Park carbon neutralization intelligent micro-grid system

Abstract

The invention discloses a park carbon neutralization intelligent micro-grid system, and relates to the technical field of power grid control. The system comprises a master control data bus and a plurality of energy nodes, establishing a unique energy mapping identifier for each energy node on the master control data bus; collecting the power, the energy storage state and the carbon emission value of the energy node in real time through the energy mapping identifier, and uploading the power, the energy storage state and the carbon emission value to the master control data bus; the master control data bus converges all the collected values to form a park energy mapping integrated data set; generating a first-layer instantaneous carbon shadow sub-account book and a second-layer recursive carbon shadow sub-account book of the energy nodes; and comparing with respective corresponding upper limit values, and generating a power control instruction. The real-time performance of energy dispatching, the energy storage utilization rate and the power grid operation stability are improved, and the comprehensive regulation and control capability of the intelligent micro-grid system is obviously superior to that of an existing intelligent micro-grid system.

Description

Technical Field

[0001] This invention relates to the field of power grid control technology, and in particular to a carbon-neutral smart microgrid system for industrial parks. Background Technology

[0002] In the current field of energy management and carbon emission control, as various parks, industrial zones, and research clusters increasingly rely on renewable energy, the application scenarios of smart microgrids are gradually expanding, and they are increasingly undertaking the comprehensive tasks of regional energy supply, energy storage management, and carbon emission coordination. In existing technologies, most smart microgrid systems are centered on a distributed energy dispatch platform, combined with an energy management system (EMS) and distributed controllers to achieve power allocation and energy storage dispatch at energy nodes. In these systems, equipment such as photovoltaic arrays, wind power generation, energy storage batteries, and gas-fired combined cooling, heating, and power (CCHP) units collect power, energy storage status, and load conditions in real time through monitoring nodes and upload this data to a host computer platform, which then generates power adjustment commands based on economic dispatch or demand response strategies. However, such systems typically treat carbon emission data as an independent indicator, using it only for periodic statistics or auxiliary optimization, failing to couple carbon emission intensity into the dispatch logic in real time. Therefore, they have significant shortcomings in dynamic dispatch under carbon neutrality constraints.

[0003] Publicly available smart microgrid technologies mainly fall into two categories: one is energy dispatching schemes based on static optimization models, which calculate power allocation and energy storage charging / discharging strategies for a given dispatching cycle by establishing multi-objective linear or nonlinear optimization models and setting load demand, equipment efficiency, and some emission factors as constraints; the other is dynamic dispatching schemes based on prediction and rolling optimization, which achieve dynamic power allocation and indirect carbon emission constraints by predicting future load and renewable energy output and using model predictive control (MPC) methods. While these two schemes can meet the energy economy and power balance requirements of most park-level applications, they have significant shortcomings in real-time carbon emission control and peak suppression. Specifically, static optimization schemes lack real-time feedback capabilities. When renewable energy output fluctuates drastically or load surges, the system cannot promptly synchronize carbon emission increases with power adjustments, easily leading to instantaneous carbon emission intensity exceeding limits. While dynamic prediction schemes can adjust energy storage strategies in advance to some extent, their control strategies are highly dependent on prediction accuracy. If the prediction error is large, control commands will deviate, causing frequent start-ups and shutdowns of energy storage devices, or even overcharging and discharging, increasing system losses and posing safety hazards. Summary of the Invention

[0004] In view of this, the present invention provides a carbon-neutral smart microgrid system for industrial parks. This system collects and aggregates power load, energy storage charge status, and carbon emission increments of energy nodes in real time via a main control data bus. It achieves millisecond-level data freezing and synchronization using a two-layer carbon shadow ledger structure. It dynamically labels high-emission, low-emission, and neutral-emission clusters using voxel-based clustering, and uses a sixteen-phase interleaved recursive scanning path constructed by reverse-order odd-even segmentation, cross-step switching, and oblique backtracking to discretely control the sequence in time and space. This coordinates the linkage adjustment of chain-like power and carbon weights. Through virtual carbon release valves, reverse carbon weight compensation, and power compensation pulses, it achieves dynamic balance of carbon weights between nodes. Thus, while ensuring that the overall carbon emission intensity of the industrial park remains stable within a preset threshold, it improves the real-time performance of energy dispatch, energy storage utilization, and grid operation stability, significantly outperforming the comprehensive control capabilities of existing smart microgrid systems.

[0005] The technical solution adopted in this invention is as follows:

[0006] A carbon-neutral smart microgrid system for industrial parks includes: a main control data bus and several energy nodes; a unique energy mapping identifier is established for each energy node on the main control data bus; the power, energy storage status, and carbon emission values ​​of the energy node are collected in real time through the energy mapping identifier and uploaded to the main control data bus; the main control data bus aggregates all collected values ​​to form an integrated energy mapping dataset for the industrial park; based on the dataset, a dynamic bidirectional carbon weight self-balancing control algorithm is invoked to perform multi-round chain-like power and carbon weight linkage adjustments on all energy nodes in the industrial park to ensure that the overall carbon emission intensity of the industrial park remains within a preset carbon emission threshold, generating a first-level instantaneous carbon shadow ledger and a second-level recursive carbon shadow ledger for the energy nodes; the main control data bus reads the real-time power load, energy storage status, and carbon emission increment of each energy node according to the second-level recursive carbon shadow ledger within a fixed period, compares them with their respective upper limit values, and generates power control commands.

[0007] Furthermore, based on the dataset, the main control data line invokes a dynamic bidirectional carbon weight self-balancing control algorithm to perform multi-round chain-like power and carbon weight linkage adjustments on all energy nodes in the park. Specifically, this process includes: simultaneously configuring a first-layer instantaneous carbon shadow ledger and a second-layer recursive carbon shadow ledger for each energy mapping identifier, and achieving seamless switching between the two ledgers in milliseconds through a ring-shaped double-buffering technology; performing voxel-based clustering of all energy nodes based on real-time power load, energy storage state of charge, and carbon emission increment three-dimensional coordinates, and accordingly labeling high-emission clusters, low-emission clusters, and neutral-emission clusters; and for each… A cluster is formed by combining reverse-order odd-even segmentation, cross-step rotation, and oblique backtracking to create a sixteen-phase interleaved recursive scanning path, avoiding instantaneous spikes caused by batch adjustments; time-series double interpolation is performed on all neutral emission nodes, injecting repair power compensation pulses and binding unique traceability tags; virtual carbon slow-release valves are set up at high emission nodes to form gradient reduction curves, and power downgrading is triggered immediately when the carbon emission curve touches the baseline of the reduction valve; the surplus energy generated after power is increased at low emission nodes is preferentially backfilled into their energy storage units, and if the state of charge approaches the safe upper limit, it is switched to the external grid backfeed channel.

[0008] Furthermore, after the main control data bus completes the synchronous broadcast of instructions to all energy nodes, it moves the first-level instantaneous carbon shadow ledger up to cover the second-level recursive carbon shadow ledger.

