Power grid dynamic carbon emission intensity prediction method

Abstract

The present application relates to carbon emission management technical field, specifically relates to a kind of power grid dynamic carbon emission intensity prediction method.The present application is first obtained the equipment discrete power consumption sequence of target batch product, further obtains the local dynamic carbon emission factor sequence of the production time window of target check;Further construct state transition diagram, search output lowest carbon consumption power reference sequence;Further obtain the local offset of each time step, and accumulate to obtain carbon emission time deviation degree;Finally, according to the difference between carbon emission time deviation degree and industry average time deviation degree, combined with preset adjustment weight and the basic static carbon emission intensity of target batch product, obtain dynamic prediction carbon emission intensity, objectively quantifies the real carbon reduction performance of peak-shaving scheduling, improves the accuracy of prediction result.

Description

Technical Field

[0001] This invention relates to the field of carbon emission management technology, and more specifically to a method for predicting the dynamic carbon emission intensity of power grids. Background Technology

[0002] The allocation of regional carbon emission allowances is based on product lifecycle carbon footprint accounting. In existing accounting practices, a static allocation is typically performed using the average grid emission factor combined with the enterprise's total electricity consumption. However, under the operating environment of large-scale grid connection of new energy sources, grid output fluctuations intensify. Thermal power units need to frequently participate in deep peak shaving to balance grid fluctuations. Boiler combustion efficiency decreases during low-load operation, leading to dramatic fluctuations in the actual carbon emissions per unit of electricity generated at different times. Some energy-intensive manufacturing enterprises have objectively acquired the ability to stagger production by constructing upstream and downstream intermediate material warehouses: that is, stopping equipment operation during periods of high grid carbon emissions and operating at full capacity to stockpile materials during periods of low grid carbon emissions. Furthermore, heavy industrial equipment exhibits inseparable continuous energy consumption characteristics during cold starts and continuous operation; each startup must complete a full operating cycle.

[0003] Traditional static distribution methods directly allocate the total electricity consumption evenly within the production scheduling cycle, completely masking the actual start-up and shutdown patterns of equipment in industrial settings. They fail to objectively identify the differences in the distribution of production scheduling electricity consumption over time, resulting in inaccurate analysis of the actual dynamic carbon emission intensity of products throughout their life cycle. Summary of the Invention

[0004] To address the technical problem that existing static accounting methods cannot objectively quantify the true carbon reduction performance of enterprises' cross-period peak-shifting scheduling, leading to distorted prediction results of dynamic carbon emission intensity of products, the present invention aims to provide a method for predicting the dynamic carbon emission intensity of power grids. The specific technical solution adopted is as follows:

[0005] Based on the fluctuation characteristics of the historical continuous operation flow data corresponding to the target batch of products, the discrete power consumption sequence of the equipment is obtained; the production scheduling time window for target verification is divided into time steps, and the thermal power load rate of each time step is obtained by combining the power grid settlement log. The load rate-carbon emission mapping function of the thermal power unit is substituted into the local dynamic carbon emission factor sequence.

[0006] Material data is discretized, and triple nodes are formed by time step index, intermediate material warehouse inventory, and cumulative material consumption. The discrete power consumption sequence of the equipment is used as the transition edge, and the path weight is calculated using the local dynamic carbon emission factor sequence to construct a state transition graph. The lowest carbon emission power consumption benchmark sequence is searched and output. Based on the sum of the actual time-series power consumption sequence and the sum of the lowest carbon emission power consumption benchmark sequence, and combined with the actual time-series power consumption sequence up to each time step and the lowest carbon emission power consumption benchmark sequence, the local offset of each time step is obtained. All the local offsets are accumulated to obtain the carbon emission time deviation.

[0007] Based on the difference between the carbon emission time deviation and the industry average time deviation, and combined with the preset adjustment weight and the basic static carbon emission intensity of the target batch of products, the dynamic predicted carbon emission intensity is obtained.

[0008] Furthermore, the method for obtaining the discrete power consumption sequence of the device includes:

[0009] The matrix contour algorithm is used to extract features from the historical continuous operation flow data, and the repeating subsequences corresponding to local minima are extracted as precise equipment power consumption segments. The starting time of the precise equipment power consumption segment is used as the alignment reference, and time slices are performed according to a preset fixed step size. The instantaneous power in each time slice is integrated with respect to time to obtain the power consumption value of the corresponding time step. All the power consumption values ​​are concatenated in chronological order to construct a one-dimensional time series array as the discrete power consumption sequence of the equipment.

[0010] Furthermore, the method for obtaining the local dynamic carbon emission factor sequence includes:

[0011] Extract the actual total electricity load of the province, the actual settlement amount of green electricity transmitted across provinces, and the actual output value of local wind and solar new energy corresponding to each time step in the power grid settlement log; add the actual total electricity load of the province and the actual settlement amount of green electricity transmitted across provinces, and subtract the actual output value of local wind and solar new energy to obtain the actual load undertaken by local thermal power at each time step; combine this with the total available installed capacity of thermal power units to obtain the thermal power load factor at each time step.

[0012] Substitute all the aforementioned thermal power load rates into the load rate-carbon emission mapping function to generate the local dynamic carbon emission factor sequence.

[0013] Furthermore, the method for constructing the state transition diagram includes:

[0014] All discrete power consumption sequences of all devices are treated as directed transfer edges emanating from the triplet node; the discrete power consumption sequence of the device corresponding to the directed transfer edge is multiplied and accumulated with the data values ​​of all time steps with the same sequence number in the local dynamic carbon emission factor sequence to obtain the path weight of the directed transfer edge.

[0015] Furthermore, the method for obtaining the lowest carbon emission power consumption benchmark sequence includes:

[0016] Starting from the initial state, a path search is performed in the state transition graph. During the search expansion process, if the intermediate material warehouse inventory in the triplet node after the transition is less than zero or greater than the maximum design capacity of the intermediate material warehouse, or the updated time step index exceeds the total number of steps in the production scheduling time window of the target verification, the corresponding transition edge is removed to stop the expansion to the triplet node.

[0017] According to the principle of minimizing cumulative path weight, the non-removed triplet nodes are expanded outward to generate transferred triplet nodes. When generating the transferred triplet nodes, if the cumulative material consumption in this node is greater than the discretized total order demand, the excess material difference is obtained. The material difference is multiplied by the smallest physical unit of measurement to restore the physical overproduction difference. The physical overproduction difference is multiplied by a preset material equivalent carbon emission penalty factor, and the multiplication result is added to the cumulative path weight of the node for correction. Based on the corrected cumulative path weight, the non-removed transferred triplet nodes are included in the queue to be expanded for subsequent comparison, and the expansion and comparison steps are repeated until the node with the smallest cumulative path weight in the current queue to be expanded has a cumulative material consumption greater than or equal to the total order demand. At this point, the expansion stops and the corresponding shortest path is extracted.

