Paper mill load green power consumption potential evaluation method fusing ProbSparse and FEDformer

Abstract

The invention discloses a paper mill load green power consumption potential evaluation method fusing ProbSparse and FEDformer, and belongs to the technical field of power system regulation and control, and the method comprises the steps: constructing an up-regulation capability evaluation model and a down-regulation capability evaluation model, and obtaining a maximum up-regulation capability and a maximum down-regulation capability; based on time-frequency domain decoupling modeling, predicting a future green power factor sequence; and based on the maximum up-regulation capacity, the maximum down-regulation capacity and the future green power factor sequence, a load green power consumption potential evaluation model is constructed, and according to the ProbSparse and FEDformer fused paper mill load green power consumption potential evaluation method, a ProbSparse-based paper mill load operation feature extraction method, the maximum up-regulation capacity, the maximum down-regulation capacity and the future green power factor sequence are extracted, and the maximum up-regulation capacity, the maximum down-regulation capacity and the future green power factor sequence are calculated. Improving the global perception capability of long-sequence load data by using a sparse attention mechanism, and accurately extracting the running rhythm, state change and key load mutation characteristics of each device; and on the basis of a green power factor prediction model of FEDformer, time sequence decomposition and frequency domain modeling capabilities are fused, and high-precision prediction of a future green power ratio factor is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system regulation and control, and specifically relates to a method for evaluating the green electricity consumption potential of a paper mill load by integrating ProbSparse and FEDformer. Background Art

[0002] Assessing the green electricity absorption potential of paper mill loads helps to accurately identify production links with the ability to absorb green electricity and promote the efficient matching of green electricity with high-energy-consuming equipment. By optimizing the energy consumption structure and improving the utilization rate of green electricity, it can not only reduce the carbon emission intensity, but also support the green transformation and carbon asset management of enterprises, and help achieve the "dual carbon" goals. However, the existing methods for assessing the green electricity absorption potential of paper mill loads have the following shortcomings: 1) The extraction of load operation characteristics is rough and lacks global time series perception capabilities. Traditional methods are usually based on fixed window statistics or single-step predictions, which make it difficult to accurately capture the dynamic behavior of various loads in paper mills (such as pulping, refining, drying, etc.) on the time axis, and ignore the complex relationship between production rhythm, start-stop frequency and process sequence, resulting in inaccurate characterization of operation characteristics, which makes it difficult to support subsequent absorption capacity modeling. 2) The prediction of green electricity factors does not reflect the multi-scale climate and load coupling characteristics. Most existing methods are based on simple regression or short-term trend extrapolation, lacking the ability to integrate modeling between the time variability of wind and solar power output and the load operation rhythm, and are unable to fully utilize multi-time scale information (such as daily cycles and weekly cycles), resulting in inaccurate identification of green power utilization windows and affecting the effectiveness of the assessment. 3) The assessment of green power consumption potential lacks contextual association and strategy evaluation mechanisms. Most methods statically quantify the load consumption potential, failing to fully consider the linkage between the time-series green power matching, process adjustability, and green power fluctuations. It is difficult to dynamically reflect the actual absorbable capacity, and the assessment results are conservative or distorted, which cannot guide refined scheduling and optimization decisions. Therefore, it is necessary to develop a new paper mill load green power consumption potential assessment method that integrates ProbSparse and FEDformer to solve the existing problems. Summary of the Invention

[0003] The purpose of the present invention is to provide a paper mill load green electricity consumption potential assessment method that integrates ProbSparse and FEDformer to solve the problem of poor accuracy in load green electricity consumption potential assessment.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for evaluating the green electricity consumption potential of paper mill loads by integrating ProbSparse and FEDformer, comprising:

[0005] Construct an upward adjustment capacity assessment model and a downward adjustment capacity assessment model to obtain the maximum upward adjustment capacity and the maximum downward adjustment capacity;

[0006] Based on time-frequency domain decoupling modeling, predict future green power factor series;

[0007] Based on the maximum adjustable capacity, the maximum adjustable capacity and the future green electricity factor sequence, a load green electricity absorption potential assessment model is constructed to dynamically calculate the amount of green electricity that the paper mill can absorb in different time periods, and define the unit load green electricity absorption potential index.