[0009] Furthermore, the on-chip high-speed storage area of ​​the main control data bus is divided into contiguous storage segments for the first-level instantaneous carbon shadow ledger and the second-level recursive carbon shadow ledger, with both segments having equal capacity. To avoid address jitter, the starting addresses of both storage segments are aligned with the address boundaries of the main control data bus. At system startup, write and read pointers are set for the first-level instantaneous carbon shadow ledger, with both initially pointing to the first address of the ledger's storage segment. Independent write and read pointers are set for the second-level recursive carbon shadow ledger, with their initial values ​​also pointing to the first address of the ledger's storage segment. After receiving the latest data from a certain energy node, the main control data bus locates the corresponding record position in the first-level instantaneous carbon shadow ledger according to the energy mapping identifier.

[0010] Furthermore, whenever the master control data bus detects the arrival of a millisecond time slice boundary, it triggers a data freeze operation: First, the write pointer position of the first-level instantaneous carbon shadow ledger is written to the circular double-buffered control table and marked as the freeze boundary; then, the read pointer and write pointer are swapped atomically, so that the read pointer points to the latest freeze boundary and the write pointer points to the position after the previous round of freeze boundary; at this time, the data consumer only reads the frozen complete data segment from the start of the read pointer to the position before the write pointer, ensuring that the read content remains unchanged; after the data reading is completed, the master control data bus copies the frozen data segment of the first-level instantaneous carbon shadow ledger to the record position of the same energy mapping identifier in the second-level recursive carbon shadow ledger; the copying action is completed in one go using a block-level direct access method, avoiding the latency caused by field-by-field write operations.

[0011] Furthermore, the main control data bus reads the real-time power load, energy storage state of charge, and carbon emission increment corresponding to each energy mapping identifier from the dataset one by one, and assembles them into three-dimensional coordinate points according to the field order of real-time power load, energy storage state of charge, and carbon emission increment. Based on the overall real-time power load range, energy storage state of charge range, and carbon emission increment range of the park, the main control data bus divides each of the three coordinate axes into equally spaced planes. The planes of the three coordinate axes are pairwise orthogonal, forming a three-dimensional voxel grid. Each voxel unit corresponds to a unique grid number in three-dimensional space, and the grid number is generated by incrementing the coordinate axis sequence. The main control data bus traverses all energy nodes and sets their three-dimensional coordinates... Points are mapped to corresponding voxel units. If multiple energy nodes exist within a voxel unit, they are added sequentially according to the mapping order to form a node list. If the coordinates of an energy node are located on the voxel boundary surface, the endpoint offset rule towards zero scale is used to classify it into an adjacent voxel unit close to zero scale. The main control data bus calculates the number of nodes in each voxel unit. If the number of nodes is lower than the preset density lower limit, the voxel unit is marked as a sparse voxel and temporarily not included in the cluster determination. The remaining voxel units are retained as candidate voxel units. For each candidate voxel unit, the main control data bus counts the average real-time power load, average energy storage state of charge, and average carbon emission increment in the node list, and records them as a voxel mean set.

[0012] Furthermore, the main control data bus reads the real-time average power load, average energy storage state of charge, and average carbon emission increment of each energy node from the voxel average set at fixed intervals, and compares them one by one with their respective preset power upper limit, energy storage safety lower limit, and carbon emission safety threshold: if the real-time average power load is higher than the power upper limit, a power reduction command is issued to the corresponding energy node; if the average energy storage state of charge is lower than the energy storage safety lower limit, a stop discharge command is issued to the corresponding energy node and charging is allowed; if the average carbon emission increment is higher than the carbon emission safety threshold, a power limiting command is issued to the corresponding energy node; the above commands are simultaneously broadcast to all energy nodes.

[0013] Furthermore, the process of the main control data line segmenting each cluster in reverse order and alternating between odd and even numbers and performing cross-step rotation includes: the main control data bus first reads the current cluster node list and assigns a sequential index to the list according to the order of the energy mapping identifier; the sequential index is divided into odd and even segments according to the odd sequence and even sequence; the odd segments are rearranged in descending order of the sequential index to obtain reverse odd segments; the even segments are arranged in ascending order of the sequential index to obtain sequential even segments; the reverse odd segments and sequential even segments are concatenated to generate the initial segment sequence, laying the sequential foundation for subsequent scanning; the main control data bus assigns a rotation index to the initial segment sequence and performs segment traversal from the first index with a fixed step length of two; after each traversal, the step starting point is shifted one position to the right to form a cross-step rotation sequence; after the cross-step rotation sequence is cycled eight times, the initial segment sequence is rearranged into a cross-step rotation sequence, and the node access order is fully discretized.

[0014] By adopting the above technical solutions, this invention achieves the following beneficial effects: The proposed carbon-neutral smart microgrid system for industrial parks establishes a unique energy mapping identifier for each energy node on the main control data bus. A two-layer ledger structure, consisting of a first-layer instantaneous carbon shadow ledger and a second-layer recursive carbon shadow ledger, supports millisecond-level data freezing, replication, and switching, enabling seamless connection of data streams for acquisition, storage, and control, effectively avoiding data tearing problems in high-frequency scheduling. Simultaneously, the system projects all nodes onto a three-dimensional voxel grid by real-time acquisition of power load, energy storage state of charge, and carbon emission increments. Based on voxel density and mean results, it dynamically divides nodes into high-emission, low-emission, and neutral emission clusters, achieving rapid spatial grouping and emission level identification between nodes. Combined with a sixteen-phase interwoven recursive scanning path constructed using reverse-order odd-even segmentation, cross-step rotation, and oblique backtracking chains, the system synchronizes the control sequence of discrete nodes in time and space, avoiding instantaneous spikes and frequency fluctuations caused by simultaneous adjustments of multiple nodes. Simultaneously, high-emission nodes achieve smooth power reduction under the action of virtual carbon slow-release valves, low-emission nodes can increase output and prioritize the backfilling of energy storage units through reverse carbon weight compensation, and neutral-emission nodes obtain power compensation pulses through time-series double interpolation. Overall, a dynamic carbon weight balance and energy utilization maximization are formed among high- and low-carbon nodes. Through the above closed-loop process, this system effectively improves the real-time performance of energy dispatch, the utilization efficiency of energy storage devices, and the stability of grid operation while ensuring that the overall carbon emission intensity of the park remains within the preset threshold. This is significantly superior to the distributed energy and carbon emission management capabilities of existing smart microgrids. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of a carbon-neutral smart microgrid system for a park, as described in an embodiment of the present invention. Detailed Implementation

[0016] All features disclosed in this specification, or all steps in all disclosed methods or processes, may be combined in any way, except for mutually exclusive features and / or steps.

[0017] Any feature disclosed in this specification (including any appended claims and abstract) may be replaced by other equivalent or similar features, unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is merely one example of a series of equivalent or similar features.