[0018] The discrete power consumption sequences of the devices corresponding to the shortest path are concatenated in chronological order to output the minimum carbon emission power consumption benchmark sequence.

[0019] Furthermore, when there are no unremoved triple nodes to expand outward, and the cumulative material consumption of the triple nodes in the current queue to be expanded has not reached the total order demand, the search is determined to have failed and the basic static carbon emission intensity of the target batch of products is directly output.

[0020] Furthermore, the method for obtaining the local offset includes:

[0021] For each time step, calculate the percentage of actual time-series power consumption accumulated from the starting time step to each time step, and the percentage of minimum carbon emission power consumption benchmark sequence accumulated from the starting time step to each time step; based on the difference between the percentage of actual time-series power consumption and the percentage of minimum carbon emission power consumption benchmark sequence, obtain the local offset of the corresponding time step.

[0022] Furthermore, the method for obtaining the dynamically predicted carbon emission intensity includes:

[0023] The difference between the industry average time deviation and the carbon emission time deviation is obtained, and the proportional difference relative to the industry average time deviation is used as a dynamic correction coefficient; the preset adjustment weight and the dynamic correction coefficient are fused, the fusion result is negatively correlated and truncated to a preset safety threshold range to obtain a correction factor;

[0024] By combining the basic static carbon emission intensity and the correction factor, a dynamic predicted carbon emission intensity is obtained.

[0025] Furthermore, discretizing the material data includes: the material data includes the silo input and consumption during the discrete power consumption sequence of the equipment operation, the maximum design capacity of the intermediate silo, and the total order demand; the silo input, the silo consumption, the maximum design capacity of the intermediate silo, and the total order demand are divided by the smallest physical unit of measurement and rounded down to complete the discretization of the material data.

[0026] Furthermore, the method for obtaining the load factor-carbon emission mapping function includes:

[0027] Obtain the monthly average load rate data and approved carbon emission data per unit of electricity from the historical verification ledger of thermal power units; perform polynomial fitting with the monthly average load rate data as the independent variable and the approved carbon emission data per unit of electricity as the dependent variable to generate the load rate-carbon emission mapping function.

[0028] The present invention has the following beneficial effects:

[0029] This invention first obtains the discrete power consumption sequence of the equipment for the target batch of products, revealing the objectively existing fixed fluctuations and start-up / shutdown patterns of the equipment. It then obtains the local dynamic carbon emission factor sequence for the production scheduling time window of the target verification, quantifying the real environmental cost of grid power consumption at each time period within the production scheduling cycle, providing accurate time-dimensional weight guidance for subsequent planning of the theoretically lowest carbon emission path. Furthermore, it constructs a state transition diagram, abstracting the enterprise's production scheduling process into a mathematical search model constrained by physical storage boundaries and the law of material conservation. The search outputs the lowest carbon emission power consumption benchmark sequence, representing the optimal off-peak carbon reduction production power consumption curve that the enterprise can achieve under objective hardware conditions. It further obtains the local offset at each time step, accurately reflecting the magnitude of the enterprise's power consumption deviating from the optimal benchmark at a single time section, accumulating these to obtain the carbon emission time deviation degree. Finally, based on the difference between the carbon emission time deviation degree and the industry average time deviation degree, combined with preset adjustment weights and the basic static carbon emission intensity of the target batch of products, it introduces a reward and punishment mechanism for the enterprise's off-peak power consumption behavior on the basis of traditional static accounting, enabling the final calculated dynamic predicted carbon emission intensity to more objectively reflect the actual situation of production scheduling. Attached Figure Description

[0030] Figure 1A flowchart illustrating a method for predicting dynamic carbon emission intensity of a power grid according to an embodiment of the present invention;

[0031] Figure 2 This is a flowchart illustrating a method for obtaining discrete power consumption sequences of a device, as provided in an embodiment of the present invention. Detailed Implementation

[0032] The following description, in conjunction with the accompanying drawings, details the specific scheme of the power grid dynamic carbon emission intensity prediction method provided by the present invention.

[0033] Please see Figure 1 The diagram illustrates a flowchart of a method for predicting the dynamic carbon emission intensity of a power grid according to an embodiment of the present invention, specifically including:

[0034] Step S1: Based on the fluctuation characteristics of the historical continuous operation data corresponding to the target batch of products, obtain the discrete power consumption sequence of the equipment; divide the production scheduling time window of the target verification into time steps, combine the power grid settlement log to obtain the thermal power load rate of each time step, substitute it into the load rate-carbon emission mapping function of the thermal power unit, and generate the local dynamic carbon emission factor sequence.

[0035] Considering that heavy industrial equipment has an inseparable continuous energy consumption characteristic during cold start and continuous operation, and the equipment must complete a complete operating cycle for each start-up, the discrete power consumption sequence of the equipment is obtained based on the fluctuation characteristics of the historical continuous operation flow data corresponding to the target batch of products. This shows the objectively existing fixed fluctuation segments and start-up and shutdown rules of the equipment, effectively preventing the accounting model from deviating from the real physical constraints of the industrial site, and providing basic support for the subsequent construction of a reasonable power consumption transfer path.

[0036] In the context of large-scale grid connection of new energy sources, thermal power units frequently participate in deep peak shaving, resulting in a decrease in combustion efficiency when the units are operating at low loads. The actual carbon emissions per unit of electricity generated fluctuate dramatically at different times. Therefore, the production scheduling time window for target verification is divided into time steps. The thermal power load rate of each time step is obtained by combining the grid settlement log. The load rate-carbon emission mapping function of the thermal power unit is substituted into the data to generate a local dynamic carbon emission factor sequence. This quantifies the real environmental cost of using grid electricity in each period within the production scheduling cycle and provides accurate time dimension weight guidance for subsequent planning of the theoretical minimum carbon emission path.

[0037] In a preferred embodiment of the present invention, the system connects to a regional government database to obtain historical verification records of all thermal power units in the target region over the past three years; and extracts two sets of corresponding data from the historical verification records, namely, the monthly average load rate data and the approved carbon emission data of thermal power units in the historical verification records.

