A heterogeneous energy collaborative prediction and scheduling method considering complementary and matching of source load low carbon characteristics

Abstract

The application discloses a heterogeneous energy collaborative prediction and scheduling method considering source-load low-carbon characteristics complementation and matching, relates to the technical field of power system source-load prediction and low-carbon scheduling, and comprises the following steps: based on Laguerre polynomials, pseudo-inverse learning and integrated learning, robust local mean decomposition (RLMD) and weighted permutation entropy (WPE), a wind power (WP), photovoltaic power (PVP) and power load (PL) prediction model is constructed, which takes into account the prediction accuracy and stability; a HECM mathematical model is constructed, the HECM mathematical model framework comprises an integrated flexible operation mode of carbon capture power plant (IFCCPP), a wind farm (WF), a photovoltaic power station (PP), a concentrated solar power (CSP), a load active demand response (LADR), a ladder carbon trading mechanism (LCTM), an objective function and related constraints; in the source-load prediction, the two targets of accuracy and stability are considered simultaneously, a compromise solution is selected in the Pareto frontier through the MORUN algorithm, and the wind power, photovoltaic power and power load prediction results are more suitable for the power system scheduling scene with high robustness requirements.

Description

Technical Field

[0001] This invention relates to the field of power system source-load prediction and low-carbon dispatch technology, specifically to a heterogeneous energy collaborative prediction and dispatch method that considers the complementarity and matching of low-carbon characteristics of source and load. Background Technology

[0002] With the further development of global energy conservation, emission reduction, and energy transition, the efficient utilization of renewable energy has received widespread attention. Among these efforts, vigorously developing heterogeneous energy generation is an effective way to achieve energy transition and pursue a sustainable development path. As the main energy consumer, the power industry accounts for a large proportion of total carbon emissions. Increasing the proportion of heterogeneous renewable energy and carrying out low-carbon transformation on both the power source and load sides are important means to achieve a low-carbon transformation of the power system.

[0003] Currently, my country is vigorously developing technologies related to the decarbonization of its power system, but due to the uncertain nature of renewable energy, the level of renewable energy consumption is not high. Current technologies also have the following shortcomings and deficiencies: 1. While there is considerable research on load-side PDR combined with source-side shunt carbon capture power plants, the more effective IFCCPP (Integrated Intermediate Carbon Capture Program) has not been adopted on the source side, nor has the CHES (Carbon Utilization and Storage) low-carbon power supply mode (combining wind, solar, and concentrated solar power, CSP) been implemented. The low-carbon performance of power systems incorporating IFCCPP, CHES, and DR needs further exploration. Source-side IFCCPP, CHES, and DR are all low-carbon measures, but each has certain limitations. Therefore, it is difficult to further improve the system's low-carbon performance based on the shortcomings of these three low-carbon measures. Furthermore, the complementary scheduling advantages of the low-carbon characteristics of the source and load methods have not been fully explored, and research on the operational mechanism of their complementary cooperation to achieve low-carbon operation is lacking.

[0004] 2. In the process of source-load coordinated CHES low-carbon scheduling, there is a lack of research on key scheduling technologies that simultaneously address scheduling costs and source-load matching, resulting in a failure to balance the system's economic costs and security.

[0005] 3. Accurate source-load forecasting results are crucial for dispatching. Existing technologies have limited research on how to obtain forecasting data for renewable energy and load in source-load dispatching, which affects the source-load forecasting results.

[0006] 4. Although intelligent optimization algorithms have been widely used in the field of scheduling, scheduling problems that consider uncertainty often exhibit complex coupling characteristics, making it difficult for traditional optimization algorithms to find feasible solutions.

[0007] Therefore, a heterogeneous energy collaborative prediction and scheduling method that considers the complementary and matching low-carbon characteristics of source and load is needed to solve the above problems. Summary of the Invention

[0008] The purpose of this invention is to provide a heterogeneous energy collaborative prediction and scheduling method that considers the complementary and matching low-carbon characteristics of source and load, so as to solve the problems existing in the prior art mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: A heterogeneous energy collaborative prediction and scheduling method that considers the complementarity and matching of low-carbon characteristics of source and load includes the following steps: S1: Based on Laguerre polynomials, pseudo-inverse learning and ensemble learning, robust local mean decomposition (RLMD), and weighted permutation entropy (WPE), a prediction model for wind power (WP), photovoltaic power (PVP), and electricity load (PL) is constructed that balances prediction accuracy and prediction stability. S2: Construct a HECM (a heterogeneous energy collaborative forecasting and scheduling method considering source-load low carbon complementarity and matching) mathematical model. The HECM mathematical model framework includes an integrated flexible operation mode of carbon capture power plant (IFCCPP), wind farm (WF), photovoltaic power plant (PP), concentrated solar power (CSP), logistic active demand response (LADR), ladder-type carbon trading mechanism (LCTM), objective function, and related constraints. The load of the HECM mathematical model is jointly borne by high-carbon emission coal-fired power plant (CFPP), low-carbon emission IFCCPP, and zero-carbon emission complementary power supply from heterogeneous energy sources (CHES). By changing the user's electricity consumption behavior through LADR, peak-valley optimization of PL is performed, and LCTM is introduced to control system carbon emissions and reduce overall economic costs. S3: Introduce the Source-load difference index (SLDI) to measure the degree of matching between power output and grid load, so as to achieve a balance between the economic cost and safety performance of system scheduling results; S4: The HECM is solved using the multi-objective Runge-Kutta algorithm based on a population-based parallel search mechanism (PSMORUN) to obtain the scheduling prediction results.

