceeds the hardware safety threshold, the legal metrology core sends a wake-up signal to the edge management core through a dedicated unshielded hardware interrupt pin, and simultaneously starts a hardware timer to open the interrupt aggregation blanking window. If repeated triggering caused by electromagnetic interference continues to occur within the interrupt aggregation blanking window, the hardware logic gate maintains a single wake-up and suppresses repeated pull-ups; if the lock continues to time out, it switches to the software serial peripheral interface polling path. Therefore, the legal metrology core maintains its legal metrology independence, while the edge management core obtains a pure cross-core wake-up signal.

[0109] As a supplement: Hardware security threshold The parameters are determined during the factory calibration stage based on the sampling front-end range, installation scenario, and typical switch disturbance envelope, and written into the one-time configuration area or the edge management core read-only parameter area; interrupt aggregation blanking window. The hardware timer configuration value is given by the legal metrology core, and its duration covers one or more inverter switching cycles and at least one charging message update cycle. Among these, the hardware safety threshold... Its purpose is to distinguish between background noise and effective disturbances; it is obtained through factory calibration. Interruption aggregation blanking window. Its purpose is to suppress storm lock-up caused by repeated interrupts; it is obtained by setting hardware timer parameters.

[0110] After the edge management core is awakened, it first suspends the local billing reporting task and the heartbeat maintenance task. Then, it reads the mirror data pair from the cross-cell shared read-only FIFO, and reads the mirror values ​​of the grid connection point net power, fixed energy storage health status, fixed energy storage adjustable margin, and charging interface acceptance level from the local buffer. In this embodiment, the above statuses are collected by the edge management core through the communication interface board: the fixed energy storage side preferably connects to the fixed energy storage management unit via the CAN bus to read the cell body temperature. State of charge The system reads the handshake phase fields, allowable current boundaries, and minimum dwell time from the charging side via a communication transceiver conforming to GB / T27930; from the photovoltaic inverter side, it reads the current DC bus voltage and derating status via an RS485 or Ethernet industrial interface. The edge management core encodes the above objects into a microgrid status vector and performs offset addressing in the SA-SOH sparse tensor tree pre-installed in local Flash memory to directly extract the anti-reverse current target power regulation command. The local Flash memory preferably uses a dual-partition layout, with one partition for execution and one partition for updating, to prevent the tensor tree from overwriting the currently executing lookup table during upgrades.

[0111] In this embodiment, the dynamic mask filter is embedded in the control program of the edge management core, or it can be implemented by a fixed-point digital signal processor external to the edge management core. The edge management core reads the temperature of the battery cell body. and state of charge Then, the diffusion parameters, particle characteristic length parameters, and thermodynamic mapping parameters corresponding to the current fixed energy storage model are retrieved from the parameter memory to calculate the digital cutoff frequency. Then use that number as the cutoff frequency General command for power regulation of anti-reverse flow target Frequency domain tearing is performed to obtain low-frequency smooth components. With high-frequency dense components Among them, the low-frequency smoothing component Write to the fixed energy storage power setpoint register via the fixed energy storage communication port; high-frequency dense component This information is then written into the protocol compliance wave packet shaper execution region. If the stationary energy storage is in a low-temperature, low-diffusion state, the edge management core switches the dynamic mask filter to a protective state, making the high-frequency dense components... It does not enter the fixed energy storage execution chain.

[0112] The protocol-compliant wave packet shaper is not a standalone hardware chip, but rather a piece of shaping execution logic coupled to the charging interface communication stack within the edge management core. Its input is a high-frequency dense component. The output is a time-division multiplexing step current request sequence. The edge management core, based on the current handshake phase, current allowable range, and minimum dwell time, assigns high-frequency dense components... The data is rewritten into multiple low-frequency step segments and written to the charging interface controller according to the message rhythm. This way, the onboard battery management system receives a power request that meets the requirements of the current stage, rather than a sudden, instantaneous power jump. Simultaneously, the photovoltaic inverter communication port continuously transmits back the DC bus voltage and overvoltage threshold. When the temporarily unreleased energy caused by step shaping pushes up the DC bus voltage, the edge management core does not block subsequent scheduling in the software queue, but allows the DC bus capacitor to absorb the energy for a short time. If the DC bus voltage approaches the overvoltage threshold... The edge management core immediately sends a derating and curtailment command to the photovoltaic inverter and suspends new high-frequency dense component solar power. The integral input completes the physical entity compensation loop in step four.