[0038] Specifically, the system uses the average monthly load factor as the independent variable, which is the ratio of the unit's actual power generation to its rated power generation capacity; and the corresponding approved carbon emissions per kilowatt-hour as the dependent variable, which is the number of grams of carbon dioxide produced by generating one kilowatt-hour of electricity. The system utilizes existing polynomial regression algorithms to perform polynomial fitting with the average monthly load factor as the independent variable and the approved carbon emissions per kilowatt-hour as the dependent variable, searching for a polynomial curve that minimizes the error at all data points, thus generating a load factor-carbon emission mapping function.

[0039] To prevent the polynomial function from diverging when the input independent variable deviates from the historical data range, thus leading to abnormal carbon emission values, upper and lower safety thresholds for carbon emissions per kilowatt-hour are set based on the highest and lowest theoretical thermal efficiencies of thermal power units. Over-limit truncation is applied to the carbon emissions per kilowatt-hour output of the load factor-carbon emission mapping function, expressed by the formula: , For the truncation time step Carbon emissions per kilowatt-hour The load factor-carbon emission mapping function, For time step thermal power load factor For time step Carbon emissions per kilowatt-hour; This is a truncation function used to restrict the input value to a specific range. Within the interval, The upper limit safety threshold for carbon emissions per kilowatt-hour. This is the lower limit safety threshold for carbon emissions per kilowatt-hour.

[0040] This load factor-carbon emission mapping function will be stored in the system as a fixed static parameter, which will be used to convert the grid load of different time periods into corresponding carbon emission costs.

[0041] Further, please refer to Figure 2 The flowchart illustrates a method for obtaining discrete power consumption sequences of a device according to an embodiment of the present invention, specifically including:

[0042] Step S101: The matrix contour algorithm is used to extract features from the historical continuous operation flow data, and the repeating subsequences corresponding to the local minimum values ​​are extracted as the accurate equipment power consumption segments.

[0043] Acquire historical continuous operation log data of the target equipment executing the processing technology for the target batch of products. This historical continuous operation log data is taken from the master control ledger of the dedicated production line for the target batch of products. This log data includes at least the instantaneous power of the equipment, the inbound flow rate of the intermediate material silo, and the outbound flow rate of the intermediate material silo. The log data is an operation log recorded at a fixed acquisition frequency of seconds or minutes.

[0044] Considering that heavy industrial equipment must complete a full processing cycle with each startup, the power consumption curve of the equipment does not fluctuate randomly, but presents fixed fluctuation segments. The matrix contour algorithm can efficiently measure the similarity between any two subsequences in long time series data. By calculating the distance distribution, it can accurately locate the power consumption patterns that repeatedly appear in historical operation and have highly consistent characteristics, and then accurately extract the complete processing cycle as the precise power consumption segment of the equipment.

[0045] As an example, an existing matrix contour algorithm can be used for calculation. The specific steps are as follows: Set a sliding time window of a preset length, the length of which is set according to the number of time steps corresponding to the rated single processing cycle of the target device (for example, the number of steps corresponding to 2 to 4 hours); calculate the Euclidean distance between the data sequence within the sliding time window and all other historical sequence segments to generate a distance distribution vector; extract the repeating subsequences whose values ​​are in local minima in the distance distribution vector as the precise device power consumption segments.

[0046] Step S102: Using the start time of the precise power consumption segment of the device as the alignment reference, time slicing is performed according to a preset fixed step size.

[0047] Considering that the actual extracted equipment processing time may not be an integer multiple of the fixed step size, direct aggregation can easily cause phase aliasing. By using the start time of the precise equipment power consumption segment as an absolute reference for time slicing, it can be ensured that the data division within the segment is strictly aligned with the time scale for subsequent calculation of the grid carbon emission cost, effectively eliminating the timing misalignment error during data aggregation.

[0048] Step S103: Integrate the instantaneous power within each time slice with respect to time to obtain the power consumption value for the corresponding time step.

[0049] By employing existing numerical integration techniques in this field, the instantaneous power contained in each time slice is integrated along the time axis, thereby accurately converting the power into the actual cumulative power consumption value within the corresponding time step.

[0050] Step S104: Concatenate all power consumption values ​​in chronological order to construct a one-dimensional time series array as the discrete power consumption sequence of the device.

[0051] Using the divided time slices as basic data units, the power consumption values ​​obtained by integration within each time slice are spliced ​​together in natural chronological order to construct a one-dimensional time series array. This array is then used as a discrete power consumption sequence of the equipment to characterize the energy consumption distribution characteristics of a single processing cycle.

[0052] Finally, all the extracted discrete power consumption sequences of the devices are used to construct a set of discrete power consumption sequences of devices.

[0053] For each discrete power consumption sequence of the device in the set, the corresponding associated static parameters are extracted and saved, including: the number of time steps of the working condition, which represents the total number of time steps that the complete processing cycle takes from start to finish;

[0054] The material input quantity indicates the total amount of material input into the intermediate silo during the current cycle.

[0055] The material consumption rate indicates the total amount of material consumed from the intermediate silo during the current cycle.

[0056] Meanwhile, to characterize the legitimate state of equipment shutdown and non-production, a standby discrete power consumption sequence is additionally defined and supplemented in the set. The operating time step corresponding to the standby discrete power consumption sequence is set to 1, and the corresponding power consumption value, hopper input quantity, and hopper consumption quantity are all set to 0. All equipment discrete power consumption sequences and standby discrete power consumption sequences in this set serve as alternative transition bases for subsequent outward expansion when constructing the state transition diagram.

[0057] Preferably, in one embodiment of the present invention, the production scheduling time window for target verification includes the absolute start and end time of the batch of products from the start of production to complete warehousing. A fixed step size (e.g., one hour) is set for the time step, and the production scheduling time window for target verification is divided equally according to the fixed step size. Each time step corresponds to a continuous time segment and is assigned an incremental time step index, thereby establishing a globally consistent time benchmark.

[0058] Retrieve the power grid settlement log, which includes at least the province's actual total electricity load data, the actual settlement amount of green electricity transmitted across provinces, and the actual output value of local wind and solar renewable energy. Extract the province's actual total electricity load, the actual settlement amount of green electricity transmitted across provinces, and the actual output value of local wind and solar renewable energy for each time step in the power grid settlement log. Add the province's actual total electricity load and the actual settlement amount of green electricity transmitted across provinces, and subtract the actual output value of local wind and solar renewable energy to obtain the actual load undertaken by local thermal power at each time step. Combine this with the total available installed capacity of thermal power units to obtain the thermal power load factor at each time step.