[0010] Preferably, the modeling process for the wind power (WP), photovoltaic power (PVP), and power load (PL) prediction models is as follows: S11: Robust Local Mean Decomposition Robust local mean decomposition (RLMD) is employed to decompose the time series of wind power, photovoltaic power, and electricity load into high-frequency and low-frequency components to reduce their volatility and improve the accuracy of the prediction model. RLMD optimizes the boundary conditions, signal envelope estimation, and stopping criteria of local mean decomposition, effectively solving its endpoint effects and mode mixing problems. RLMD can adaptively decompose the number of high-frequency and low-frequency components according to the characteristics of the time series. Its core formula is as follows: (1) (2) (3) (4) (5) In the formula, It is the average of adjacent local extreme points. Local extreme points The corresponding signal value, This is a local envelope estimate. and These are the frequency modulation signal and the envelope signal, respectively. and These are the product function and the residual signal, respectively. Mirror extension of time series signals For component index, For discrete time series indexing, The first PF component A single component, The number of amplitude-modulated signals involved in the synthesis; S12: Weighted permutation entropy When the number of high-frequency and low-frequency components in the RLMD adaptive decomposition is large, the model's prediction efficiency will decrease. Therefore, the weighted permutation entropy (WPE) is used to analyze the complexity of the subsequences after RLMD decomposition, merging subsequences with similar complexity to reduce the model's prediction complexity. The core formula of WPE is as follows: Decomposition of subsequences conduct Phase space reconstruction: (6) In the formula, and These represent the embedding bit length and latency, respectively. , The total length of the original decomposed subsequence; Calculate subsequence WPE value: (7) S13: Hybrid Pseudo-Inverse Laguree Neural Network (1) Hybrid pseudo-inverse Laguerreneural network (HLNN) Orthogonal polynomials are widely used in constructing feedforward neural networks due to their powerful nonlinear approximation capabilities. Wang et al. proposed a novel Laguerre polynomial and combined it with the original Laguerre polynomial to form a hybrid pseudo-inverse Laguerreural network (HLNN), which showed excellent prediction performance in wind power forecasting.

[0011] The orthogonality of the two sets of Laguerre orthogonal polynomials with respect to the weighting function is described as follows: (8) (9) The recurrence equations for the two sets of Laguerre orthogonal polynomials are as follows: (10) (2) Pseudo-reverse learning Pseudo-inverse learning is used to train the Hybrid Laguerre Neural Network (HLNN) and establish the Hybrid Pseudo-Inverse Laguerre Neural Network (HPLNN) model. The specific process is as follows: When the input matrix of the training set is And the output matrix is At that time, the optimal weight matrix of HPLNN The calculation is as follows: (11) in: (12) in It is a matrix Moore Penrose pseudo-retrogression, It is the output of the hidden layer of HPLNN. These are Laguerre orthogonal basis functions; the HPLNN structure is as follows: Figure 7 As shown.

[0012] (3) AdaBoost.R2 To ensure strong stability of the prediction results, AdaBoost.R2 is used to further improve the prediction accuracy of HPLNN. AdaBoost.R2 trains multiple weak predictors, and the final strong predictor is constructed through adaptive correction and recombination of the prediction error weights of these multiple predictors. This assumes the training sample set... , It is the input vector. It is the output vector; the core mathematical formula of Adaboost.R2 is as follows: (13) (14) (15) In the formula, The initial weights of the samples, and These are the weak predictor and the relative error, respectively. The normalization constant is yes The weighted median.

[0013] Preferably, the mathematical model of the IFCCPP is: (16) in, and They are respectively Period IFCCPP Total output power and net output power; and Carbon capture equipment Fixed energy consumption and operating energy consumption; The energy consumption coefficient for capturing one unit of CO2; for Period IFCCPP The amount of CO2 captured; This refers to the maximum operating condition coefficient of the regeneration tower and compressor; , and IFCCPP Carbon capture efficiency, flue gas split ratio, and carbon emission intensity coefficient; For IFCCPP The upper limit of output power, For solution storage exist The amount of CO2 to be captured provided during the time period; for The total amount of CO2 generated by IFCCPP during the period; The formula for converting the amount of CO2 extracted from a solution storage device into solution volume is: (17) in, For IFCCPP solution storage exist The volume of solution required to release CO2 at any given time; and Let be the molar mass of ethanolamine and CO2, respectively; This refers to the regeneration tower's analytical value; and These represent the concentration and density of the alkanolamine solution, respectively.

[0014] Preferably, the CSP mathematical model is as follows: (18) (19) (20) in, Let t be the heat power absorbed by the heat collection device at time t. The thermal power generated directly by the solar collector. To supply the thermal power of the thermal storage system to the thermal collector, Let be the thermal storage power of the thermal storage device at time t. For the heat storage system's charging efficiency, The heat energy transferred from the heat storage device to the heat transfer fluid at time t. Let be the heat release power of the thermal storage device at time t. A product of zero indicates that the thermal storage device cannot store and release heat simultaneously. This indicates the heat release efficiency of the thermal storage system. Let be the electrical power of the solar thermal power plant at time t. This refers to the thermoelectric conversion efficiency of a solar thermal power plant.

[0015] Preferably, the LADR model is: LADR fully considers users' willingness to participate in the response and the smoothness of the response curve changes between different electricity price segments, giving demand response a fuzzy attribute.