[0113] To facilitate understanding, consider this specific scenario: The smart meter is installed at the public access point of a photovoltaic-storage-charging integrated cabinet in the park. During the day, the rooftop photovoltaic power suddenly increases, an electric vehicle is connected to the DC charging port, and the fixed energy storage is in a medium-to-high charge state. At this time, the legal metering core first captures the steepening of the current edge from the sampling front-end board, and simultaneously delivers the high-frequency raw code stream and the compliant smoothed fundamental frequency to the edge management core through a unidirectional DMA channel and a cross-chip shared read-only FIFO; the edge management core then retrieves the anti-reverse current target power adjustment command from its local Flash memory. Then, a portion of it was written as a low-frequency smoothing component. The other part is sent to stationary energy storage, and written as a high-frequency dense component. The protocol-compliant wave packet shaper is then used. The vehicle side sees a smooth step request, while the photovoltaic inverter side handles the bus voltage buffering during the step transition. The entire process is completed within the same smart meter.

[0114] The smart meter in this embodiment can also be equipped with a local maintenance interface, an event log storage area, and a parameter upgrade interface. The parameter upgrade interface only allows writing to the operating parameters, SA-SOH sparse tensor tree update package, and dynamic mask filter parameter table on the edge management core side; it does not allow rewriting of the metering parameters and filter coefficients in the legal metering core. As a parallel extension, the smart meter can also be connected to an external independent buffer capacitor unit, extending the DC bus capacitor absorption path from inside the inverter to a dedicated external buffer branch; the fixed energy storage communication port can also be replaced from a CAN bus to an Ethernet industrial protocol interface, as long as the edge management core can still read the cell body temperature. State of charge You can use the health status field.

[0115] In the second embodiment, the smart meter implements the above four methods and steps into a fabricable, connectable, and operable internal physical structure through the dual-core legal metrology core-edge management core, unidirectional DMA, cross-core shared read-only FIFO, SA-SOH sparse tensor tree, dynamic mask filter, protocol compliant wave packet shaper, and DC bus voltage compensation chain.

[0116] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0117] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart meter energy management method for photovoltaic, energy storage, and charging scenarios, characterized by: include, The legal metrology core mirrors the high-frequency raw code stream acquired by the ADC and the compliant smoothed fundamental wave processed by the legal FIR digital low-pass filter to the cross-core shared read-only FIFO through a one-way DMA channel, and outputs a pure cross-core wake-up signal based on the difference between the two. The edge management core responds to the pure cross-core wake-up signal, offsets and addresses in the SA-SOH sparse tensor tree in the local Flash to obtain the anti-reverse current target power adjustment general command; the anti-reverse current target power adjustment general command is input into the dynamic mask filter, and combined with the cell temperature and state of charge, low-frequency smooth component and high-frequency dense component are obtained. The low-frequency smooth component is sent to the fixed energy storage, and the high-frequency dense component is reconstructed by the protocol compliant wave packet shaper and sent to the electric vehicle. Compensation and exit control are performed in conjunction with the DC bus capacitor voltage.

2. The smart meter energy management method for photovoltaic, energy storage, and charging scenarios according to claim 1, characterized in that: The legal metrology core simultaneously mirrors and writes the high-frequency raw code stream and the compliant smooth fundamental wave corresponding to the same sampling point into the cross-core shared read-only FIFO via a one-way DMA channel. Based on the difference between the two, a twin residual is formed. When the twin residual meets the trigger condition, a pure cross-core wake-up signal is output to the edge management core and the interrupt aggregation blanking window is started.

3. The smart meter energy management method for photovoltaic-storage-charging scenarios according to claim 2, characterized in that: The legal metrology core determines the residual slope based on the twin residuals of adjacent sampling points, compares the residual slope with the hardware safety threshold, and switches to software SPI polling when the interrupt aggregation blanking window is locked for a timeout, in order to maintain the subsequent acquisition path of the pure cross-chip wake-up signal.