[0018] refer to Figure 1 The key to implementing the carbon-neutral smart microgrid system in the industrial park lies in mapping energy flow, information flow, and carbon emission flow onto a unified spatiotemporal coordinate system maintained by the main control data bus. This allows for the construction of a finely addressed, real-time closed-loop dynamic carbon emission mapping model at the digital level. The core premise is that the power output, energy storage state of charge, and carbon emission increment of any energy node are not isolated factors but collectively determine its instantaneous gradient value on the overall carbon emission intensity surface of the industrial park. Therefore, the system first establishes a unique energy mapping identifier for each energy node, reorganizing the previously scattered metering instrument signals into a unified message structure before writing them into the integrated energy mapping dataset of the industrial park. This process not only eliminates node naming conflicts and timestamp drift problems in traditional distributed monitoring but also lays a highly consistent time base for subsequent high-frequency regulation. Subsequently, the dynamic bidirectional carbon weight self-balancing control algorithm was designed as a discrete event-driven controller that directly acts on the gradient of the three-dimensional coordinate surface. It uses the overall carbon emission threshold of the park as a rigid boundary condition and couples power scheduling and carbon emission constraints into two sides of the same objective function: when a positive deviation of the local gradient is detected, a negative carbon flow quota is triggered and the power of the corresponding node is deducted. At the same time, the scheduling command is delayed through a virtual carbon slow-release valve to prevent the instantaneous power reduction from causing an impact on the electrical side. When the gradient deviates in the opposite direction, the power of low-emission nodes is increased through a reverse carbon weight compensation value or their energy storage is allowed to be backfilled, thus forming a real-time balancing mechanism that "borrows" carbon weight from the high-carbon area to the low-carbon area.

[0019] The algorithm employs a circular double-buffering technique, consisting of a first-layer instantaneous carbon shadow ledger and a second-layer recursive carbon shadow ledger, because under millisecond-level control cycles, the write and read threads must not block each other. The double-layer ledger enables continuous acquisition of the latest data by the write end and stable referencing of frozen data by the read end. The write and read pointers are atomically swapped at the millisecond time slice boundary, immediately completing the ledger rotation, thus forming a low-latency, highly consistent carbon emission snapshot channel. To reduce the group resonance effect that occurs when multiple nodes are adjusted simultaneously, the system adopts a voxelized clustering method based on real-time power load, energy storage state of charge, and carbon emission increments. All nodes are projected into discrete voxel units, and the voxel mean is used to represent the comprehensive emission characteristics of the region. This not only compresses the high-dimensional space into a finite voxel grid, reducing the global search time, but also allows subsequent scanning paths to be expanded at the cluster level rather than the single-point level. The purpose of the sixteen-phase interleaved recursive scan path is to further disrupt the execution order of nodes within the cluster, approximating random roaming with a combination of reverse-order odd-even segmentation, cross-step rotation, and oblique backtracking chains, so that nodes that receive power adjustment commands within the same time slice have sufficient distance in both physical space and electrical topology, thereby smoothing out the spikes caused by batch power changes.

[0020] Time-series dual interpolation is used in neutral emission clusters because these nodes have neither significant over-emissions nor significant emission reduction potential, but their power curves and energy storage margins often contain usable residual degrees of freedom. By inserting fine-grained repair power compensation pulses into the original curves and binding unique traceability tags in the ledger, the system can slightly stretch or compress local power without introducing new frequency fluctuations. The main control data bus performs threshold comparisons on the second-level recursive carbon shadow ledger within a fixed period. The power upper limit, energy storage safety lower limit, and carbon emission safety threshold constitute three orthogonal safety boundaries. Triggering any boundary generates a corresponding power control command. This process is similar to a hard real-time task triggered by guard conditions. It does not rely on prediction models but relies entirely on the recursive ledger data accumulated in the previous period to implement rigid correction. After completing one round of regulation, the system immediately moves the frozen fragment of the first-level instantaneous carbon shadow ledger up to cover the second-level recursive carbon shadow ledger, and clears the first-level ledger to prepare for the next round of writing. This process is repeated on a millisecond timescale to achieve a complete closed loop from data acquisition, status assessment, strategy decision-making, instruction issuance to result archiving.

[0021] Thanks to this closed loop, the park's carbon-neutral smart microgrid system can maintain a quasi-linear relationship between computational complexity and node size even as the number of multi-source heterogeneous energy nodes continues to expand. This is because voxel clustering sparsifies the global search matrix, the sixteen-phase interleaved recursive scan path completes instruction discretization in both the time and spatial domains, and the dual-ledger structure provides queue read and write operations with constant time complexity. In summary, this embodiment introduces the concept of carbon weights to replace traditional passive monitoring, coupling power scheduling and carbon emission intensity to the same control scepter, thereby constructing an adaptive, multi-scale, and prediction-independent real-time carbon emission balancing mechanism. Simultaneously, by utilizing the integrated storage and communication architecture of the main control data bus, it establishes a complete data flow from bottom-level metering to top-level decision-making, ultimately achieving continuous and controllable carbon emissions at the park level without sacrificing energy supply security and energy economy.

[0022] Furthermore, in this embodiment, the main control data bus first synchronously opens the first-level instantaneous carbon shadow ledger and the second-level recursive carbon shadow ledger for all energy mapping identifiers based on the park's energy mapping integrated dataset. The two have equal storage capacity and adopt a ring double buffering technology. At the millisecond-level time slice boundary, the ledger role is rotated by atomically swapping the write pointer and the read pointer. This ensures that the write end keeps the latest collected data flowing in continuously, while the read end can stably reference the frozen snapshot, avoiding data tearing caused by write-read conflicts. After completing a pointer switch, the system immediately copies the frozen first-level instantaneous carbon shadow ledger fragment to the corresponding record position in the second-level recursive carbon shadow ledger. The copying latency is shortened by block-level direct access, and the write pointer position is updated synchronously to reserve a blank area for the next second-level cycle. Subsequently, the main control data bus constructs a three-dimensional coordinate point set for three types of fields: real-time power load, energy storage charge status, and carbon emission increment of all energy nodes. Based on the overall field value range of the park, it divides the area along the three coordinate axes at equal intervals to generate a three-dimensional voxel grid. Each voxel unit is assigned a unique grid number. After the system traverses the nodes and completes the mapping, if the number of nodes in a voxel unit is lower than the preset density lower limit, it is marked as a sparse voxel and temporarily does not participate in the cluster determination. The remaining voxels are compared with the average carbon emission increment and the high emission threshold and low emission threshold, and then marked as high emission cluster, low emission cluster, or neutral emission cluster.

[0023] When the algorithm enters the cluster scanning phase, it first performs reverse parity segmentation for each cluster, splitting the node sequential index into reverse odd segments and sequential even segments. Then, it rearranges the access order eight times in two cycles with a fixed step length through cross-step rotation. Finally, it inserts a diagonal backtracking chain in the two-dimensional virtual coordinate plane and divides the sequence into sixteen segments, recursively wrapping them into a closed loop, forming a sixteen-phase interwoven recursive scanning path. This scanning path ensures that the target nodes of any two adjacent control commands maintain sufficient spacing in terms of physical location and electrical topology, thereby weakening the instantaneous spikes caused by batch power adjustments. For nodes in neutral emission clusters, the system performs time-series double interpolation within the power curve, inserts repair power compensation pulses between adjacent sampling points, and writes a unique traceability tag for each pulse, so as to stretch or compress local power output in a fine-grained manner without destroying spectral stability.