[0016] (twenty one) (twenty two) In the formula, Indicates the actual load transfer rate. and These represent the load transfer rates of the pessimistic and optimistic response curves, respectively. These are the weighting coefficients. For the peak-valley electricity price difference, and The nodes are divided into dead zone, response zone, and saturation zone based on the electricity price difference. Using the same method, the actual load transfer rates from peak to flat and from flat to valley were calculated respectively. and The load transfer amount and load value after responding to user demands are calculated as follows: (twenty three) (twenty four) in, , , These represent peak, flat, and trough periods, respectively. , and This represents the average load for each time period before the implementation of peak-valley time-of-use pricing; This represents the amount of load transfer caused by demand response at time t; , These represent the load values ​​at time t before and after the implementation of peak-valley electricity pricing; fuzzy membership functions are used to divide the load into peak, valley, and flat loads. The LADR model is constructed with the objectives of minimizing load fluctuations and maximizing household electricity satisfaction. The LADR optimization objective function is as follows: (25) (26) (27) (28) Where T is the scheduling period. To accurately measure the average load under demand response, and These are respectively user power comfort and economy. and They are respectively Electricity prices are calculated for both the period before and after demand response.

[0017] Preferably, the overall economic cost of the HECM mathematical model is as follows: (29) in, To take into account the overall economic costs, To enable the combined output of high-carbon and low-carbon power units, and It is time Unit fuel cost and valve point effect VPE cost at that time For the overall cost of a solar thermal power plant, and They are time CHES curtailment costs and load shedding costs at that time For carbon trading costs, For the cost of CCPP modification and depreciation, The cost of solvent loss during the carbon capture process; (1) Fuel cost of thermal power units (30) in, , and It is the fuel cost coefficient. It is the sum of low-carbon units and high-carbon units. It is time Combined output power of IFCCPP and IFCCPP; (2) Valve point effect cost (31) in, and It is the VPE cost coefficient; (3) CSP Comprehensive Cost (32) in, This is the CSP configuration cost per unit of heat storage capacity. It is the CSP thermal storage capacity. Environmental benefit coefficient It is CSP in Output power over a time period; (4) CHES curtailment cost When the CHES (Current Equivalent to Spinning Reserve) exceeds the scheduling plan demand, and the system's negative spinning reserve capacity cannot offset this excess power, CHES power curtailment will occur. The mathematical formula for the cost of curtailment is as follows: (33) in, To underestimate the cost coefficient, and They are respectively Real-time system network loss and demand response load, and They are respectively Wind and solar power generation capacity at all times for The negative spinning reserve capacity provided by the combined high-carbon and low-carbon units at all times; (5) System load shedding cost When the actual output of wind-solar hybrid power generation is less than the demand in the dispatch plan, and the system's reserved forward rotation reserve capacity is insufficient to make up for the power shortage, load shedding will occur. The mathematical formula for the load shedding cost of HECM is as follows: (34) in, To prevent the system from overestimating cost coefficients, for The positive-rotation reserve capacity provided by the combined high-carbon and low-carbon units at all times; (6) Costs of CCPP modification and depreciation (35) This includes the depreciation costs of CCPP and solution storage devices. and For FSCCPP depreciation costs and depreciation periods, The discount rate for CCPP projects. and For the volume and depreciation period of the solution storage device, The unit cost of solution storage devices; (7) Solvent loss cost (36) in, and These are the ethanolamine solvent cost coefficient and the solvent operating loss coefficient, respectively. For IFCCPP The amount of CO2 captured; (8) LCTM cost The mathematical formula for tiered carbon trading is as follows: (37) (38) (39) in, and These represent the carbon emissions and carbon emission quotas of thermal power units, respectively. For the number of IFCCPPs, and These represent the quota factor and carbon emission intensity factor of thermal power units, respectively. This represents the total carbon trading cost of the system. This indicates the length of the carbon emission range in carbon trading costs. This indicates the growth rate of carbon trading costs.

[0018] Preferably, the objective function mathematical expression of the Source-Load Difference Index (SLDI) is as follows: (40) This represents the variance of the source load, while , and These correspond to the standard deviations of power supply side output fluctuation, residual load, and curve inflection point, respectively; in addition, , and It is the weighting coefficient.