[0008] Preferably, the constructing of the upward adjustment capability assessment model and the downward adjustment capability assessment model includes:

[0009] Construct a multi-dimensional time series feature vector of key operation information, as shown in formulas (1)-(2);

[0010] X(t)=[P BJ (t),P MJ (t),P YZ (t),P HG (t),T(t),H(t),D(t),E(t)] (1)

[0011] χ t =[X(t-L+1),...,X(t)] (2)

[0012] Where: X(t) is the input feature vector at time step t, P BJ (t) is the actual load power of the beater at time t (kW), P MJ (t) is the refiner load, P YZ (t) is the press load, P HG (t) is the dryer load, T(t) is the workshop temperature (℃), H(t) is the workshop humidity (%), D(t) is the capacity load ratio, which is equal to the current output divided by the design capacity, E(t) is the unit electricity price at time t (yuan / kWh), χ t is the historical feature sequence ending at the current time t, and L is the length of the time window;

[0013] ProbSparse Transformer is used to extract the key attention time slices in the paper mill load sequence, retain the state information with high correlation with the regulation capability, and generate the regulation potential factor vectors of various load devices; as shown in formula (3) and formula (4);

[0014] Z t =ProbSparseEncoder(χ t ) (3)

[0015] z(t)=[z BJ (t),z MJ (t),z YZ (t),Z HG (t)] (4)

[0016] Where: Z t is the temporal embedding representation extracted by the attention module, Z t is the historical feature sequence ending at the current time t, L is the time window length, d is the embedding dimension, z(t) is the adjustment potential factor vector of each device at the current time t, z i (t) is the conditioning potential embedding of the i-th device at time t; the attention module is a submodule in the MLP aggregator;

[0017] Based on the extracted adjustment factors, an upward adjustment capacity evaluation model and a downward adjustment capacity estimation model are constructed to evaluate the maximum load increase and maximum load reduction capabilities of each type of key equipment under the current operating state. Combined with the rated operating range of the equipment and the adjustment response coefficient, the load adjustable boundary value that can be used for green power response or low-carbon scheduling is formed, as shown in formulas (5)-(7);

[0018] ΔP i up (t) = min{P i max -P i (t),γ i ·σ(z i (t))} (5)

[0019] ΔP i down (t) = min{P i (t)-P i min ,η i ·σ(z i (t))} (6)

[0020]

[0021] Where: ΔP i up (t) is the maximum adjustable capacity of the i-th type equipment at time t (kW), P i max is the rated maximum power upper limit of device i (kW), P i (t) is the current actual load power of device i, γ i The equipment adjustment response capability coefficient, σ(z i (t)) is the normalized output of the Sigmoid function for the regulatory potential factor, ΔP i down (t) is the maximum load that can be reduced for equipment i, P i min is the minimum safe operating power of the equipment (kW), η i Lower the responsiveness factor for the device.

[0022] Preferably, the prediction of future green power factor sequences based on time-frequency domain decoupling modeling includes:

[0023] Construct a multi-dimensional time series input tensor including total load, distributed photovoltaic output, energy storage discharge power, energy storage charge state, and meteorological data elements, as shown in formula (8);

[0024]

[0025] Where: y t is the input sequence tensor at time t, containing L time steps of data, P load (t) is the total load power of the paper mill (kW), P res (t) is the distributed renewable energy output within the plant (kW), is the energy storage system discharge power (kW), SOC(t) is the state of charge of the energy storage system, W(t) is a meteorological factor such as light intensity or wind speed, T(t) is the ambient temperature (°C), L is the time window length, and d is the characteristic dimension of each time step;

[0026] Based on the FEDformer architecture of frequency domain enhanced decoupling modeling, the multi-source power data of the paper mill is processed in a trend-cycle dual-channel manner, as shown in formula (9);

[0027]

[0028] Where: is the green power factor prediction sequence for the next H time steps, FEDformer(·) is the frequency domain enhanced decomposition Transformer model, and H is the prediction duration;

[0029] Define the green power factor of the paper mill load, that is, the proportion of distributed renewable energy and energy storage provided per kilowatt-hour per unit time, as shown in formulas (10)-(11);

[0030]

[0031] Where: GPF(t) is the green power factor, which represents the proportion of green power in each kilowatt-hour. res (t) is the output of distributed renewable energy at time t (kW), is the discharge power of the energy storage system at time t (kW), is the total load of the paper mill at time t (kW), Green electricity supply capacity for paper mills, P load (t) is the total load power of the paper mill (kW).