[0024] For nodes in high-emission clusters, the system inserts a virtual carbon mitigation valve into their power regulation link. This valve exhibits a gradient decrease curve as power declines. When the carbon emission curve touches the baseline of the decreasing valve, it immediately triggers power downgrading and deducts the corresponding negative carbon flow allowance, ensuring that carbon emission intensity declines rapidly without causing a step shock to the power grid. For nodes in low-emission clusters, the system allows power to be increased and simultaneously increases the reverse carbon weight compensation value. The surplus energy generated by the increase will be preferentially backfilled into the node's own energy storage unit. When the energy storage charge state approaches the safe upper limit, it automatically switches to the external power grid feed-back channel to avoid the risk of overcharging. Throughout the process, the dynamic bidirectional carbon weight self-balancing control algorithm uses a sixteen-phase interleaved recursive scanning path as the access sequence. After each node performs power adjustment or carbon weight settlement, the results are immediately written to the corresponding fields of the first-level instantaneous carbon shadow ledger and batch-sinked to the second-level recursive carbon shadow ledger when the pointer switches. The main control data bus performs hard threshold comparisons based on the real-time power load, energy storage charge status, and carbon emission increment provided by the second-level recursive carbon shadow ledger, and the upper limit of power, the lower limit of energy storage safety, and the carbon emission safety threshold. If an out-of-bounds condition is found, a targeted power control command is generated and simultaneously broadcast to all energy nodes, realizing a millisecond-level closed loop of collection, evaluation, decision-making, execution, and archiving. This ensures that the overall carbon emission intensity of the park remains stable within the preset carbon emission threshold, while maintaining the energy storage safety boundary and improving the renewable energy utilization efficiency of low-emission nodes.

[0025] Furthermore, after the master control data bus broadcasts the power control command to all energy nodes and receives execution confirmation feedback from the node side, the write and read threads enter the ledger rotation preparation state. The system first freezes the write channel of the current first-level instantaneous carbon shadow ledger, locks the write pointer position and records it as the latest frozen boundary. Then, it calls a block-level direct access instruction to copy the entire data segment of the first-level instantaneous carbon shadow ledger from the frozen boundary to the starting address to the starting address of the corresponding storage segment of the second-level recursive carbon shadow ledger. Here, continuous address mapping is used to eliminate gaps within the segment, ensuring that the copy operation is completed within a single bus burst cycle, avoiding the bus handshake overhead caused by block writing. After the copy is completed, the system uses atomic instructions to synchronously jump the read and write pointers of the second-level recursive carbon shadow ledger to the end of the new write segment, ensuring that subsequent monitoring threads can immediately reference the latest recursive data without additional offset calculations. Immediately afterwards, the master control data bus clears historical residual records outside the frozen area of ​​the first-level instantaneous carbon shadow ledger, resets the write pointer to the starting address of the storage segment, and restarts the write channel, providing a clean buffer for the next sampling cycle. The entire overwrite process is completed within a millisecond-level time slice. During this time, the read end always points to the second-level recursive carbon shadow ledger that has been copied. The write end only pauses briefly during the locking period and resumes normal writing immediately after the clearing operation is completed. This enables the continuous alternation of ledger roles in the time domain and ensures that the data referenced by subsequent hard threshold comparison operations are the latest cyclic sedimentation results.

[0026] Furthermore, during the initialization phase after system power-on, the main control data bus first allocates a continuous and equal-length storage area in the on-chip high-speed memory area for both the first-level transient carbon shadow ledger and the second-level recursive carbon shadow ledger. The two storage segments have identical capacities, and their starting addresses are strictly aligned to the data bus address boundaries according to the main control data bus's alignment rules. This ensures that burst writes and reads do not need to cross page limits, eliminating instruction-level wait cycles caused by address jitter at the source. The system then uses the address mapping table to initially position the write and read pointers of the first-level transient carbon shadow ledger to the starting address of its own storage segment, and configures independent write and read pointers for the second-level recursive carbon shadow ledger, also pointing to the starting address of the corresponding storage segment. Thus, the two ledgers are physically isolated from each other while maintaining symmetrical capacity, forming a symmetrical ring data channel. To facilitate rapid indexing, each ledger record is arranged in a fixed-length order, and the order of fields within the record strictly follows the arrangement rules of carbon emission increment, power drift trend, and net carbon weight. When the master control data bus receives the latest data message from a certain energy node through the energy mapping identifier, it first uses hashing to quickly locate the row offset of the energy mapping identifier in the first-level instantaneous carbon shadow ledger, and then moves the write pointer to the first byte of the corresponding record and writes the complete record data at once.

[0027] After writing is complete, the write pointer increments sequentially by one record length and immediately updates the loop counter. When the write pointer reaches the end of a storage segment, it automatically rolls back to the beginning address of the storage segment, forming a circular write operation within the physically enclosed memory area. Since both storage segments have equal capacity and their beginning addresses are aligned with the master control data bus address boundary, the write and read pointers maintain consistent alignment granularity during incrementing or rolling back, ensuring that each bus handshake completely covers the entire block of data for one or more records, thus avoiding a half-write / half-read state at the hardware level. The read thread is completely independent; it continuously fetches stable data segments from the first-level instantaneous carbon shadow ledger before the latest freeze based on the read pointer, ensuring that the snapshot received by the consumer does not contain any intermediate states being written. Meanwhile, the second-level recursive carbon shadow ledger, after the last time slice rotation, has already carried the complete dataset settled from the previous cycle. Its write and read pointers are synchronized to the end of the settled segment, ready for use by the threshold comparison thread. When the system reaches the millisecond-level time slice boundary, the scheduler triggers the dual-ledger rotation process. First, it locks the first-level write channel and records the write pointer position as the frozen boundary. Then, it atomically swaps the first-level read and write pointers, making the frozen segment immediately visible to the read end. Next, it calls a high-speed transfer instruction to copy the entire frozen segment to the current write pointer position of the second-level recursive carbon shadow ledger, verifying that the total number of bytes written matches the number of records multiplied to ensure no records are truncated or omitted. After copying, it simultaneously updates the second-level write and read pointers to ensure the recursive ledger remains aligned. Finally, it resets the first-level write pointer back to the beginning address of the storage segment and releases the write lock, allowing the write and read threads to cycle through to the next time slice. The entire process is completed within a single millisecond-level time slice, with all pointer movements performed in fixed steps, thus eliminating unaligned access on the bus side. The dual-ledger structure ensures that write and read threads never contend for the same address space, further reducing bus arbitration conflicts. Through this strictly symmetrical and aligned storage segment partitioning, independent pointer management, and ring write strategy, the system not only achieves high-throughput real-time data acquisition and settling, but also provides millisecond-level reentrant ledger snapshots for the subsequent dynamic bidirectional carbon weight self-balancing control algorithm, ensuring the data integrity and timeliness of the carbon emission control process.