[0019] Preferably, the constraints of the HECM mathematical model are: (1) Power balance constraint In the HECM model, the source-side outputs are the combined outputs of high-carbon units, IFCCPP units, WF, PP, and CSP units. The load-side includes power system network losses and the power load after LFDR. Its constraints are equality constraints, and the mathematical expressions are as follows: (41) in, The mathematical formula is as follows: (42) In the formula, , and This is the network loss coefficient. and They are respectively Time of the first and the The output of the thermal power unit; (2) Upper and lower limits of unit output constraints (43) In the formula, and They are respectively The upper and lower limits of power generation for PCPP high-carbon units and IFCCPP units at all times; (3) Generator slope limit (44) In the formula, and These are the first in HECM The maximum power of each unit during the rise and fall per unit time; (4) Rotational spare constraint To address CHES uncertainties, this invention employs a combination of IFCCPP and CCPP to provide spinning reserve constraints; the mathematical expression for the spinning reserve capacity is as follows; (45) (46) In the formula, and These are the rated power of wind farms and photovoltaic power plants, respectively. To maximize the output of the solar thermal power plant, The rotational backup response time is, in this invention, Hour; (5) CSP output constraint The power generation capacity of a CSP must be within a certain range, and the mathematical formula is as follows: (47) In the formula, for At any given moment, the CSP's minimum power output is [not specified]. (6) CSP ramp constraint The power output of CSP power generation per unit time meets the following constraints: (48) In the formula, , The maximum power of the CSP during its rise and fall per unit time; (7) CSP charging and discharging thermal power constraint CSP thermal storage system in The heat storage and heat release power at any given time must meet the following constraints: (49) In the formula, and The lower limit for charging and releasing heat in CSP thermal storage systems; and The upper limit for the charging and releasing of heat in the thermal storage system; (8) CSP heat charging and heat dissipation constraints CSP The mathematical formula for the inability to simultaneously perform heat charging and heat release is as follows: (50) (9) IFCCPP alcoholamine solution volume and CO2 mass constraint CO2 exists in the form of a compound in the alcoholamine solution of the solution storage device. The relationship between the mass of CO2 extracted and the volume of the alcoholamine solution is as follows: (51) In the formula, for Time of the first The volume of amine solution required for an IFCCPP unit to release CO2; and These are the molar masses of the alkanolamine solution and CO2, respectively. , and These represent the regeneration tower desorption rate, alkanolamine solution concentration, and density, respectively. for Time of the first CO2 capture quality of the IFCCPP unit; (10) IFCCPP storage capacity constraint (52) in, and Power plants The installed rich liquid storage tank and poor liquid storage tank are in The volume of solution over a given time period; For power plants The capacity of the installed solution storage device; , , and Power plants The initial and final solution volumes of the installed rich and poor solution storage tanks; (11) FSCCPP constraint FSCCPP does not include a solution storage device, and its constraints are as follows: (53) (12) LADR related constraints 1) User battery level fluctuates within a certain range: (54) In the formula, It is the power fluctuation ratio; 2) To avoid excessive price differences between peak, off-peak, and valley periods, it is necessary to restrict electricity prices during these periods: (55) In the formula, , and These represent prices during peak load periods, normal load periods, and off-peak periods, respectively. 3) To ensure user participation in LADR, the user's electricity expenditure after LADR optimization should not be higher than the electricity expenditure before optimization: (56) 4) The average electricity price after LADR cannot be higher than before optimization: (57) In the formula, and These are the average electricity prices before and after LADR optimization, respectively.

[0020] Compared with the prior art, the beneficial effects of the present invention are: 1. Multi-objective collaborative optimization: Accuracy and stability are considered simultaneously in source-load forecasting. The MORUN algorithm is used to select a compromise solution in the Pareto front, making the forecast results of wind power, photovoltaic power and power load more suitable for power system dispatching scenarios with high robustness requirements.

[0021] 2. High solution efficiency: Latin hypercube sampling and synchronous backband elimination are used to generate typical scenarios of wind power, photovoltaic power and power load. The source-load uncertainty optimization problem is transformed into a deterministic typical probabilistic scenario problem for solution. This can simulate the uncertainty characteristics of wind power, photovoltaic power and power load, reduce the difficulty of modeling and solving, and obtain a more reasonable and robust decision scheme.

[0022] 3. Effectively reduce carbon emissions: On the source side, the combined output of multiple energy sources such as IFCCPP, CFPP, WF, PP, and CSP reduces system carbon emissions. LCTM is introduced to further control system carbon emissions. On the load side, a PSMORUN-LADR electricity price-load optimization model is established. By optimizing the power load and adjusting the output ratio of high-carbon and low-carbon units on the source side, a source-load low-carbon resource complementary and coordinated scheduling HECM model is established. This can minimize the carbon emissions of the power system. With the goal of maximizing the total output of the power source side to track the load curve and minimizing the overall economic cost of the system, the proposed PSMORUN algorithm is used to solve the model to achieve low-carbon, economical, and safe dispatch of the power system. Attached Figure Description

[0023] Figure 1 This is a structural diagram of the PCPP system of the present invention.

[0024] Figure 2 This is a comparison chart of the net output power of the three operating modes of the PCPP of this invention.

[0025] Figure 3This is a schematic diagram of the energy flow in the IFCCPP of the present invention.

[0026] Figure 4 This is a schematic diagram illustrating the LADR low-carbon principle of the present invention.

[0027] Figure 5 This is a schematic diagram comparing the unit power generation costs under the carbon trading mechanism of this invention.

[0028] Figure 6 This is a schematic diagram illustrating the low-carbon principle of source-load coordination during peak and off-peak load periods in this invention.

[0029] Figure 7 This is a diagram of the hybrid pseudo-inverse Laguree neural network structure of the present invention.

[0030] Figure 8 This invention presents a multi-source collaborative optimization scheduling framework that considers the complementary low-carbon characteristics of source and load.

[0031] Figure 9 This is a flowchart of the PSMORUN-HECM optimized scheduling method of the present invention.

[0032] Figure 10 This is a graph showing the training and testing data of wind power, photovoltaic power, and power load of this invention.

[0033] Figure 11 This is an RLMD data decomposition diagram of WP, PVP, and PL in this invention.

[0034] Figure 12 This is a combined time series component diagram of wind power, photovoltaic power, and electricity load for this invention.

[0035] Figure 13 This is a statistical analysis of the evaluation index results for the source and load of this invention, which are 1 hour to 24 hours ahead of both sides.

[0036] Figure 14 This is a comparison chart of the selection and prediction curves of the PF non-dominated solution predicted by the source and load sides 24 hours ahead in this invention.

[0037] Figure 15 This is a probability density distribution diagram of the prediction error of wind power, photovoltaic power and power load 24 hours ahead of time in this invention.

[0038] Figure 16 This is a diagram showing the generation and reduction results of wind power, photovoltaic power, and power load scenarios in this invention.

[0039] Figure 17 This invention presents the selection of non-dominated solutions in PSMORUN-LADR and the comparison of electricity price and load before and after optimization.

[0040] Figure 18This presents the comparison results of electricity price and load before and after PSMOUN-LADR optimization under different probability scenarios of this invention.