[0032] Preferably, based on the maximum adjustable capacity, the maximum adjustable capacity and the future green electricity factor sequence, a load green electricity absorption potential assessment model is constructed to dynamically calculate the amount of green electricity that can be absorbed by the paper mill in different time periods, and define the unit load green electricity absorption potential index including:

[0033] Calculate the basic absorption capacity of green electricity;

[0034] Based on the maximum green electricity absorption capacity of load regulation, determine the actual total amount of green electricity that can be absorbed;

[0035] Define green electricity consumption potential indicators;

[0036] Time series evaluation indicator analysis and trend judgment.

[0037] Preferably, the calculation of the green electricity basic absorption capacity includes: multiplying the green power factor (GPF) predicted by the FEDformer model by the current total load to obtain the basic green power that can be absorbed at each moment, as shown in formula (12);

[0038]

[0039] Where: is the green power that the paper mill can absorb without adjusting the load at any time (kW), GPF(t) is the green power factor at time t, P load (t) is the total load power of the paper mill (kW).

[0040] Preferably, the maximum green electricity absorption capacity based on load regulation to determine the actual total amount of green electricity that can be absorbed includes:

[0041] The upward adjustment capacity is superimposed on the basic green electricity absorption capacity to construct the maximum green electricity boundary that the paper mill can theoretically absorb, and compared with the actual available green electricity power to determine the actual total amount of green electricity that can be absorbed, as shown in formulas (13)-(14);

[0042]

[0043] Where: is the maximum green power absorption power (kW) that can be achieved by increasing the load at the paper mill at time t, ΔP up (t) is the total adjustable load capacity (kW), is the green power that the paper mill can actually absorb at time t (kW), is the current green electricity supply capacity (kW), including the output of distributed renewable energy and energy storage discharge power. For the actual amount of green electricity that can be absorbed in the end.

[0044] Preferably, the green electricity consumption potential index is defined as follows:

[0045] As shown in formula (15);

[0046]

[0047] Where: GPI(t) is the green power consumption potential index at time t, is the amount of green electricity absorbed by the current load regulation, P load (t) is the current total load of the paper mill, which is used as the normalized basis.

[0048] Preferably, the time series evaluation indicator analysis and trend judgment include:

[0049] The GPI is accumulated and visualized in the time dimension to form the daily total green electricity regulation absorption and time series trend curve indicators, as shown in formula (16);

[0050]

[0051] Where: GPI total The total amount of green electricity (kWh) and GPI absorbed by adjusting the whole cycle total is the duration of each time step, T is the total number of time steps, is the green power absorbed by regulation at time t (kW).

[0052] The technical effects and advantages of the present invention are as follows: the paper mill load green electricity consumption potential assessment method that integrates ProbSparse and FEDformer, the paper mill load operation feature extraction method based on ProbSparse, uses its sparse attention mechanism to improve the global perception ability of long-sequence load data, and accurately extracts the operation rhythm, state changes and key load mutation characteristics of each device; the green electricity factor prediction model based on FEDformer integrates time series decomposition and frequency domain modeling capabilities to achieve high-precision prediction of future green electricity proportion factors, providing continuous and reliable data input for subsequent consumption analysis; constructs a paper mill load green electricity consumption potential assessment method, combines the load increase capacity, production adjustable links and green electricity proportion change trends, and dynamically calculates the actual amount of green electricity that can be absorbed in different time periods, providing a basis for achieving accurate matching and formulating green electricity priority consumption strategies;

[0053] This solution accurately depicts the load operation characteristics and green electricity utilization capacity of the paper mill, improving the timeliness and accuracy of the green electricity absorption assessment; through ProbSparse to extract key load change characteristics, combined with FEDformer's high-precision prediction of green electricity factors, it achieves time-series matching of load and green electricity output, and constructs a dynamic assessment model to provide decision support for the paper mill to formulate green electricity priority utilization strategies and carbon emission reduction paths; and realizes high-precision dynamic assessment of the paper mill's ability to absorb green electricity under different operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] The present invention provides Figure 1 A paper mill load green power consumption potential assessment method integrating ProbSparse and FEDformer is shown in the paper mill load green power consumption potential assessment method, including:

[0057] First, based on the ProbSparse load operation feature extraction method, a sparse attention mechanism is introduced for key processes such as pulping, refining, pressing, and drying to identify time segments in historical load data that are highly correlated with regulation capabilities, thereby extracting the up and down regulation capabilities of various types of equipment in the current state, providing load boundaries for subsequent green power scheduling; secondly, based on the FEDformer green power factor prediction method, multi-source data such as external calls, distributed photovoltaics, energy storage output, and meteorological data are integrated, and through time-frequency domain decoupling modeling, the proportion of green power contained in each kilowatt-hour of paper mill electricity at each moment in the future is predicted, providing a basis for evaluating the availability of green power on the energy supply side; finally, the outputs of the above two models are integrated to construct a paper mill load green power absorption potential assessment model, using a "basic absorption + adjustable expansion" two-layer structure to dynamically calculate the amount of green power that the paper mill can absorb in different time periods, and define a unit load green power absorption potential indicator; this method can provide quantitative support for enterprises to formulate green power priority operation strategies, optimize production scheduling plans, participate in carbon market transactions, etc., and promote green transformation and low-carbon development in the paper industry. Specifically, it includes the following steps:

[0058] Step 1. Propose a method for extracting the load operation characteristics of a paper mill based on ProbSparse, extract the adjustable characteristics of the paper mill load, and provide an upward adjustment capacity evaluation model and a downward adjustment capacity evaluation model. This step focuses on the load operation characteristics of key processes within the paper mill (such as pulping, refining, pressing, and drying). By introducing the ProbSparse attention mechanism, a Transformer model is constructed to identify historical state information that is strongly correlated with load changes in key time segments. Multidimensional time series including equipment load, workshop environment, electricity price, capacity utilization, etc. are input into the model to extract the embedded factors of the adjustment potential of various types of equipment at the current moment. Then, an upward adjustment capacity evaluation model and a downward adjustment capacity evaluation model are constructed to estimate the maximum adjustable upward capacity and the maximum adjustable downward capacity of each type of equipment under the current operating state, respectively, to provide load-side adjustment boundary conditions for subsequent green electricity response adjustment.

[0059] Step 2: The FEDformer-based method for predicting the green electricity factor of the paper mill load comprehensively considers the output of external electricity, internal distributed new energy, and internal distributed energy storage to predict the proportion of green electricity contained in each kilowatt-hour of the paper mill load; this step focuses on predicting the proportion of green electricity contained in each kilowatt-hour of the paper mill at each future moment, that is, the "green electricity factor". In order to achieve accurate prediction, the FEDformer model based on time-frequency domain decoupling is introduced to fully explore the long-term trends and short-term periodicity of multi-source data such as paper mill load, grid power supply, distributed photovoltaic output, energy storage system output, charge status, and meteorological conditions. The model will predict a future green electricity factor sequence, reflecting the distribution of the proportion of green electricity that the paper mill can absorb under given operating and meteorological conditions. This result provides a basis on the energy supply side for the subsequent evaluation of whether the adjustable load of the paper mill has sufficient green electricity response value;

[0060] Step 3: Propose a method for evaluating the green electricity consumption potential of paper mill loads. This step integrates the results of the first two steps to construct a model for evaluating the green electricity consumption potential of paper mill loads. The model uses the upward adjustment capacity output of the adjustable capacity assessment as the capacity boundary for the paper mill to increase electricity consumption to absorb green electricity, and uses the green electricity factor prediction results and the actual available green electricity output as the supply-side upper limit to establish a two-layer absorption capacity structure of "basic absorption + adjustable expansion". By comparing the actual green electricity available with the maximum absorption value supported by the paper mill's upward and downward adjustment capacity, the actual green electricity absorption capacity of the paper mill at each moment is dynamically calculated, and the unit load green electricity consumption potential indicator is further defined. This model can provide quantifiable data support for paper mills to formulate green electricity priority operation strategies and participate in carbon trading.

[0061] For step 1, the paper mill load operation feature extraction method based on ProbSparse is shown in formulas (1)-(7);

[0062] By extracting the adjustable characteristics of the paper mill load, an upward adjustment capacity evaluation model and a downward adjustment capacity evaluation model are proposed, as follows:

[0063] 1. Multi-process load state modeling;

[0064] This step focuses on the typical load links in the paper mill (such as pulping, refining, pressing, and drying), and constructs a multi-dimensional time series feature vector covering key operating information such as equipment power, workshop temperature and humidity, electricity price, and capacity utilization, reflecting the dynamic energy consumption status of the paper mill under the complex production process. This vector serves as the basis for model input and provides high-quality time series feature support for subsequent adjustable capacity extraction, as shown in formulas (1)-(2). Formula (1) is for constructing a feature vector (including environmental and production information). Formula (2) is for constructing an input sequence with a sliding window length of L;