[0059] The actual load on local thermal power plants is divided by the total available installed capacity to obtain the thermal power load factor for the corresponding time step.

[0060] The formula for calculating the actual load borne by local thermal power plants at each time step t is:

[0061] ;

[0062] in, For time steps The actual load borne by local thermal power plants This represents the total actual electricity load of the province. This represents the actual settlement amount of green electricity transmitted across provinces. Actual output value of local wind and solar new energy sources.

[0063] The formula for calculating the load factor of thermal power plants is:

[0064] ;

[0065] in, For time steps thermal power load factor This represents the total available installed capacity of thermal power units.

[0066] If the actual load borne by local thermal power is less than 0, it indicates that the local electricity consumption at the corresponding time step is completely covered by new energy sources, and the actual load borne by local thermal power is forcibly assigned to 0.

[0067] Substituting all thermal power load rates into the load rate-carbon emission mapping function, the carbon emission value per kilowatt-hour corresponding to each time step is obtained. Following the natural time order of increasing time step index, all carbon emission values ​​per kilowatt-hour are concatenated into a one-dimensional array, which is then combined to generate a local dynamic carbon emission factor sequence. This sequence reflects the environmental cost of using grid electricity in each period within the production scheduling cycle. The larger the value, the more the grid relies on inefficient thermal power in that period, and the higher the carbon emission cost.

[0068] It should be noted that the production scheduling time window for target verification is determined based on the absolute start and end time of the target batch of products from the start of production to complete warehousing. The total available installed capacity of thermal power units is taken as the sum of the rated installed power of all normally grid-connected thermal power units in the target area.

[0069] Step S2: Discretize the material data, forming triplet nodes with time step index, intermediate material warehouse inventory, and cumulative material consumption. Use the discrete power consumption sequence of equipment as transition edges and calculate path weights with the local dynamic carbon emission factor sequence to construct a state transition graph and search for and output the minimum carbon emission power consumption benchmark sequence. Based on the sum of the actual time-series power consumption sequence and the sum of the minimum carbon emission power consumption benchmark sequence, and combining the actual time-series power consumption sequence up to each time step with the minimum carbon emission power consumption benchmark sequence, obtain the local offset of each time step, and accumulate all local offsets to obtain the carbon emission time deviation.

[0070] To objectively measure the true extent of a company's efforts in peak-shifting and carbon reduction during production scheduling, and to provide an absolute evaluation benchmark that meets hardware limits, it is necessary to find a production scheduling power consumption trajectory that theoretically minimizes carbon emission costs. To find this trajectory, it is necessary to first discretize the material data to avoid the problem of state space explosion caused by continuous values, and establish a finite and boundary-defined search node space. Using time step index, intermediate material warehouse inventory, and cumulative material consumption to form triplet nodes, using the discrete power consumption sequence of equipment as transition edges, and calculating path weights using the local dynamic carbon emission factor sequence, a state transition graph is constructed. This abstracts the company's production scheduling process into a mathematical search model constrained by physical storage boundaries and the law of material conservation.

[0071] The system then searches and outputs the lowest carbon emission power consumption benchmark sequence, which represents the optimal peak-shifting and carbon emission reduction production and power consumption curve that enterprises can achieve under objective hardware conditions. This effectively overcomes the shortcomings of the traditional static equalization method, which masks the actual start-up and shutdown patterns of equipment, and ensures that the accounting benchmark no longer deviates from the actual physical limitations and business progress of the industrial site.

[0072] Considering that the conventional comparison method of directly subtracting from the same time step cannot quantify the inter-period scheduling performance of enterprises using material silos to shift equipment power consumption from high-carbon periods to low-carbon periods, the sum of the actual time-series power consumption sequence and the sum of the minimum carbon emission power consumption benchmark sequence represent the actual power consumption scale and the optimal power consumption scale within the production scheduling time window, respectively. Combining the actual time-series power consumption sequence and the minimum carbon emission power consumption benchmark sequence up to each time step, it also represents the cumulative ratio of actual power consumption progress to ideal power consumption progress from the start of production to the current time step. In this way, the local offset of each time step can be obtained, which accurately reflects the extent to which the enterprise's power consumption deviates from the optimal benchmark at a single time section.

[0073] Finally, the carbon emission time deviation is obtained by summing up all local offsets, which summarizes the deviation status at all times within the production scheduling cycle. The misalignment distance of the time series data is converted into the difference aggregation of the cumulative progress ratio, which directly quantifies the time scheduling efforts made by enterprises to actively stagger peak periods, and provides a reliable correction basis for the subsequent objective generation of dynamic prediction of carbon emission intensity.

[0074] Preferably, in one embodiment of the present invention, the method for discretizing material data includes:

[0075] Material data includes the input and consumption of materials in the silo during the discrete power consumption sequence of the equipment, the maximum design capacity of the intermediate silo, and the total order demand.

[0076] Divide the input quantity of the silo, the consumption quantity of the silo, the maximum design capacity of the intermediate silo, and the total order demand by the smallest physical unit of measurement and round down. For data items that are greater than zero but whose rounding result is zero, correct the rounding result to 1 to complete the discretization of material data. The smallest physical unit of measurement is set to, for example, 1 ton, 1 batch, or 1 standard pallet. In this example, 1 ton is used.

[0077] By dividing by the smallest physical unit of measurement and rounding down, the originally continuous material parameters are transformed into a finite set of integers, thereby establishing a node space with defined boundaries. This eliminates the risk of state space explosion caused by continuous values ​​for the subsequent construction of the state transition diagram.

[0078] Furthermore, indexed by time step intermediate material warehouse inventory Cumulative material consumption The nodes that form triplet nodes, i.e. Among them, cumulative material consumption The introduction of this feature enables the system to accurately identify whether the corresponding scheduling path has completed the order task at each step of the search graph expansion, thus solving the search deadlock problem caused by the lack of a business progress dimension in traditional algorithms.

[0079] It should be noted that, in order to unify the measurement standards of the state space, the total order demand has been numerically equivalent to the total material consumption necessary to complete the target batch of products, and both were processed using the same smallest physical unit of measurement during the material data discretization stage. Therefore, the cumulative material consumption can accurately represent the order completion progress and can be directly compared with the discretized total order demand.