[0041] Figure 19 This invention provides a compromise solution selection and source-load power balance verification for the PSMORUN-HECM model.

[0042] Figure 20 The diagram shows the source-load power balance verification results of the PSMORUN-HECM model under scheduling scenarios 1-4 of this invention.

[0043] Figure 21 This is the optimal scheduling scheme and source-load power balance verification diagram for the multi-objective scheduling model of this invention. Detailed Implementation

[0044] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0045] Please see Figure 1-21 The present invention provides the following technical solutions: A heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load includes the following steps: (1) Establish a WP-PVP-PL source-load prediction model to provide typical probabilistic scenarios of WP, PVP and PL for HECM.

[0046] (2) The source side combines multiple energy sources such as IFCCPP, CFPP, WF, PP and CSP to reduce system carbon emissions, and introduces LCTM to further control system carbon emissions.

[0047] (3) Establish a PSMORUN-LADR electricity price-load optimization model on the load side, and adjust the output ratio of high-carbon and low-carbon units on the source side by optimizing the power load.

[0048] (4) Based on the low-carbon characteristics of both source and load, establish a source-load low-carbon resource complementary and coordinated scheduling HECM model.

[0049] (5) The PSMORUN algorithm is used to solve the HECM model in typical probabilistic scenarios.

[0050] (6) The probability of typical scenarios is used as the weight coefficient of the objective function of PSMORUN-HECM in the corresponding scenario. After linear weighting, the expected value of the objective function of power system optimization in typical scenarios is obtained.

[0051] This invention delves into the low-carbon dispatch potential of IFCCPP, CHES, and LADR by analyzing their low-carbon operating principles and complementary characteristics. Thermal power, power plants (WP), power plants (PVP), concentrated solar power (CSP), and power plants (PL) are all considered dispatchable resources and categorized into three classes based on their different carbon emission characteristics. Class 1 consists of high-carbon units, such as CFPP. Class 2 consists of low-carbon units, such as IFCFPP. Class 3 consists of zero-carbon units, such as WF, PP, CSP, and LADR providing virtual power. The complementary low-carbon operating principle of source and load is analyzed below.

[0052] The IFCCPP operates in two modes: flue gas separation and solution storage. Flue gas separation adjusts carbon capture energy consumption and the net output power of the CCPP by controlling the proportion of flue gas directly emitted into the air. Solution storage operation introduces lean and rich liquor storage systems, ensuring that the amount of rich liquor absorbed from the absorber is no longer equal to the amount entering the regeneration tower. This decouples, to some extent, the CO2 absorption process, which determines the carbon capture rate, from the solution regeneration process, which determines the carbon capture energy consumption. The structural block diagram of the IFCCPP is shown below. Figure 1 As shown: ① and ② are auxiliary carbon capture facilities for PCPP. Without these auxiliary facilities, PCPP mainly consists of equipment such as a flue gas cooler, an absorber heat exchanger, and a regeneration tower. CO2 generated by CFPP enters the absorber through the flue gas cooler, where it is converted into CO2 compounds and transported to the regeneration tower for processing. Subsequently, CO2 in the eluent is collected by a compressor through a separation tank, and the remaining eluent is returned to the absorber for recycling, thereby reducing CO2 emissions. However, while this operation mode can achieve low-carbon or zero-carbon CO2 emissions, its flexibility and economic efficiency are relatively low. Therefore, it is necessary to add auxiliary carbon capture facilities ① or ② to improve the flexibility of PCPP operation.

[0053] Currently, PCPP operation modes can be divided into three categories: split-flow, storage-flow, and integrated flexible operation. Split-flow PCPP adjusts carbon capture energy consumption and net output power by adding a "① flue gas diversion device" to control the proportion of flue gas directly emitted into the air. Storage-flow PCPP introduces "② lean and rich liquid storage devices" to ensure that the rich liquid absorbed from the absorber is no longer equal to the rich liquid entering the regeneration tower, thus decoupling the energy between CO2 capture volume and the energy consumption of generating the captured CO2 solution.

[0054] The addition of "① flue gas diversion device" and "② lean and rich liquid storage" constitutes the IFCCPP. Compared to a diversion-type PCPP, the IFCCPP can shift CO2 capture energy consumption during peak electricity load periods to off-peak periods, alleviating the contradiction between electricity demand and CO2 capture energy consumption during these periods. It can also increase CO2 capture energy consumption during off-peak periods to increase the absorption of renewable energy and reduce the net output of the IFCCPP. Compared to a liquid storage-type PCPP, the IFCCPP can control the amount of CO2 released into the atmosphere according to system scheduling needs, flexibly adjusting carbon capture energy consumption and improving scheduling flexibility.

[0055] The time-shifted CO2 capture energy consumption of IFCCPP depends on the amount of liquid stored in the solution storage tank. When the amount of rich liquid increases and the amount of lean liquid decreases in both the rich and lean solution storage tanks within a certain period, carbon capture energy consumption decreases, and the net output of IFCCPP increases. During this period, CO2 only undergoes absorption and not desorption. Conversely, when the amount of rich liquid decreases and the amount of lean liquid increases in both solution storage tanks (indicating that the CO2-storing solution is undergoing desorption at this moment), carbon capture energy consumption increases, and the net output of IFCCPP decreases. Therefore, by adjusting the amount of liquid stored in the solution storage tank, the time-shifted carbon capture energy consumption of IFCCPP can be achieved, thereby tapping the system's low-carbon potential and improving scheduling flexibility. A comparison of net output power for the three operating modes of PCPP is provided below. Figure 2 As shown.