[0065] 2. ProbSparseAttention extracts regulatory feature factors;

[0066] ProbSparse Transformer is used to extract the key attention time slices in the paper mill load sequence, retain the state information with high correlation with the regulation capability, and generate the regulation potential factor vectors of various load devices. This process not only improves the model's perception of sparse key moments, but also effectively compresses redundant inputs. The output results provide core feature support for device-level regulation capability modeling as shown in formulas (3)-(4). Formula (3) is to input the input sequence into ProbSparseTransformer for training and extract the regulation potential factors of each load in the paper mill. Formula (4) is based on the extracted regulation factors of each load in the paper mill, and the regulation embedding vector is obtained through the MLP aggregator;

[0067] 3. Upward adjustment of the capability assessment model and downward adjustment of the capability assessment model;

[0068] Based on the adjustment factors extracted in the previous stage, an upward adjustment capacity evaluation model and a downward adjustment capacity estimation model are constructed to evaluate the maximum load increase and maximum load reduction capabilities of each type of key equipment under the current operating state. Combined with the rated operating range of the equipment and the adjustment response coefficient, the load adjustable boundary value that can be used for green power response or low-carbon scheduling is formed, which is an important prerequisite for absorbing green power and demand response control as shown in formulas (5)-(7); Formula (5) is a ProbSparse-based evaluation model for the upward adjustment capacity of each equipment in the paper mill load. Formula (6) is a ProbSparse-based evaluation model for the downward adjustment capacity of each equipment in the paper mill load. Formula (7) is a ProbSparse-based evaluation model for the total downward adjustment capacity of the paper mill load.

[0069] X(t)=[P BJ (t),P MJ(t),P YZ (t),P HG (t),T(t),H(t),D(t),E(t)] (1)

[0070] χ t =[X(t-L+1),...,X(t)] (2)

[0071] Z t =ProbSparseEncoder(χ t ) (3)

[0072] z(t)=[z BJ (t),z MJ (t),z YZ (t),Z HG (t)] (4)

[0073] ΔP i up (t) = min{P i max -P i (t),γ i ·σ(z i (t))} (5)

[0074] ΔP i down (t) = min{P i (t)-P i min ,η i ·σ(z i (t))} (6)

[0075]

[0076] Where: X(t) is the input feature vector at time step t, P BJ (t) is the actual load power of the beater at time t (kW), P MJ (t) is the refiner load, P YZ (t) is the press load, P HG (t) is the dryer load, T(t) is the workshop temperature (℃), H(t) is the workshop humidity (%), D(t) is the capacity load ratio, which is equal to the current output divided by the design capacity (dimensionless), E(t) is the unit electricity price at time t (yuan / kWh), χ t is the historical feature sequence ending at the current time t, L is the time window length (number of steps), Z t is the temporal embedding representation extracted by the attention module, Z tis the historical feature sequence ending at the current time t, d is the embedding dimension, z(t) is the adjustment potential factor vector of each device at the current time t, z i (t) is the regulation potential embedding (normalized value) of the i-th device (such as BJ, MJ, etc.) at time t, ΔP i up (t) is the maximum adjustable capacity of the i-th type equipment at time t (kW), P i max is the rated maximum power upper limit of device i (kW), P i (t) is the current actual load power of device i, γ i Adjust the response capacity coefficient (experience value or data training result) and σ(z i (t)) is the normalized output of the Sigmoid function for the regulation potential factor, ranging from (0,1), ΔP i down (t) is the maximum load that can be reduced for equipment i, P i min is the minimum safe operating power of the equipment (kW), η i Lower the responsiveness factor for the device.

[0077] For step 2, the green electricity factor prediction method of paper mill load based on FEDformer is shown in formulas (8)-(11);

[0078] Taking into account the output of external electricity, internal distributed renewable energy, and internal distributed energy storage, we predict the proportion of green electricity contained in each kilowatt-hour of the paper mill's load; the details are as follows:

[0079] 1. Construct the input sequence of the paper mill load green electricity factor prediction model;

[0080] This step is based on the characteristics of the paper mill power system. A multidimensional time series input tensor is constructed, which includes elements such as total load, distributed photovoltaic output, energy storage discharge power, energy storage charge state, and meteorological data. This input fully reflects the dynamic relationship between the load demand and green electricity supply within the plant, laying a data foundation for the subsequent use of FEDformer to predict the green electricity factor, ensuring that the model can perceive the temporal evolution characteristics of power supply and consumption, as shown in formula (8). Formula (8) is used to construct the time series tensor.