[0028] Furthermore, when the high-precision clock of the master control data bus reaches the millisecond time slice boundary, it immediately enters the atomic process of data freezing operation. This process first writes the current write pointer position of the first-level instantaneous carbon shadow ledger to the circular double-buffered control table, marks the position with a freeze boundary label and records a timestamp, so that the moment when this freeze snapshot was generated can be accurately located in subsequent tracing. Immediately afterwards, the master control data bus executes an indivisible atomic instruction to instantaneously swap the read pointer and write pointer of the first-level instantaneous carbon shadow ledger, so that the read pointer directly points to the previously marked freeze boundary, while the new write pointer jumps to the first free record position after the previous round of freeze boundary. Since this operation is latched at the hardware level as a single-cycle completion, no concurrent access can intervene during the write-read pointer swap, ensuring that the data plane is completely consistent. At this moment, the data consumer thread can only read all records sequentially from the new read pointer to the new write pointer, thus forming a frozen data segment with a completely determined and unchanging start and end boundary. The reading process is not affected by the subsequent operations of the write thread, so the read content remains absolutely stable throughout the entire consumption window. Once the consumer confirms that the data segment has been completely extracted and returns a consumption completion signal to the master control data bus, the master control data bus immediately invokes the block-level direct access method to move the frozen data segment to the corresponding energy mapping identifier record position in the second-level recursive carbon shadow ledger in one go. During the movement, equal-length continuous address mapping is used, and all bytes are written to the target memory segment in a single DMA cycle using bus burst transmission, completely eliminating the instruction queuing and handshake overhead caused by writing field by field. At the same time, a hardware check counter compares the total number of bytes before and after the movement. If any mismatch is found, an error interrupt is immediately triggered to roll back the entire copy transaction. After the movement is completed, the system updates the write pointer of the second-level recursive carbon shadow ledger to point to the end of the segment that was just written, while the read pointer remains in place so that the monitoring thread can directly read the latest settling data during the next threshold comparison. Subsequently, the write channel of the first-level instantaneous carbon shadow ledger is unlocked, and the write pointer is now located in the new starting free area, allowing the data stream of the next sampling cycle to continue writing continuously, while the read pointer statically points to the segment before the frozen boundary, preparing for the next millisecond time slice trigger. The entire freezing, copying, and pointer reset process is strictly limited to a single millisecond time slice, and all critical steps are completed through bus-level atomic operations and block-level direct access, thus avoiding address jitter or bus starvation, as well as race conditions between write and read threads. Through this mechanism, the system can continuously provide highly consistent, low-latency two-layer ledger snapshots in high-frequency acquisition scenarios, providing a millisecond-level real-time and reliable data foundation for the subsequent dynamic bidirectional carbon weight self-balancing control algorithm's chain-linked power and carbon weight adjustment.

[0029] Furthermore, when the main control data bus enters the 3D voxelization preprocessing flow, it first performs a sequential traversal of the park's energy mapping integrated dataset. Following the record order retrieved by the energy mapping identifier, it sequentially reads the three fields: real-time power load, energy storage state of charge, and carbon emission increment. These three fields are then written into a fixed-length field cache in the strict order of real-time power load, energy storage state of charge, and carbon emission increment. This cache is then directly mapped to a 3D coordinate point, with real-time power load as the horizontal axis, energy storage state of charge as the vertical axis, and carbon emission increment as the depth axis. As the traversal progresses, the system synchronously calculates the minimum and maximum values ​​for each of the three coordinate axes, thereby obtaining the overall real-time power load range, energy storage state of charge range, and carbon emission increment range for the park.

[0030] Next, the main control data bus divides the three coordinate axes into several segments with equal spacing. These segments are orthogonal to each other in space, thus forming an equivalent three-dimensional voxel mesh composed of stacked regular hexahedrons. Each voxel unit occupies a three-dimensional block of equal axial length and is assigned a unique mesh number during generation. This number is formed by splicing together the numbers in the order of increasing horizontal axis number, increasing vertical axis number, and increasing depth number. Thus, the position of the voxel unit in three-dimensional space can be directly located through a simple integer sequence. The master control data bus then performs a second traversal of all energy nodes, mapping the previously obtained 3D coordinates to their corresponding voxel units one by one. If all three coordinate values ​​of a point fall within the closed interval of a voxel unit, the node is added to the node list of that voxel unit. If the point falls on the boundary where two or three dividing planes intersect, the system triggers the endpoint offset rule, pushing the point to an adjacent voxel unit closer to zero by a slight offset towards the zero-scale direction, ensuring that no node is simultaneously counted in two voxel units. As mapping continues, if multiple energy nodes accumulate within a voxel unit, the master control data bus appends the energy mapping identifiers of these nodes to the end of the node list according to the mapping order, thus maintaining the insertion order of nodes in the list consistent with the traversal order. After the traversal is complete, the system counts the length of the node list for each voxel unit. If the number of nodes is lower than a pre-set density limit, the voxel unit is marked as a sparse voxel and removed from subsequent cluster determination; the remaining voxel units are retained as candidate voxel units.

[0031] For each candidate voxel, the main control data bus allocates an accumulation register in its local register to sequentially read the real-time power load, energy storage state of charge, and carbon emission increment from the node list and perform item-by-item accumulation. After traversing the entire node list, the system performs integer division on each of the three accumulated values ​​using the number of nodes, obtaining the average real-time power load, average energy storage state of charge, and average carbon emission increment. The system then rewrites these three averages into the voxel mean set in the field order of average real-time power load, average energy storage state of charge, and average carbon emission increment, and adds the unique grid number of the voxel and the node list length to the voxel mean set entry, thus forming a complete voxel mean record. This record is then stored in the voxel mean set buffer, awaiting reference by subsequent emission level calculation threads. In this way, the main control data bus completes all the steps from single-node 3D coordinate mapping, voxel mesh partitioning, node list aggregation to voxel mean set statistics. It not only compresses massive heterogeneous node information into an iterable candidate voxel set, but also ensures that the subsequent determination of high-emission clusters, low-emission clusters, and neutral-emission clusters can be carried out at the statistical level rather than at the node-by-node level, which significantly reduces the algorithm complexity and improves real-time performance.