[0056] During peak load periods, the total output power of FSCCPP is high, generating a large amount of CO2. Full capture of this CO2 would increase carbon capture operation energy consumption, leading to a decrease in the net output power of FSCCPP and conflicting with load demand. Using IFCCPP, excess CO2 can be stored in a solution storage tank. This reduces CO2 emissions, lowers carbon capture energy consumption, promotes output from carbon capture power plants, and enables low-carbon units to replace high-carbon units.

[0057] During periods of low load, FSCCPP reduces net output and expands its operating range by increasing CO2 capture. Figure 2 Operating mode 2). However, due to the low load demand and low total output power, the amount of CO2 generated is also small, making it impossible to effectively reduce the net output of the system. If IFCCPP is used, the CO2 in the solution storage can be released and captured by the carbon capture device, thereby increasing the system's carbon capture energy consumption and further increasing the lower limit and range of the net output of CCPP. Figure 2 (Operating mode 3). This operating mode is conducive to further promoting the consumption of renewable energy sources such as wind power and photovoltaics, and replacing the output of high-carbon units.

[0058] In summary, solution storage can transfer CO2 from peak load to off-peak load for processing, thereby shifting the energy consumption of CO2 capture from peak to off-peak load. This is equivalent to replacing the expensive, high-carbon thermal power output during peak load with the inexpensive, low-carbon carbon capture power output during off-peak load, along with the additional zero-carbon output from wind, photovoltaic, and solar thermal power, which is beneficial for promoting the low-carbon and economical operation of the power system. The energy flow of this invention (IFCCPP) is as follows: Figure 3 As shown.

[0059] However, the IFCCPP's low-carbon operation mode has certain limitations. Firstly, it is... Figure 2 It is known that IFCCPP and FSCCPP have the same upper limit of net output, but the former has a lower lower limit. Therefore, when the load peak is large, IFCCPP has a weaker ability to provide spinning reserve compared to FSCCPP. This necessitates other high-carbon units to provide more spinning reserve to meet load demand, reducing the low-carbon operation advantage of IFCCPP. Secondly, the peak-shaving and valley-filling effect of IFCCPP is limited by the maximum operating state of the regeneration tower and compressor, as well as the volume of the solution storage tank; its high cost restricts the development of IFCCPP. Therefore, to maximize the advantages of IFCCPP, it is necessary to use demand response to peak and valley loads and prevent excessively high load peaks.

[0060] LADR (Low-Carbon Utilization) guides users to change their electricity consumption habits by optimizing electricity pricing, encouraging them to reduce electricity consumption during peak hours and increase it during off-peak hours, thereby reducing electricity load fluctuations. The low-carbon principle of LADR is as follows: Figure 4 As shown.

[0061] pass Figure 4 It can be seen that during peak PL periods, when the net output of IFCCPP units reaches its upper limit, this power load gap will be filled by high-cost, high-carbon-emission CFPP units. After LADR peak shaving and valley filling, this power load deficit is transferred to low-carbon-emission and low-cost IFCCPP units or low-cost, zero-carbon-emission WP, PVP, and CSP units working together, which can effectively reduce system carbon emissions.

[0062] However, the LADR low-carbon approach also has certain shortcomings. The peak load transferred by LADR will be handled by two parts: first, the curtailment of wind, solar, and solar thermal power during off-peak hours; second, the remaining high-carbon units with lower carbon emission intensity. When the curtailment of wind, solar, and solar thermal power is too small, the alternative power source is still a high-carbon unit. Although the system's economic efficiency will improve and carbon emissions will decrease, the overall carbon emission performance of the system is still not ideal. Therefore, it is necessary to complement the low-carbon characteristics of IFCCPP and LADR to further explore the system's low-carbon potential.

[0063] LCTM (Limited Carbon Migration Technology) is a low-carbon mechanism that uses CO2 emission rights as a trading commodity to incentivize high-carbon-emitting enterprises to proactively reduce emissions, effectively curbing carbon emissions from high-carbon units. The low-carbon mechanism of LCTM is analyzed as follows: Figure 5 As shown.

[0064] Figure 5 In this study, unit costs are categorized into four types: low-carbon low-cost, high-carbon low-cost, low-carbon high-cost, and high-carbon high-cost. Without a carbon trading mechanism, units 1 and 2, with lower unit thermal power costs, will be prioritized for operation. This will lead to unit 2 maintaining consistently high carbon emissions, which is detrimental to the system's low-carbon operation. With the introduction of a carbon trading mechanism, high-carbon emission units incur high carbon trading costs and high overall economic costs, while low-carbon emission units incur low carbon trading costs and low overall economic costs. In this scenario, units 1 and 3, with lower unit overall costs, will be prioritized for operation, thereby suppressing high-carbon units from reducing carbon emissions and improving the low-carbon operation capabilities of the units.

[0065] The above analysis shows that both IFCCPP and LADR can promote the low-carbon operation of the system, but both have certain limitations. Therefore, this invention combines the two low-carbon strategies to further explore the low-carbon potential of the system. 1) IFCCPP can compensate for the shortcomings of LADR's low-carbon capabilities. During demand response, the expensive high-carbon unit power generation during peak loads can be replaced by renewable energy curtailment and low-carbon unit power generation during off-peak loads, thus compensating for the low-carbon shortcomings of LADR. 2) LADR can reduce peak loads, effectively reducing the net output pressure on IFCCPP and effectively improving the spinning reserve capacity of IFCCPP. 3) LADR effectively optimizes the low-carbon capabilities of the system, reducing the solution storage capacity of IFCCPP while ensuring the low-carbon performance of the system, thereby reducing the economic cost of the system. Figure 6 The low-carbon principle of IFCCPP and LADR working together during peak and off-peak load periods is described.