[0080] Treat all discrete power consumption sequences (including standby discrete power consumption sequences) in the set of device discrete power consumption sequences as directed transition edges emanating from the triple node. When starting from a triple node with a known state... When expanding outward along a certain directed transfer edge, the intermediate warehouse inventory of the triple node after transfer is generated by iteratively updating the warehouse input quantity and warehouse consumption quantity after rounding, and the cumulative material consumption quantity of the triple node after transfer is generated by iteratively accumulating the warehouse consumption quantity after rounding.

[0081] As an example: when a directed transition edge originates from a triple node with a known state... When initiated, let the directed transfer edge sent out correspond to the discrete power consumption sequence of the m-th device, the corresponding operating time step number be L(m), the rounded hopper input quantity be V(m), and the rounded hopper consumption quantity be U(m). Then, the state update method of the target triplet node after the transfer is as follows:

[0082] ;

[0083] in, Index for the current time step, For the time step index after the transfer; For the first The number of time steps in the discrete power consumption sequence of each device under operating conditions; This is the rounded-down input quantity for the hopper; This is the rounded-down consumption amount from the silo. This represents the current inventory of intermediate materials in the warehouse. This refers to the inventory of intermediate materials in the warehouse after the transfer. This represents the current cumulative material consumption. This represents the cumulative material consumption after the transfer.

[0084] After obtaining the target triple node, illegal scheduling paths need to be eliminated: the system determines whether the inventory of the target node is less than 0 or greater than the maximum design capacity of the discretized intermediate silo. If the value is less than 0, it indicates that the scheduling will lead to a production disruption; if it is greater than 0... This indicates that the scheduling will lead to silo overflow. Simultaneously, the system checks whether the time step index of the target node is greater than the total number of time steps in the production scheduling window. If any of the above out-of-bounds conditions are met, the system will directly remove the directed transition edge to prevent the algorithm from continuing to search for that illegal state.

[0085] It should be noted that in this embodiment, the input quantity of the material silo occurs before the consumption quantity of the material silo during the operation of the corresponding discrete power consumption sequence of the equipment. Therefore, the net value of the intermediate material warehouse inventory at the end of the cycle is used to determine the boundary, which can ensure that no production interruption occurs during the cycle. In other embodiments, if a discrete power consumption sequence of a certain equipment has a condition where consumption occurs before input within the cycle, the intermediate material warehouse inventory can also be checked step by step within the discrete power consumption sequence of the equipment according to the time step. When the intermediate material warehouse inventory corresponding to any time step is less than zero, the corresponding directed transfer edge is also removed.

[0086] Then, all discrete power consumption sequences of all devices (including standby discrete power consumption sequences) are used as directed transfer edges emanating from the triplet nodes. The discrete power consumption sequence of the device corresponding to the directed transfer edge is multiplied and accumulated with the data values ​​of all time steps with the same sequence number in the local dynamic carbon emission factor sequence to obtain the path weight of the directed transfer edge.

[0087] The calculation formula is: ;

[0088] In the formula, For the first The path weights corresponding to the discrete power consumption sequences of each device; For the local dynamic carbon emission factor sequence at time step The carbon emissions per kilowatt-hour; For the first The number of time steps corresponding to the discrete power consumption sequence of each device (i.e., the total number of time steps experienced by the device from start to finish in the complete processing cycle). For the first In the discrete power consumption sequence of the device, the first... The power consumption value at each time step.

[0089] Each directed transfer edge represents the execution of a specific discrete power consumption sequence of a device. When calculating the path weight, the time step index contained in the triple node that issued the directed transfer edge is used as the starting point, and a sequence segment is extracted from the local dynamic carbon emission factor sequence. The number of steps extracted from the sequence segment is equal to the number of operating time steps corresponding to the directed transfer edge, thus ensuring that the extracted sequence segment has the same length as the discrete power consumption sequence of the device and that the span of the corresponding time step sequence is completely consistent. Based on this, the power consumption value of each time step in the discrete power consumption sequence of the device is multiplied one-to-one with the carbon emission value of the same time step in the sequence segment, and all the multiplication results are summed to accurately quantify the real carbon emission cost generated by executing the directed transfer edge.

[0090] Ultimately, the system uses the existing Dijkstra shortest path algorithm to perform optimization, starting from the initial state (such as time step index 1, inventory 0, cumulative material consumption 0, or the first time step of the time series) and performing path search in the state transition graph.

[0091] During the search expansion process, if the intermediate material warehouse inventory in the triple node after the transfer is less than zero or greater than the maximum design capacity of the intermediate material warehouse, or the updated time step index exceeds the total number of steps in the production scheduling time window of the target verification, then the path is determined to violate physical reality, and the corresponding transfer edge is removed to stop expanding to the triple node.

[0092] Following the principle of minimizing cumulative path weight, the remaining triplet nodes are expanded outwards to generate transferred triplet nodes. When generating these transferred triplet nodes, if the cumulative material consumption in this node exceeds the discretized total order demand, the excess material difference is obtained. This material difference is multiplied by the smallest physical unit of measurement to reconstruct the physical overproduction difference. This physical overproduction difference is then multiplied by a preset material equivalent carbon emission penalty factor, and the result is added to the cumulative path weight of the node for correction. The calculation formula is as follows:

[0093] ;

[0094] In the formula, This is the corrected cumulative path weight. The cumulative path weights before correction; This represents the cumulative material consumption in the triplet node after the transfer. This represents the total order demand after discretization. It is the smallest physical unit of measurement; This is a preset material equivalent carbon emission penalty factor.

[0095] Based on the corrected cumulative path weight, the transferred triplet nodes that were not removed are included in the queue to be expanded for subsequent comparison, and the expansion and comparison steps are repeated until the node with the smallest cumulative path weight in the current queue to be expanded has a cumulative material consumption that is greater than or equal to the total order demand. At this point, the expansion stops and the corresponding shortest path is extracted.

[0096] By splicing together the discrete power consumption sequences of the devices corresponding to the shortest path in chronological order, the lowest carbon emission power consumption benchmark sequence is output. This represents the most perfect peak-shifting and carbon emission reduction power consumption curve that a company can theoretically achieve under objective hardware conditions, and will serve as an absolute benchmark for evaluating the company's actual performance.

[0097] It should be noted that if the total number of time steps in the output minimum carbon emission power consumption benchmark sequence is less than the total number of steps in the target production time window (that is, it indicates that the scheduling scheme has completed the total order demand ahead of schedule), then an empty time step with a power consumption value of zero is added at the end of the minimum carbon emission power consumption benchmark sequence until the total length of the sequence is strictly equal to the total number of steps in the production time window.