[0081] 2. FEDformer modeling process;

[0082] This step introduces the FEDformer architecture of frequency domain enhanced decoupling modeling to perform trend-cycle dual-channel processing on the multi-source power data of the paper mill; the model uses wavelet decomposition to extract long-term changes and short-term periodicity, enhances the time-frequency recognition ability of the load-green electricity supply and demand relationship, and thus achieves high-precision prediction of the green electricity factor (the proportion of green electricity in each kilowatt-hour) in multiple time steps in the future. As shown in formula (9), formula (9) is based on the encoder output of the future green electricity factor sequence'

[0083] 3. Definition of green electricity ratio per kilowatt-hour

[0084] This step defines the green power factor (GPF) of the paper mill load, which is the proportion of distributed renewable energy and energy storage in each kilowatt-hour of electricity per unit time. By dividing the sum of photovoltaic and energy storage output by the current total load, the actual proportion of green electricity in production electricity consumption is quantified. This indicator reflects the level of green electricity utilization of the factory under the current scheduling conditions and is an important parameter for subsequent green electricity scheduling and carbon emission reduction assessment, as shown in formulas (10) and (11). Formula (10) is the green power factor of the paper mill per kilowatt-hour. Formula (11) is the predicted green electricity supply capacity.

[0085]

[0086] Where: y t is the input sequence tensor at time t, containing L time steps of data, P load (t) is the total load power of the paper mill (kW), P res (t) is the output of distributed renewable energy (such as photovoltaic) in the factory (kW), is the energy storage system discharge power (kW), SOC(t) is the energy storage system state of charge (range 0–1), W(t) is the meteorological factor such as light intensity or wind speed (related to P res correlation), T(t) is the ambient temperature (℃), L is the time window length, d is the feature dimension of each time step (such as 6 for 6 variables), is the green power factor prediction sequence for the next H time steps, FEDformer(·) is the frequency domain enhanced decomposition Transformer model, H is the prediction duration (number of future time steps), GPF(t) is the green power factor, which represents the proportion of green power in each kilowatt-hour (0-1), P res (t) is the output of distributed renewable energy at time t (kW), is the discharge power of the energy storage system at time t (kW), is the total load of the paper mill at time t (kW), Green electricity supply capacity for paper mills;

[0087] For step 3, the green electricity consumption potential assessment method for the paper mill load is shown in formulas (12)-(16). Based on the results of steps (1) and (2), the green electricity consumption potential assessment model for the paper mill load is constructed as follows:

[0088] 1. Calculate the basic green electricity absorption capacity;

[0089] This step is used to evaluate the green electricity capacity that the paper mill can naturally absorb without adjusting the load. The green power factor (GPF) predicted by the FEDformer model is multiplied by the current total load to obtain the basic green electricity power that can be absorbed at each moment. This indicator reflects the "green electricity matching degree" under the current operating conditions and is the basis for subsequent judgment on whether there is room for adjustment and implementation of green electricity priority operation. As shown in formula (12), formula (12) is the basic green electricity absorption capacity under the current operating state of the factory;

[0090] 2. Consider the maximum green power absorption capacity of load regulation;

[0091] On the basis of natural absorption, we further consider the "extra green electricity absorption capacity" achieved by increasing the load. This step superimposes the upward adjustment capacity evaluated in step 1 with the basic green electricity absorption capacity to construct the maximum green electricity boundary that the paper mill can theoretically absorb, and compares it with the actual green electricity power available. Finally, the actual total amount of green electricity that can be absorbed is determined, which is an important basis for promoting the demand side to actively respond to green electricity. As shown in formulas (13) and (14); formula (13) is the potential green electricity absorption space introduced by increasing the capacity; formula (14) is the final actual amount of green electricity that can be absorbed;

[0092] 3. Define the Green Power Consumption Potential Index (GPI);

[0093] In order to measure the effect of load regulation on the green electricity absorption capacity, the green electricity absorption potential index (GPI) is defined, that is, the proportion of newly absorbed green electricity due to the current load increase in the total load. The higher the GPI, the stronger the current flexible green electricity absorption capacity is. It helps to identify high-potential periods, plan regulation windows and achieve low-carbon optimized operation. It is an important reference indicator for green electricity response scheduling as shown in formula (15); formula (15) is the green electricity regulation absorption potential per unit load'

[0094] 4. Time series evaluation indicators and trend judgment;

[0095] This step accumulates and visualizes the GPI over time, generating indicators such as daily green power regulation and absorption and a time-series trend curve. These indicators are used to assess the plant's green power response capacity over different time periods. By analyzing dynamic trends, regulation bottlenecks can be identified, along with periods of low green power absorption risk. This provides support for subsequent regulation strategy design, green power time-sharing priority access, and the achievement of carbon reduction targets, as shown in Formula (16). Formula (16) represents the daily green power absorption potential.