[0066] Figure 6 This paper compares five operating modes to illustrate the complementary advantages of the low-carbon characteristics of combining IFCCPP and LADR. The five operating modes are: 1) CFPP; 2) CFPP combined with LADR; 3) FSCCPP; 4) IFCCPP; and 5) IFCCPP combined with LADR. Figure 5It can be seen that during peak load periods, compared to other models, Model 5 can convert more high-carbon unit output into low-carbon unit output and LADR virtual output, effectively increasing the system's on-spin reserve capacity and further compressing the output of high-carbon units. During off-peak load periods, Model 5 can better absorb the abandoned renewable energy, achieving full substitution for high-carbon unit power generation. The peak-shaving pressure during normal load periods is not as significant as during peak and off-peak periods, and the conflict between IFCCPP energy consumption demand and load demand is not as significant as during peak periods. Therefore, the scheduling plan can be flexibly allocated according to the situation during normal load periods. Furthermore, the CO2 captured during normal load periods can be stored in a solution storage device to ensure that IFCCPP has sufficient energy consumption to absorb renewable energy during off-peak load periods.

[0067] In conclusion, a power system that combines the low-carbon characteristics of IFCCPP and LADR can achieve complementary advantages throughout the entire time period. Combined with the low-carbon policy mechanism of LCTM, the low-carbon performance of the system can be further improved.

[0068] Example 1. Data processing and parameter settings (1) WP-PVP-PL data collection The experimental data for WP, PVP, and PL in this invention are sourced from Hami City, Xinjiang Uygur Autonomous Region, China. This invention collected actual data on wind power, photovoltaic power, and electricity load in this region in 2018. Historical data from July to September (sampling interval of 1 hour) were selected as the training sample set for the source-load two-sided prediction model. The first 70% (1512 sampling points) of the training sample was used for training, and the last 30% (648 sampling points) was used for testing, performing source-load two-sided predictions with a lead time of 1 hour to 24 hours to verify the performance of the proposed RLMD-MORUN-HPLNN-AdaBoost.R2 prediction model. The training and testing data for WP, PVP, and PL are as follows: Figure 10 As shown in the diagram, the green, orange, and blue areas represent the training dataset, while the red area represents the test dataset.

[0069] (2) WP-PVP-PL data processing This invention employs Restricted Flow Decomposition (RLMD) to decompose WP, PVP, and PL time series into high-frequency and low-frequency components, thereby reducing their volatility and improving the accuracy of the prediction model. RLMD can adaptively decompose the number of high-frequency and low-frequency components based on the characteristics of the time series. The results of RLMD decomposition of WP, PVP, and PL time series components (TCs) are as follows: Figure 11 Show.

[0070] (3) WPE complexity calculation This invention uses WPE to qualitatively analyze the complexity of the WP, PVP, and PL subsequences after RLMD decomposition, and merges subsequences with similar complexity to improve the prediction efficiency of the RLMD-MORUN-HPLNN-AdaBoost.R2 source-load two-sided prediction model. The entropy calculation results of the WP, PVP, and PL time series components by WPE are shown in Table 1.

[0071] Table 1. Statistical table of entropy values ​​for WP, PVP, and PL time series components.

[0072] To reduce the prediction complexity of the RLMD-MORUN-HPLNN-AdaBoost.R2 source-load prediction model, this invention merges time series components with similar entropy values. The merged partitioning results of the WP, PVP, and PL time series components are as follows: Figure 12 As shown.

[0073] pass Figure 12 It can be seen that the entropy values ​​of the time series components of WP, PVP, and PL after RLMD decomposition are similar. Therefore, this invention merges and divides the time series components with similar entropy values. The number of time series components of WP, PVP, and PL after merging is reduced to 3 each, which greatly reduces the prediction complexity of the model.

[0074] (4) Source load prediction objective function This invention constructs an RLMD-MORUN-HPLNN-AdaBoost.R2 source-load multi-objective prediction model with accuracy and stability as optimization objectives. The mathematical formula for its objective function is as follows: (58) (59) In the formula, The prediction accuracy index reflects the degree of difference between the predicted value and the actual value. The prediction stability index reflects the degree of dispersion of the prediction error. , and These represent the predicted values, actual values, and sample sizes for WP, PVP, and PL, respectively. This invention employs a fuzzy decision-making method to select the non-dominated solution for PF, using satisfaction as the evaluation index for the decision.

[0075] (5) Source-load prediction and evaluation indicators This invention uses the stability index (SDEX), mean absolute error (MAE), root mean square error (RMSE), index of agreement (IA), and median absolute percentage error (MdAPE) as evaluation metrics for the source-load prediction model to comprehensively evaluate the prediction performance of the RLMD-MORUN-HPLNN-AdaBoost.R2 source-load prediction model.

[0076] (6) HECM related parameter settings This invention uses a 10-unit power system based on the IEEE-39 node as a case study to analyze relevant comparative experiments of the PSMORUN-HECM dispatch model. The relevant HECM parameter settings are shown in Table 2.

[0077] Table 2 HECM related parameter settings

[0078] 2. Experimental Results and Analysis This invention conducts a comprehensive comparative experiment on the performance of the proposed RLMD-MORUN-HPLNN-AdaBoost.R2 prediction model and the PSMORUN-HECM scheduling model. The WP-PVP-PL prediction experiment mainly includes prediction experiments 1-24 hours in advance, comparison experiments of multiple prediction models, and generation of typical source-load probabilistic scenarios. The PSMORUN-HECM experiment mainly includes PSMORUN-LADR analysis, source-load matching coefficient sensitivity analysis, PSMORUN-HECM typical probabilistic scenario scheduling result analysis, PSMORUN-HECM multi-scheduling scenario comparative analysis, and PSMORUN-HECM multi-objective scheduling model comparative analysis. The parameter settings for the MORUN and PSMORUN algorithms are derived from publicly available literature.