[0098] When there are no triple nodes that have not been removed for outward expansion, and the cumulative material consumption of the triple nodes in the current queue to be expanded has not reached the total order demand, it indicates that production cannot be completed on schedule under objective physical constraints. At this time, the search is judged to have failed and the calculation is directly stopped. The basic static carbon emission intensity of the target batch of products is directly output as the calculation result.

[0099] Since the penalty for overproduction has already been adjusted for the cumulative path weights in the preceding node expansion phase, the physical dimensions of material quantity and carbon emission cost are unified. This allows the algorithm to proactively avoid planning ineffective overproduction to piece together low-carbon-emission durations during comparison and optimization. The preset material equivalent carbon emission penalty factor is a specified multiple of the basic static carbon emission intensity; in this example, it is set to 1.5 times.

[0100] It should be noted that the parameters involved in the embodiments of this invention, such as the maximum design capacity of the intermediate silo, the total order demand, the total number of time steps in the production scheduling window, and the preset material equivalent carbon emission penalty factor, are all obtained from the specific industrial production scenario corresponding to the target batch of products, the enterprise's registered equipment parameters, or by connecting to the enterprise's production execution system and government platform. Furthermore, the analysis and expansion process for each triple node, each directed transition edge, and the corresponding path weight in the state transition diagram remains consistent; only one example is described here.

[0101] In a preferred embodiment of the present invention, the idea of ​​a one-dimensional optimal transmission algorithm is adopted. By calculating the area difference between the two power consumption sequences on the cumulative distribution function (CDF), the peak scheduling behavior in the time dimension is quantified into a specific absolute scalar value, namely the local offset, which represents the deviation between the actual power consumption progress of the enterprise and the optimal ideal progress at a single time section.

[0102] Before calculating the local offset, retrieve the actual operating power log of the target batch of products within the production scheduling time window; take the start time of the production scheduling time window as the alignment reference, slice it according to the same time step, and perform integral calculation on the instantaneous power in each time slice to obtain the actual power consumption value of the corresponding time step; concatenate all the actual power consumption values ​​in time order to obtain the actual time-series power consumption sequence.

[0103] Then, for each time step within the production scheduling time window of the target verification, calculate the percentage of actual time-series power consumption sequence accumulated from the starting time step to each time step, and the percentage of minimum carbon emission power consumption benchmark sequence accumulated from the starting time step to each time step.

[0104] Based on the difference between the actual time-series power consumption sequence ratio and the minimum carbon emission power consumption benchmark sequence ratio, the local offset of the corresponding time step is obtained.

[0105] As an example, for each time step within the production scheduling time window of the target verification, the actual time-series power consumption accumulated from the starting node to the corresponding time step is used as the numerator, the sum of the actual time-series power consumption sequence and the sum of the first preset positive parameter are used as the denominator, and the ratio of the fractions is used as the proportion of the actual time-series power consumption sequence; the minimum carbon emission power consumption benchmark accumulated from the starting node to the corresponding time step is used as the numerator, the sum of the minimum carbon emission power consumption benchmark sequence and the sum of the second preset positive parameter are used as the denominator, and the ratio of the fractions is used as the proportion of the minimum carbon emission power consumption benchmark sequence.

[0106] The absolute value of the difference between the actual time-series power consumption percentage and the lowest carbon emission power consumption benchmark percentage is used as the local offset for the corresponding time step. This represents the severity of the enterprise's actual power consumption deviating from the optimal benchmark at the corresponding time section. The calculation formula is as follows:

[0107] ;

[0108] In the formula, For time steps Local offset; The actual time-series power consumption sequence is in the first... The power consumption value at each time step; The lowest carbon emission power consumption benchmark sequence is in the first The power consumption value at each time step; This represents the total number of time steps within the production scheduling window. This is the first preset positive parameter. This is the second preset positive parameter.

[0109] By iterating through all time steps within the production scheduling window, the sum of the local offsets of all corresponding time steps is taken as the carbon emission time deviation, which represents the overall span of the enterprise's power consumption focus deviating from the grid's low-carbon emission period on the time axis, directly quantifying the enterprise's proactive peak-shifting and carbon reduction scheduling performance.

[0110] Carbon emission time deviation The calculation formula is: .

[0111] The dimensions of the first preset positive parameter and the second preset positive parameter are the same as those of the parameters being added. In this example, the values ​​of the two parameters are set to minimum values ​​according to their respective dimensions and orders of magnitude (e.g., 0.0001 in their respective dimensions), which have a minimal impact on the actual percentage calculation results.

[0112] Step S3: Based on the difference between the carbon emission time deviation and the industry average time deviation, and combined with the preset adjustment weights and the basic static carbon emission intensity of the target batch of products, obtain the dynamic predicted carbon emission intensity.

[0113] To transform the micro-level inter-period production scheduling performance of enterprises into objective and horizontally comparable evaluation coefficients, and to adjust the traditional accounting results, this paper analyzes the difference between the carbon emission time deviation and the industry average time deviation to determine the superiority or inferiority of the target enterprise's off-peak production scheduling performance compared to the overall industry level. Then, it combines preset adjustment weights with the basic static carbon emission intensity of the target batch of products for smoothing correction to obtain the dynamic predicted carbon emission intensity. Based on the traditional static accounting, this paper introduces a reward and punishment mechanism for enterprises' off-peak electricity consumption behavior, so that the final calculated dynamic predicted carbon emission intensity can more objectively reflect the actual situation of production scheduling.

[0114] Preferably, in one embodiment of the present invention, the system uses a conventional carbon emission accounting method to obtain the basic static carbon emission intensity of the target batch of products. That is, the system obtains the total power consumption and total order demand of the batch of products during the production scheduling cycle, multiplies the total power consumption by the average grid emission factor for the current year, and divides it by the total order demand to obtain the basic static carbon emission intensity. The calculation formula is as follows: ;

[0115] In the formula, Based on static carbon emission intensity, This represents the total power consumption of this batch of products during the production cycle. This is the average grid emission factor for this year; This represents the total order demand.

[0116] Among them, total power consumption, total order demand, and the average grid emission factor for the year are all conventional and known parameters in the fields of industrial production and carbon verification, which can be obtained by directly reading enterprise production ledgers and publicly available grid data.

[0117] The system connects to the regional government approval database and reads the preset industry average time deviation. This industry average time deviation is derived from the statistical average of the time deviation calculated by the government platform for manufacturing enterprises of the same type and scale in the previous year within the jurisdiction.