4. The smart meter energy management method for photovoltaic, energy storage, and charging scenarios according to claim 3, characterized in that: The edge management core responds to the pure cross-core wake-up signal, freezes the net power mirror value of the grid connection point, the fixed energy storage health status mirror value, the fixed energy storage adjustable margin mirror value, and the charging interface acceptance level mirror value, and encodes each frozen mirror value into a microgrid state vector for the transformer area, which serves as the addressing input for the SA-SOH sparse tensor tree.

5. The smart meter energy management method for photovoltaic-storage-charging scenarios according to claim 4, characterized in that: The edge management core performs trunk addressing of the SA-SOH sparse tensor tree according to the scene index, disturbance index, and health index, and locates the sub-tables under the healthy leaf nodes by combining the margin index and the acceptance index. It prioritizes reading the leaf value code, and reads the backoff leaf value code when there is no corresponding leaf value code, so as to obtain the general command for anti-reverse flow target power adjustment.

6. The smart meter energy management method for photovoltaic, energy storage, and charging scenarios according to claim 5, characterized in that: The edge management core reads the cell body temperature, state of charge, and cell diffusion calibration parameters in the local parameter area to determine the digital cutoff frequency of the dynamic mask filter. Based on this, it filters and splits the anti-reverse current target power adjustment command to obtain a low-frequency smooth component and a high-frequency dense component.

7. The smart meter energy management method for photovoltaic-storage-charging scenarios according to claim 6, characterized in that: The edge management core writes the low-frequency smooth component into the fixed energy storage power setpoint register, writes the high-frequency dense component into the protocol compliance wave packet shaper execution area, and synchronously writes the current digital cutoff frequency into the filter state area. It also restricts the alternating command from entering the fixed energy storage execution chain when the cell diffusion capability is lower than the preset boundary.

8. The smart meter energy management method for photovoltaic, energy storage, and charging scenarios according to claim 7, characterized in that: The edge management core reads the handshake phase field, current allowable range, minimum dwell time and corresponding message sequence number returned by the charging interface controller, reconstructs the high-frequency dense components into a stepped current request sequence ordered by the current transmission cycle, and writes it into the current request field according to the message cycle.

9. The smart meter energy management method for photovoltaic, energy storage, and charging scenarios according to claim 8, characterized in that: Within the current shaping window, the edge management core determines the remaining energy based on the target energy corresponding to the high-frequency dense component and the actual delivered energy corresponding to the step current request sequence, and maps the remaining energy to the DC bus capacitor voltage rise to update subsequent shaping cycles or step currents; When the DC bus voltage corresponding to the DC bus capacitor voltage rise reaches the overvoltage exit condition, the edge management core stops integrating new high-frequency dense components, maintains the transmission path of low-frequency smooth components, clears the unsent step current request sequence in the current shaping window, and issues a derating curtailment command to the photovoltaic inverter.

10. A smart meter for photovoltaic-storage-charging scenarios, comprising a legal metering core, an edge management core, a unidirectional DMA channel, a cross-chip shared read-only FIFO, local Flash, a dynamic mask filter, and a protocol-compliant wave packet shaper, characterized in that: The legal metrology core mirrors the high-frequency raw code stream acquired by the ADC and the compliant smoothed fundamental wave processed by the legal FIR digital low-pass filter to the cross-core shared read-only FIFO via a one-way DMA channel, and outputs a pure cross-core wake-up signal to the edge management core based on the difference between the two. The edge management core obtains the anti-reverse current target power adjustment general command by offset addressing in the SA-SOH sparse tensor tree in the local Flash. After passing through the dynamic mask filter, it obtains the low-frequency smooth component and the high-frequency dense component. The low-frequency smooth component is sent to the fixed energy storage, and the high-frequency dense component is reconstructed by the protocol compliant wave packet shaper and sent to the electric vehicle. It also performs compensation and exit control in combination with the DC bus capacitor voltage.

Images