[0032] Furthermore, when the main control data bus enters the fixed-cycle safety threshold review process, it first freezes the latest snapshot of the voxel mean set and loads it into the register-level cache. Then, it reads the three statistical data items contained therein, namely the real-time power load average, the energy storage state of charge average, and the carbon emission increment average, in a pipeline manner. In order to ensure that the comparison results are synchronized with the latest control benchmark, the system will also call the power upper limit, energy storage safety lower limit, and carbon emission safety threshold in the static configuration table at this stage, and map them to the fast lookup table stored on the same chip in read-only mode. During the comparison process, the main control data bus first performs an unsigned integer comparison between the real-time average power load and the power limit. If the real-time average power load is found to be higher than the power limit, a power reduction instruction is immediately generated in the local control buffer. This instruction includes the target energy mapping identifier, the required reduction level, and the execution effective timestamp. Next, the system compares the average energy storage state of charge (SBC) with the energy storage safety lower limit. If the average SBC is found to be lower than the energy storage safety lower limit, the control buffer is updated, and an instruction to stop discharging and allow charging is added under the same target energy mapping identifier. Subsequently, the average carbon emission increment is compared with the carbon emission safety threshold. If the average carbon emission increment is higher than the threshold, the control buffer is modified again to generate a power limiting instruction for that node. Since the three comparisons may be triggered simultaneously or only partially, the system establishes an instruction merging stack for the same energy mapping identifier in the control buffer. Multiple instructions are deduplicated and merged according to priority to prevent conflicting instructions from being issued to the same node. After all candidate voxel units have completed threshold comparison and generated instructions, the master control data bus enters the instruction scheduling phase. The system first sorts the instruction set in the control buffer according to the priority of power reduction, stopping discharge, and power limiting, and then writes it into the instruction broadcast register. After the broadcast register is triggered, it automatically packages the instructions in a predefined frame format and sends them to all connected energy nodes through the master control data bus. At the same time, the master control data bus records the batch number, number of instructions, and target node list of this broadcast in the system event log for subsequent auditing and backtracking. After receiving the broadcast instruction, the energy node will immediately return an execution confirmation message. After collecting the confirmation information, the master control data bus refreshes the node status flag to ensure that the chained power and carbon weight linkage adjustment is implemented at the physical layer. If any node fails to return confirmation within the specified timeout period, the system will automatically trigger the retransmission mechanism and generate an abnormal alarm after the retransmission threshold is exhausted. Through this closed loop consisting of periodic threshold comparison, command merging, synchronous broadcasting, and execution confirmation, the park's carbon-neutral smart microgrid system can constrain node power output, energy storage state of charge, and carbon emission increments in real time using hard safety boundaries without relying on predictive models. This ensures both the safety of grid operation and that the overall carbon emission intensity of the park remains within the set threshold.

[0033] Furthermore, after the main control data bus enters the cluster scan preprocessing stage, it first loads the current cluster node list continuously into the register cache and assigns it an ascending sequential index according to the natural order in which the nodes appear in the list. To ensure the reversibility and determinism of subsequent reordering operations, the system immediately performs parity classification on the entire sequential index array after index allocation, assigning all odd indices to odd segments and all even indices to even segments. Subsequently, the system reorders the index elements within the odd segments in descending order to generate reverse odd segments; the even segments maintain their original ascending order to generate sequential even segments. The reverse odd segments and sequential even segments are merged in memory using sequential concatenation to form the initial segment sequence, without introducing any extra padding bits to ensure index continuity. After the initial segment sequence is formed, the main control data bus assigns a rotating index to each node in the sequence and stores the rotating index in a one-to-one mapping with the node's original sequential index. Next, the system performs segmented traversal of the initial segmented sequence with a fixed jump length of two: starting from the node corresponding to the first rotation index, it visits the next node by skipping two elements in sequence, forming the first round of traversal subsequence; when the traversal pointer reaches the end of the sequence, the system shifts the jump start point one position to the right and starts the same jump operation again from the second rotation index, generating the second round of traversal subsequence. This shifting and traversal process is executed eight times. After eight cycles, the original initial segmented sequence is rearranged into a cross-rotation sequence, at which point the relative positions of each node in the sequence are re-dispersed. To further improve the dispersion in access time, the main control data bus adds jump backfilling logic to the cross-rotation sequence: for nodes skipped in each traversal, they are inserted into the end of the current traversal subsequence in the order they were skipped after the next round of traversal, thus ensuring that all nodes are visited once and only once within eight rounds of traversal. The system names the final sequence obtained after eight rounds of iterations as the cross-step rotation sequence and writes it into the scan path register table as the node access order for chained power and carbon weight commands. Because the reverse parity segmentation breaks the linear relationship of the original index, and the cross-step rotation further disperses the relative positions of accessed nodes in the time dimension through a combination of fixed step size and starting displacement, the cross-step rotation sequence can simultaneously widen the distance between adjacent nodes in both physical space and electrical topology, significantly reducing the probability of batch power adjustments instantaneously overlapping on the same electrical branch. When actually issuing commands, the system proceeds sequentially based on the cross-step rotation sequence. After processing each node, the sequence head pointer is incremented by one and written back to the register table. Once the pointer returns to the beginning of the sequence, one round of full cluster scanning is completed. Then, the rotation index shift process is re-triggered to ensure that the node order is re-discrete in the next scan round. Through this strictly defined reverse-order parity segmentation and cross-step rotation mechanism, the main control data bus maximizes the spatial and temporal dispersion of command triggering while maintaining the determinism of the algorithm, providing robust peak suppression protection for the entire park's carbon neutral smart microgrid system.

[0034] The process of forming a sixteen-phase interleaved recursive scan path by combining the main control data lines in a diagonal backtracking manner to avoid instantaneous spikes caused by batch adjustments includes: mapping the cross-step rotation sequence to a two-dimensional virtual coordinate plane, where the horizontal axis is the rotation index and the vertical axis is the traversal round; inserting marker points backtracking from the lower right corner to the upper left corner at a 1:1 slope, adding the diagonal node of the previous round of backtracking after each marker point insertion, forming a diagonal backtracking chain; after the diagonal backtracking chain is inserted, the cross-step rotation sequence is expanded into a diagonal backtracking sequence, and the newly added diagonal backtracking visits between nodes can be interspersed with different rounds; and then sequentially arranging the diagonal backtracking sequence... The system is evenly divided into sixteen segments, each called a phase segment. The main control data bus assigns phase numbers one to sixteen to phase segments one through sixteen, arranged in ascending order of phase number. The first and last nodes of each phase segment are recorded as the preceding and following nodes, providing location information for subsequent recursive binding. The main control data bus connects the preceding node of phase segment one to the following node of phase segment two, and the preceding node of phase segment two to the following node of phase segment three, and so on recursively until the following node of phase segment sixteen is connected to the preceding node of phase segment one in a closed loop. After the loop is closed, a recursive chain is formed, ensuring that the access path is connected end-to-end among the sixteen segments. No forks occur; based on the first recursive chain, recursive wrapping is performed fifteen times. Each recursive wrapping inserts a new link at the node interval of the previous recursive chain, causing the access path to wrap layer by layer. After the recursive wrapping is completed, a sixteen-layer wrapping chain is formed, collectively called a sixteen-phase intertwined recursive scan path. The distance between adjacent nodes in any two layers of the sixteen-layer link is no less than two, ensuring that triggering control commands will not concentrate on adjacent nodes. The main control data bus writes the sixteen-phase intertwined recursive scan path into the clustered path register table, flattening it into a one-dimensional scan array in node sequence. The scan array is output dynamically in a queue from index one to the final index value. The bidirectional carbon weight self-balancing control algorithm issues power and carbon weight commands to nodes according to the scanning array sequence. When the scanning array traversal is completed, a round of full-cluster chain adjustment is completed, the master control data bus resets the array index and waits for the next cycle to be triggered. After each round of scanning, the master control data bus monitors the instantaneous peak amplitude within the power adjustment window. If the peak amplitude is detected to exceed the preset threshold, the path fine-tuning process is automatically triggered: the phase segment positions of phase number seven and phase number eight are swapped; the recursive winding relationship is adjusted synchronously to maintain the link closed loop; after the path fine-tuning is completed, the cluster path register table is updated and the next round of scanning is entered to ensure continuous suppression of peaks.