[0118] Considering that there are absolute differences in the factory size and total power of equipment among different enterprises, direct comparison of numerical values ​​will produce magnitude errors. Therefore, the base bias caused by the absolute size is eliminated by proportional differences to obtain dynamic correction coefficients. Furthermore, considering that the dynamic correction coefficients only represent the fluctuations in relative performance, if they are directly used for indicator calculation, they are likely to cause the final carbon emission results to fluctuate violently and become distorted. Therefore, it is necessary to introduce preset adjustment weights to smooth the intervention range.

[0119] At the same time, based on the reward mechanism that the better the enterprise's off-peak performance, the lower the corresponding carbon emission quota should be, a negative correlation mapping is applied to the intervention results; furthermore, in order to avoid extreme scheduling performance leading to carbon emission quotas that are too low to zero or facing excessively strict penalties, it is also necessary to provide a safety net for the range of restricted areas.

[0120] Based on this, the difference between the industry average time deviation and the carbon emission time deviation is obtained, and the proportional difference relative to the industry average time deviation is used as a dynamic correction coefficient.

[0121] By integrating preset adjustment weights and dynamic correction coefficients, the fusion result is negatively correlated and truncated to a preset safety threshold range to obtain the correction factor;

[0122] Finally, by integrating the basic static carbon emission intensity and the correction factor, a dynamic predicted carbon emission intensity is obtained.

[0123] As an example, the difference between the industry average time deviation and the carbon emission time deviation is used as the numerator, and the industry average time deviation plus a third preset positive parameter is used as the denominator. The calculated ratio is used as the dynamic correction coefficient, and the formula is as follows: ;

[0124] In the formula, This is a dynamic correction factor; This represents the industry average time deviation. Carbon emission time deviation; This is the third preset positive parameter.

[0125] The difference between the constant 1 and the product of the preset adjustment weight and the dynamic correction coefficient is taken as the negative correlation mapping result. This difference is then truncated to a preset safety threshold range to obtain the correction factor. The calculation formula is as follows: ;

[0126] In the formula, As a correction factor; Preset adjustment weights; This is a dynamic correction factor; This is a truncation function used to restrict a value to a certain range. Within the preset safety threshold range, This is the lower limit of the preset safety threshold range. This is the upper limit of the preset safety threshold range.

[0127] In this example, the preset adjustment weight is 0.10, the preset safety threshold range is 0.8 to 1.2; the third preset positive parameter is 0.0001, which is used to prevent the denominator from being zero, and the units are consistent with the industry average time deviation.

[0128] Finally, the product of the baseline static carbon emission intensity and the correction factor is used as the dynamic predicted carbon emission intensity of the target batch of products, representing the actual carbon emission level after production scheduling adjustments. This ensures that the quota calculation results take into account both baseline energy consumption and peak-shifting carbon reduction contributions. The calculation formula is as follows: ;

[0129] In the formula, To dynamically predict carbon emission intensity; Based on static carbon emission intensity.

[0130] It should be noted that the preset adjustment weight value of 0.10 is based on the allowable floating intervention ratio for individual emission reduction scheduling measures in conventional carbon quota verification, ensuring that the adjustment range is within a reasonable range (usually around 10%), and avoiding a drastic change in the overall accounting result due to the performance of a single measure. The preset safety threshold range of 0.8 to 1.2 is based on the conventional tolerance boundary for corporate carbon emission rewards and penalties in government management, limiting the maximum reward or penalty ratio caused by scheduling to no more than 20% of the benchmark value, thus ensuring that the final accounting result is still based on the actual absolute total electricity consumption. In other embodiments, implementers can adjust the above values ​​according to local specific carbon emission quota policies and industry standards.

[0131] In another embodiment of the present invention, after obtaining the dynamically predicted carbon emission intensity, the method further includes: feeding back the obtained dynamically predicted carbon emission intensity to the production execution system of the production line where the target batch of products is located.

[0132] The dynamically predicted carbon emission intensity is compared with a preset carbon emission intensity threshold. When the dynamically predicted carbon emission intensity is greater than the carbon emission intensity threshold, the minimum carbon emission power consumption benchmark sequence is used as the target curve to generate equipment start-up and shutdown control instructions for the next production cycle. The equipment start-up and shutdown control instructions include the equipment start-up and shutdown status corresponding to each time step within the production time window. The equipment start-up and shutdown control instructions are sent to the overall control system of the target equipment to control the target equipment to start up and run at the time step with a lower value in the local dynamic carbon emission factor sequence and switch to standby mode at the time step with a higher value in the local dynamic carbon emission factor sequence. This causes the actual time-series power consumption sequence of the next production cycle to converge to the minimum carbon emission power consumption benchmark sequence, thereby reducing the actual carbon emissions generated by the target batch of products while meeting the total order demand and the maximum design capacity constraints of the intermediate silo.

[0133] In summary, to address the technical problem that existing static accounting methods cannot objectively quantify the true carbon reduction performance of enterprises' cross-time-period peak-shifting scheduling, leading to distorted prediction results of dynamic carbon emission intensity for products, this invention aims to provide a method for predicting dynamic carbon emission intensity of power grids. This invention first obtains the discrete power consumption sequence of equipment for the target batch of products, and further obtains the local dynamic carbon emission factor sequence for the production scheduling time window of the target verification; further, it constructs a state transition diagram and searches for and outputs the lowest carbon emission power consumption benchmark sequence; further, it obtains the local offset at each time step and accumulates it to obtain the carbon emission time deviation; finally, based on the difference between the carbon emission time deviation and the industry average time deviation, combined with preset adjustment weights and the basic static carbon emission intensity of the target batch of products, it obtains the dynamically predicted carbon emission intensity.

[0134] This invention constructs a state transition diagram to search for the lowest carbon emission power consumption benchmark sequence, combines it with the actual time-series power consumption sequence to obtain the carbon emission time deviation, and corrects the basic static carbon emission intensity according to industry differences. This objectively quantifies the real carbon reduction performance of peak-shifting scheduling and improves the accuracy of prediction results.

[0135] This application also provides a computer program product that, when run on a computer device, enables the computer device to execute any of the aforementioned methods for predicting dynamic carbon emission intensity of power grids.

[0136] This application also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer device, the computer device can execute any of the aforementioned methods for predicting the dynamic carbon emission intensity of a power grid.