[0096]

[0097]

[0098] Where: is the green power that the paper mill can absorb without adjusting the load at any time (kW), GPF(t) is the green power factor at time t (predicted by FEDformer), P load (t) is the total load power of the paper mill (kW), is the maximum green power absorption power (kW) that can be achieved by increasing the load at the paper mill at time t, ΔP up (t) is the total adjustable load capacity (kW), is the green power that the paper mill can actually absorb at time t (kW), is the green power supply capacity (kW) at that time, including the output of distributed renewable energy and energy storage discharge power, GPI(t) is the green power consumption potential index at time t (range [0,1]), is the amount of green electricity absorbed by the current load regulation, P load (t) is the current total load of the paper mill, which is used as the normalized benchmark, GPI total is the total amount of green electricity absorbed by regulation in the entire cycle (such as 1 day, 1 hour) (kWh), T is the total number of time steps, GPI total The duration of each time step (in hours or minutes, such as 0.25 for 15 minutes), is the green power absorbed by regulation at time t (kW).

[0099] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

[0100] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A paper mill load green power consumption potential assessment method integrating ProbSparse and FEDformer, characterized by: include: Construct an upward adjustment capacity assessment model and a downward adjustment capacity assessment model to obtain the maximum upward adjustment capacity and the maximum downward adjustment capacity; Based on time-frequency domain decoupling modeling, predict future green power factor series; Based on the maximum adjustable capacity, the maximum adjustable capacity and the future green electricity factor sequence, a load green electricity absorption potential assessment model is constructed to dynamically calculate the amount of green electricity that the paper mill can absorb in different time periods, and define the unit load green electricity absorption potential index.

2. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 1 is characterized by: The constructing of the upward adjustment capability assessment model and the downward adjustment capability assessment model includes: Construct a multi-dimensional time series feature vector of key operation information, as shown in formulas (1)-(2); X(t)=[P BJ (t),P MJ (t),P YZ (t),P HG (t),T(t),H(t),D(t),E(t)] (1) χ t =[X(t-L+1),...,X(t)] (2) Where: X(t) is the input feature vector at time step t, P BJ (t) is the actual load power of the beater at time t (kW), P MJ (t) is the refiner load, P YZ (t) is the press load, P HG (t) is the dryer load, T(t) is the workshop temperature (℃), H(t) is the workshop humidity (%), D(t) is the capacity load ratio, which is equal to the current output divided by the design capacity, E(t) is the unit electricity price at time t (yuan / kWh), χ t is the historical feature sequence ending at the current time t, and L is the length of the time window; ProbSparse Transformer is used to extract the key attention time slices in the paper mill load sequence, retain the state information with high correlation with the regulation capability, and generate the regulation potential factor vectors of various load devices; as shown in formula (3) and formula (4); Z t =ProbSparseEncoder(χ t ) (3) z(t)=[z BJ (t),z MJ (t),z YZ (t),z HG (t)] (4) Where: Z t is the time series embedding, z(t) is the adjustment potential factor vector of each device at the current time t, z i (t) is the regulation potential embedding of the i-th device at time t; Based on the extracted adjustment factors, an upward adjustment capacity evaluation model and a downward adjustment capacity estimation model are constructed to evaluate the maximum load increase and maximum load reduction capabilities of each type of key equipment under the current operating state. Combined with the rated operating range of the equipment and the adjustment response coefficient, the load adjustable boundary value that can be used for green power response or low-carbon scheduling is formed, as shown in formulas (5)-(7); ΔP i up (t)=min{P i max -P i (t),γ i ·σ(z i (t))} (5) ΔP i down (t)=min{P i (t)-P i min ,or i ·σ(z i (t))} (6) Where: ΔP i up (t) is the maximum adjustable capacity of the i-th type equipment at time t (kW), P i max is the rated maximum power upper limit of device i (kW), P i (t) is the current actual load power of device i, γ i The equipment adjustment response capability coefficient, σ(z i (t)) is the normalized output of the Sigmoid function for the regulatory potential factor, ΔP i down (t) is the maximum load that can be reduced for equipment i, P i min is the minimum safe operating power of the equipment (kW), η i Lower the responsiveness factor for the device.

3. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 1 is characterized by: The prediction of future green power factor sequences based on time-frequency domain decoupling modeling includes: Construct a multi-dimensional time series input tensor including total load, distributed photovoltaic output, energy storage discharge power, energy storage charge state, and meteorological data elements, as shown in formula (8); Where: y t is the input sequence tensor at time t, containing L time steps of data, P load (t) is the total load power of the paper mill (kW), P res (t) is the distributed renewable energy output within the plant (kW), is the energy storage system discharge power (kW), SOC(t) is the state of charge of the energy storage system, W(t) is a meteorological factor such as light intensity or wind speed, T(t) is the ambient temperature (°C), and L is the time window length; Based on the FEDformer architecture of frequency domain enhanced decoupling modeling, the multi-source power data of the paper mill is processed in a trend-cycle dual-channel manner, as shown in formula (9); Where: is the green power factor prediction sequence for the next H time steps, FEDformer(·) is the frequency domain enhanced decomposition Transformer model, and H is the prediction duration; Define the green power factor of the paper mill load, that is, the proportion of distributed renewable energy and energy storage provided per kilowatt-hour per unit time, as shown in formulas (10)-(11); Where: GPF(t) is the green power factor, which represents the proportion of green power in each kilowatt-hour. res (t) is the output of distributed renewable energy at time t (kW), is the discharge power of the energy storage system at time t (kW), is the total load of the paper mill at time t (kW), Green electricity supply capacity for paper mills, P load (t) is the total load power of the paper mill (kW).

4. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 1 is characterized by: Based on the maximum adjustable capacity, the maximum adjustable capacity, and the future green electricity factor sequence, a load green electricity absorption potential assessment model is constructed to dynamically calculate the amount of green electricity that the paper mill can absorb in different time periods. The unit load green electricity absorption potential indicators are defined as follows: Calculate the basic absorption capacity of green electricity; Based on the maximum green electricity absorption capacity of load regulation, determine the actual total amount of green electricity that can be absorbed; Define green electricity consumption potential indicators; Time series evaluation indicator analysis and trend judgment.

5. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 4 is characterized by: The calculation of the basic green electricity absorption capacity includes: multiplying the green power factor (GPF) predicted by the FEDformer model by the current total load to obtain the basic green power that can be absorbed at each moment, as shown in formula (12); Where: is the green power that the paper mill can absorb without adjusting the load at any time (kW), GPF(t) is the green power factor at time t, P load (t) is the total load power of the paper mill (kW).

6. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 4 is characterized by: The maximum green electricity absorption capacity based on load regulation, and the determination of the actual total amount of green electricity that can be absorbed include: The upward adjustment capacity is superimposed on the basic green electricity absorption capacity to construct the maximum green electricity boundary that the paper mill can theoretically absorb, and compared with the actual available green electricity power to determine the actual total amount of green electricity that can be absorbed, as shown in formulas (13)-(14); Where: is the maximum green power absorption power (kW) that can be achieved by increasing the load at the paper mill at time t, ΔP up (t) is the total adjustable load capacity (kW), is the green power that the paper mill can actually absorb at time t (kW), is the current green electricity supply capacity (kW), including the output of distributed renewable energy and energy storage discharge power. For the actual amount of green electricity that can be absorbed in the end.

7. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 4 is characterized by: The green electricity consumption potential indicators defined include: As shown in formula (15); Where: GPI(t) is the green power consumption potential index at time t, is the amount of green electricity absorbed by the current load regulation, P load (t) is the current total load of the paper mill.

8. The method for evaluating the green power consumption potential of a paper mill load by integrating ProbSparse and FEDformer according to claim 4 is characterized by: The time series evaluation indicator analysis and trend judgment include: The GPI is accumulated and visualized in the time dimension to form the daily total green electricity regulation absorption and time series trend curve indicators, as shown in formula (16); Where: GPI total is the total amount of green electricity absorbed through regulation during the entire cycle (kWh), T is the total number of time steps, is the green power absorbed by regulation at time t (kW).

Images