[0035] The following is a complete embodiment demonstrating the operational details of a carbon-neutral smart microgrid system in a park within a sampling period (sampling interval of 1 second, algorithm loop period of 100 ms). In this embodiment, the park has a total of 12 new energy nodes, including 4 photovoltaic array nodes, 3 wind turbine nodes, 1 gas-fired combined cooling, heating and power (CCHP) node, and 4 energy storage battery nodes. The system is required to ensure a stable power supply for a maximum instantaneous load of 1000 kW under the condition that the park's annual average carbon emission intensity does not exceed 0.30 kg CO2 / kWh.

[0036] The first step is for the main control data bus to assign a unique energy mapping identifier (ID) to each node. i (i = 1, ..., 12). The acquisition system obtains the power P of each node in real time. i Energy storage state of charge (SOC) i Carbon emission increment C i The initial vectors that make up the nodes: Among them, P i The unit is kilowatt (kW), SOC i The unit is percentage (%), C i The unit is carbon emissions per second (kgCO2 / s). Assume data is collected from some nodes: Node 1: P1 = 180, SOC1 = 75, C1 = 0.04; Node 2: P2 = 200, SOC2 = 82, C2 = 0.00. Other nodes are recorded in the same format.

[0037] The second step is to determine the global scope of each field: P min =0,P max =240; SOC min =10,SOC max =100; C min =0,C max =0.06; Divide each of the three coordinate axes into four equally spaced facets, resulting in 5×5×5=125 voxel elements. The coordinates of node 1 (P1,SOC1,C1)=(180,75,0.04) fall into the voxel index (3,3,4), and the mesh number is obtained by the formula: N=(x-1)×25+(y-1)×5+z, which gives N1=94. After performing the same mapping on 12 nodes, only 9 voxel elements have a node density exceeding the threshold of 2, and the rest are marked as sparse voxels.

[0038] The third step is to calculate the average power for each candidate voxel unit. Mean of state of charge of energy storage Average increase in carbon emissions With voxel N 94 For example, considering nodes 1, 4, and 7, calculate:

[0039] The fourth step is to compare the system's set safety thresholds: power limit P. lim =220kW, lower limit of safe SOC for energy storage lim =20%, carbon emission safety threshold C lim =0.05 kg CO2 / s. The judgment condition is: For N 94 ,have: Therefore, hard threshold control is not triggered.

[0040] Step 5: Classify voxel groups according to average carbon emissions: High-emission clusters: Neutral emission clusters Low-emission clusters: Voxel N 94 Belongs to the neutral cluster, assuming N 88 With N 103 Belongs to the high-emission cluster, N 12 With N 27 It belongs to the low-emission cluster.

[0041] Step 6: Construct the access sequence for high-emission clusters. Node list [2,5,6,9] (by ID) i Starting in ascending order, reverse the odd segment [2,6] to get [6,2], keep the even segment [5,9] as [5,9], and concatenate the initial segment sequence [6,2,5,9]. With a fixed step length of 2, perform 8 cycles to obtain the cross-step cycle sequence [6,5,2,9].

[0042] Step 7: Map the sequence to a two-dimensional coordinate system (x,y) and insert backtracking chains with a slope of 1:1. The starting point is the bottom right corner node (4,2). After inserting the diagonal node from the previous round into the current position, the sequence expands to [6,5,2,9,5]. Repeat this process for the remaining nodes to complete the generation of the diagonal backtracking chain.

[0043] Step 8: Divide the final sequence into 16 phase segments and assign phase numbers 1, ..., 16. The recursive chain is connected according to the following rules: Link(F k ,L k+1 ),k=1,…,15;Link(F 16 ,L1); where F k L is the leading node of phase segment k. k+1 Let be the trailing node of phase segment k+1. After recursively wrapping 15 times, a sixteen-layer link Γ is obtained.

[0044] Step 9: Flatten Γ into a scan array S with length L S =64. The control algorithm executes instructions in S order. Example: Node 6 carbon emission increment C6 = 0.055 > C limThe power reduction operation is triggered: ΔP6 = -0.10P6; Node 5 is a low-carbon exhaust wind power node, which is given reverse carbon credit compensation, and the power is increased: ΔP5 = +0.06P5.

[0045] Step 10: Scan complete, detect power spikes, and calculate peak value. If M peak If the power is >60kW, then swap phase segments numbered 7 and 8 and correct Γ. Example: M is detected. peak =48kW, no adjustment required.

[0046] Step 11: The scheduling cycle ends. The 48 frozen records in the first-level ledger are copied to the second-level ledger via DMA. The cumulative net emission reduction is: ΔQ = ∑ i∈High (C i -C lim )-∑ j∈Low (C lim -C j ) = 0.006 kg CO2;

[0047] The meanings of each parameter are as follows: ID i : Unique mapping identifier for the node; P i Node power (kW); SOC i Energy storage state of charge (%); C i Carbon emission increment (kgCO2 / s); Node initialization vector; P min ,P max Global power range; SOC min SOC max : Global energy storage coverage; C min C max : Global range of carbon emissions; N: Voxel number; Mean values ​​of three terms within a voxel; P lim Power limit; SOC lim : Lower limit of energy storage safety; C lim Γ: Carbon emission safety threshold; Γ: Sixteen-layer recursive scan chain; S: Scan array; M peak ΔP6, ΔP5: Power peak amplitude; ΔP6, ΔP5: Node power adjustment values; ΔQ: Total net emission reduction for the cycle.

[0048] This invention is not limited to the specific embodiments described above. The invention extends to any new feature or combination disclosed in this specification, as well as any new method or process step or combination disclosed herein.

Claims

1. A carbon-neutral smart microgrid system for industrial parks, characterized in that, The system includes: a main control data bus and several energy nodes; a unique energy mapping identifier is established for each energy node on the main control data bus; the power, energy storage status, and carbon emission value of the energy node are collected in real time through the energy mapping identifier and uploaded to the main control data bus; the main control data bus aggregates all collected values ​​to form an integrated energy mapping dataset for the park; based on the dataset, a dynamic bidirectional carbon weight self-balancing control algorithm is invoked to perform multi-round chain-like power and carbon weight linkage adjustments on all energy nodes in the park to ensure that the overall carbon emission intensity of the park remains within the preset carbon emission threshold, generating a first-level instantaneous carbon shadow ledger and a second-level recursive carbon shadow ledger for the energy nodes; the main control data bus reads the real-time power load, energy storage charge status, and carbon emission increment of each energy node according to the second-level recursive carbon shadow ledger within a fixed period, compares them with their respective upper limit values, and generates power control commands.