[0137] In the embodiments provided in this application, it should be understood that the computer device, computer program product and computer-readable storage medium provided are all used to perform the corresponding methods provided above, and therefore the beneficial effects they can achieve can be referred to the beneficial effects of the methods provided above, which will not be repeated here.

Claims

1. A method for predicting the dynamic carbon emission intensity of a power grid, characterized in that, The method includes: Based on the fluctuation characteristics of the historical continuous operation flow data corresponding to the target batch of products, the discrete power consumption sequence of the equipment is obtained; the production scheduling time window for target verification is divided into time steps, and the thermal power load rate of each time step is obtained by combining the power grid settlement log. The load rate-carbon emission mapping function of the thermal power unit is substituted into the local dynamic carbon emission factor sequence. Material data is discretized, and triple nodes are formed by time step index, intermediate material warehouse inventory, and cumulative material consumption. The discrete power consumption sequence of the equipment is used as the transition edge, and the path weight is calculated using the local dynamic carbon emission factor sequence to construct a state transition graph. The lowest carbon emission power consumption benchmark sequence is searched and output. Based on the sum of the actual time-series power consumption sequence and the sum of the lowest carbon emission power consumption benchmark sequence, and combined with the actual time-series power consumption sequence up to each time step and the lowest carbon emission power consumption benchmark sequence, the local offset of each time step is obtained. All the local offsets are accumulated to obtain the carbon emission time deviation. Based on the difference between the carbon emission time deviation and the industry average time deviation, and combined with the preset adjustment weight and the basic static carbon emission intensity of the target batch of products, the dynamic predicted carbon emission intensity is obtained.

2. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, The method for obtaining the discrete power consumption sequence of the device includes: The matrix contour algorithm is used to extract features from the historical continuous operation flow data, and the repeating subsequences corresponding to local minima are extracted as precise equipment power consumption segments. The starting time of the precise equipment power consumption segment is used as the alignment reference, and time slices are performed according to a preset fixed step size. The instantaneous power in each time slice is integrated with respect to time to obtain the power consumption value of the corresponding time step. All the power consumption values ​​are concatenated in chronological order to construct a one-dimensional time series array as the discrete power consumption sequence of the equipment.

3. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, The method for obtaining the local dynamic carbon emission factor sequence includes: Extract the actual total electricity load of the province, the actual settlement amount of green electricity transmitted across provinces, and the actual output value of local wind and solar new energy corresponding to each time step in the power grid settlement log; add the actual total electricity load of the province and the actual settlement amount of green electricity transmitted across provinces, and subtract the actual output value of local wind and solar new energy to obtain the actual load undertaken by local thermal power at each time step; combine this with the total available installed capacity of thermal power units to obtain the thermal power load factor at each time step. Substitute all the aforementioned thermal power load rates into the load rate-carbon emission mapping function to generate the local dynamic carbon emission factor sequence.

4. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, The method for constructing the state transition diagram includes: All discrete power consumption sequences of all devices are treated as directed transfer edges emanating from the triplet node; the discrete power consumption sequence of the device corresponding to the directed transfer edge is multiplied and accumulated with the data values ​​of all time steps with the same sequence number in the local dynamic carbon emission factor sequence to obtain the path weight of the directed transfer edge.

5. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 4, characterized in that, The method for obtaining the lowest carbon emission power consumption benchmark sequence includes: Starting from the initial state, a path search is performed in the state transition graph. During the search expansion process, if the intermediate material warehouse inventory in the triplet node after the transition is less than zero or greater than the maximum design capacity of the intermediate material warehouse, or the updated time step index exceeds the total number of steps in the production scheduling time window of the target verification, the corresponding transition edge is removed to stop the expansion to the triplet node. According to the principle of minimizing cumulative path weight, the non-removed triplet nodes are expanded outward to generate transferred triplet nodes. When generating the transferred triplet nodes, if the cumulative material consumption in this node is greater than the discretized total order demand, the excess material difference is obtained. The material difference is multiplied by the smallest physical unit of measurement to restore the physical overproduction difference. The physical overproduction difference is multiplied by a preset material equivalent carbon emission penalty factor, and the multiplication result is added to the cumulative path weight of the node for correction. Based on the corrected cumulative path weight, the non-removed transferred triplet nodes are included in the queue to be expanded for subsequent comparison, and the expansion and comparison steps are repeated until the node with the smallest cumulative path weight in the current queue to be expanded has a cumulative material consumption greater than or equal to the total order demand. At this point, the expansion stops and the corresponding shortest path is extracted. The discrete power consumption sequences of the devices corresponding to the shortest path are concatenated in chronological order to output the minimum carbon emission power consumption benchmark sequence.

6. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 5, characterized in that, The method further includes: when there are no unremoved triple nodes to expand outward, and the cumulative material consumption of the triple nodes in the current queue to be expanded has not reached the total order demand, the search is determined to be unsuccessful and the basic static carbon emission intensity of the target batch of products is directly output.

7. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, The method for obtaining the local offset includes: For each time step, calculate the percentage of actual time-series power consumption accumulated from the starting time step to each time step, and the percentage of minimum carbon emission power consumption benchmark sequence accumulated from the starting time step to each time step; based on the difference between the percentage of actual time-series power consumption and the percentage of minimum carbon emission power consumption benchmark sequence, obtain the local offset of the corresponding time step.

8. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, The method for obtaining the dynamic prediction of carbon emission intensity includes: The difference between the industry average time deviation and the carbon emission time deviation is obtained, and the proportional difference relative to the industry average time deviation is used as a dynamic correction coefficient; the preset adjustment weight and the dynamic correction coefficient are fused, the fusion result is negatively correlated and truncated to a preset safety threshold range to obtain a correction factor; By combining the basic static carbon emission intensity and the correction factor, a dynamic predicted carbon emission intensity is obtained.

9. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, Discretizing the material data includes: the material data includes the input and consumption of the silo during the discrete power consumption sequence of the equipment, the maximum design capacity of the intermediate silo, and the total order demand; the input, consumption, maximum design capacity of the intermediate silo, and total order demand are divided by the smallest physical unit of measurement and rounded down to complete the discretization of the material data.

10. The method for predicting the dynamic carbon emission intensity of a power grid according to claim 1, characterized in that, The method for obtaining the load factor-carbon emission mapping function includes: Obtain the monthly average load rate data and approved carbon emission data per unit of electricity from the historical verification ledger of thermal power units; perform polynomial fitting with the monthly average load rate data as the independent variable and the approved carbon emission data per unit of electricity as the dependent variable to generate the load rate-carbon emission mapping function.

Images