2. The carbon-neutral smart microgrid system for industrial parks as described in claim 1, characterized in that, The main control data line, based on the dataset, invokes a dynamic bidirectional carbon weight self-balancing control algorithm to perform multi-round chain-like power and carbon weight linkage adjustments on all energy nodes in the park. Specifically, this process includes: simultaneously configuring a first-layer instantaneous carbon shadow ledger and a second-layer recursive carbon shadow ledger for each energy mapping identifier, and achieving seamless switching between the two ledgers in milliseconds using a ring-shaped double-buffering technology; performing voxel-based clustering of all energy nodes based on real-time power load, energy storage state of charge, and carbon emission increment three-dimensional coordinates, and labeling them as high-emission clusters, low-emission clusters, and neutral-emission clusters accordingly; and processing each cluster... A sixteen-phase interleaved recursive scanning path is formed by combining reverse-order odd-even segmentation, cross-step rotation, and oblique backtracking to avoid instantaneous spikes caused by batch adjustments; time-series double interpolation is performed on all neutral emission nodes, and repair power compensation pulses are injected and bound with unique traceability tags; virtual carbon slow-release valves are set at high emission nodes to form gradient reduction curves, and power downgrading is triggered immediately when the carbon emission curve touches the baseline of the reduction valve; the surplus energy generated after the power of low emission nodes is increased is preferentially backfilled into their energy storage units, and if the state of charge approaches the safe limit, it is switched to the external grid backfeed channel.

3. The carbon-neutral smart microgrid system for industrial parks as described in claim 2, characterized in that, After the main control data bus completes the synchronous broadcast of the instructions to all energy nodes, it moves the first-level instantaneous carbon shadow ledger up to cover the second-level recursive carbon shadow ledger.

4. The carbon-neutral smart microgrid system for industrial parks as described in claim 3, characterized in that, The on-chip high-speed storage area of ​​the main control data bus is divided into contiguous storage segments for the first-level instantaneous carbon shadow ledger and the second-level recursive carbon shadow ledger, with both segments having equal capacity. To avoid address jitter, the starting addresses of both storage segments are aligned with the address boundaries of the main control data bus. At system startup, write and read pointers are set for the first-level instantaneous carbon shadow ledger, with both initially pointing to the first address of the ledger's storage segment. Independent write and read pointers are set for the second-level recursive carbon shadow ledger, with their initial values ​​also pointing to the first address of the ledger's storage segment. After receiving the latest data from a certain energy node, the main control data bus locates the corresponding record position in the first-level instantaneous carbon shadow ledger according to the energy mapping identifier.

5. The carbon-neutral smart microgrid system for industrial parks as described in claim 4, characterized in that, Whenever the main control data bus detects that the millisecond time slice boundary has been reached, a data freeze operation is triggered: first, the write pointer position of the first-level instantaneous carbon shadow ledger is written to the circular double buffer control table and marked as the freeze boundary; Subsequently, the atomic method swaps the read and write pointers, making the read pointer point to the latest frozen boundary and the write pointer point to the position after the previous frozen boundary. At this time, the data consumer only reads the complete frozen data segment from the position before the write pointer, ensuring that the read content remains unchanged. After the data reading is completed, the master control data bus copies the frozen data segment of the first-level instantaneous carbon shadow ledger to the record position of the same energy mapping identifier in the second-level recursive carbon shadow ledger. The copying operation is completed in one go using a block-level direct access method, avoiding the latency caused by field-by-field write operations.

6. The carbon-neutral smart microgrid system for industrial parks as described in claim 5, characterized in that, The main control data bus reads the real-time power load, energy storage state of charge, and carbon emission increment corresponding to each energy mapping identifier from the dataset one by one, and assembles them into three-dimensional coordinate points according to the field order of real-time power load, energy storage state of charge, and carbon emission increment. Based on the overall real-time power load range, energy storage state of charge range, and carbon emission increment range of the park, the main control data bus divides each of the three coordinate axes into equally spaced planes. These planes are pairwise orthogonal, forming a three-dimensional voxel mesh. Each voxel unit corresponds to a unique mesh number in three-dimensional space, and the mesh number is obtained through... The coordinate axis numbers are generated in an incremental manner; the main control data bus traverses all energy nodes and maps their three-dimensional coordinate points to the corresponding voxel units; if there are multiple energy nodes in a voxel unit, they are added sequentially according to the mapping order to form a node list; if the coordinates of an energy node are located on the voxel boundary surface, the endpoint offset rule towards the zero scale is used to classify it into the adjacent voxel unit close to the zero scale; the main control data bus calculates the number of nodes in each voxel unit; if the number of nodes is lower than the preset density lower limit, the voxel unit is marked as a sparse voxel and temporarily not involved in cluster determination. The remaining voxel units are retained as candidate voxel units; for each candidate voxel unit, the average real-time power load, average energy storage state of charge, and average carbon emission increment in the main control data bus statistical node list are denoted as the voxel mean set.

7. The carbon-neutral smart microgrid system for industrial parks as described in claim 6, characterized in that, The main control data bus reads the real-time average power load, average energy storage state of charge, and average carbon emission increment of each energy node from the voxel average set at fixed intervals, and compares them one by one with their respective preset power upper limit, energy storage safety lower limit, and carbon emission safety threshold: if the real-time average power load is higher than the power upper limit, a power reduction command is issued to the corresponding energy node; if the average energy storage state of charge is lower than the energy storage safety lower limit, a stop discharge command is issued to the corresponding energy node and charging is allowed; if the average carbon emission increment is higher than the carbon emission safety threshold, a power limiting command is issued to the corresponding energy node; the above commands are broadcast synchronously to all energy nodes.

8. The carbon-neutral smart microgrid system for industrial parks as described in claim 7, characterized in that, The process of the main control data line segmenting each cluster in reverse order and alternating between odd and even numbers and performing cross-step rotation includes: The main control data bus first reads the current cluster node list and assigns a sequential index to the list according to the order of the energy mapping identifier; the sequential index is divided into odd and even segments according to the odd sequence and even sequence; the odd segments are rearranged in descending order of the sequential index to obtain the reverse odd segments; the even segments are arranged in ascending order of the sequential index to obtain the sequential even segments; the reverse odd segments and the sequential even segments are concatenated to generate the initial segment sequence, laying the sequential foundation for subsequent scanning; the main control data bus assigns a rotation index to the initial segment sequence and performs segment traversal from the first index with a fixed step length of two; after each traversal, the step starting point is shifted one position to the right to form a cross-step rotation sequence; after the cross-step rotation sequence is cycled eight times, the initial segment sequence is rearranged into a cross-step rotation sequence, and the node access order is fully discretized.

Images