[0079] (1) Source load prediction performance test To apply source-load prediction results to power system optimal dispatch, it is necessary to predict wind power, photovoltaic power, and power load 1-24 hours in advance to obtain the prediction results of WP, PVP, and PL for the 24-hour day-ahead. To evaluate the prediction performance of the proposed RLMD-MORUN-HPLNN-AdaBoost.R2 source-load prediction model, this invention compares it with current mainstream neural network prediction models. The neural networks used for comparison include: BPNN, LSTM, RBF, RELM, and HPLNN. The parameter settings of each neural network were determined through cross-validation. The maximum number of function evaluations (FEs) was set to 1E+04. The comparative experiment was run independently 30 times, and the average value was taken as the final experimental statistical result. The prediction statistics of the five neural network prediction models on WP, PVP, and PL are shown in Table 3.

[0080] Table 3. Statistics of 1-hour lead-time prediction results of the neural network prediction model on both sides of the source and load.

[0081] As shown in Table 3, each neural network prediction model achieved different advantages across various evaluation metrics. Overall, compared to the comparative neural network prediction models, the proposed RLMD-MORUN-HPLNN-AdaBoost.R2 source-load prediction model achieved better prediction results on IA, MAE, RMSE, SDEX, and MdAPE evaluation metrics. This indicates that the proposed source-load prediction model has higher prediction accuracy and stronger prediction stability, and the HPLNN neural network has stronger prediction capabilities on WP, PVP, and PL.

[0082] In day-ahead optimal dispatching of power systems, relying solely on 1-hour lead-time source-load two-sided forecasts is insufficient to provide effective future dispatching information for WP, PVP, and PL. To improve the applicability of the proposed model's prediction results in day-ahead optimal dispatching of power systems, it is necessary to perform WP, PVP, and PL forecasts from 1 hour to 24 hours in advance. The statistical results of the source-load two-sided forecast evaluation indicators for power systems RLMD-MORUN-HPLNN-AdaBoost.R2 from 1 hour to 24 hours in advance are as follows: Figure 13 As shown.

[0083] Figure 13 This displays the statistical results of prediction error evaluation indicators for WP, PVP, and PL from 1 hour to 24 hours in advance. (Left side) The axis coordinates are MAE, RMSE, SDEX, and MdAPE index values, with the right side... The axes represent the IA index values. As shown in the graph, with increasing lead time, the IA values ​​of WP, PVP, and PL gradually decrease, while the values ​​of MAE, RMSE, SDEX, and MdAPE gradually increase. This indicates that the longer the lead time, the worse the performance of the prediction model. The comparison results of PF non-dominated solution selection and prediction curves for the RLMD-MORUN-HPLNN-AdaBoost.R2 power system source-load dual-side prediction model with a 24-hour lead time are shown below. Figure 14 As shown.

[0084] from Figure 14 It can be seen that as the prediction step size increases, the model's prediction performance deteriorates, and the source-load two-sided advance prediction results are more consistent with the actual application of power system optimal dispatch. Furthermore, from... Figure 13 It can be seen that the prediction error evaluation indicators for WP, PVP, and PL, ranging from 1 hour to 24 hours in advance, are all within acceptable ranges. Therefore, the advance prediction results of the RLMD-MORUN-HPLNN-AdaBoost.R2 source-load prediction model proposed in this invention can provide an accurate and stable data foundation for the optimal scheduling of power systems.

[0085] (2) Generation of typical scenarios with uncertain source and load This invention employs Latin hypercube sampling (LHS) and simultaneous backward reduction (SBR) to generate typical scenarios for power WP, power PVP, and power PL, transforming the source-load uncertainty optimization problem into a deterministic typical probabilistic scenario problem for solution. This approach simulates the uncertainty characteristics of WP, PVP, and PL while reducing the difficulty of model solving, resulting in a more reasonable and robust decision-making scheme. This invention uses the probability density distribution of WP, PVP, and PL prediction errors 24 hours in advance as an example for analysis. The probability density distribution of the prediction errors on both the source and load sides of the power system 24 hours in advance is as follows: Figure 15 As shown: pass Figure 15 It can be seen that the probability density distribution histograms of the 24-hour lead time prediction errors for WP, PVP, and PL exhibit "peaked" and "fat-tailed" characteristics, and the overall distribution follows a Gaussian distribution. Therefore, this study selects the Gaussian distribution as the probability density distribution for generating typical scenarios using LHS and SBR. A specific day from the September prediction results for WP, PVP, and PL is selected as the typical day for generating and reducing typical source load scenarios. To fully consider source load uncertainty, firstly, 1000 WP, PVP, and PL scenarios are generated using LHS. Then, the number of scenarios is reduced using the SBR method, and five representative typical scenarios are generated based on probability. The scenario generation and reduction results for WP, PVP, and PL are shown below. Figure 16 As shown.

[0086] Figure 16 The red solid line represents the prediction results for WP, PVP, and PL. The scenarios generated based on the source load prediction results completely "wrap" the prediction results, indicating that the 1000 generated scenarios fully characterize the uncertainty features of WP, PVP, and PL. Figure 16 Five typical probabilistic scenarios are presented, namely WP, PVP, and PL, with scenario probabilities of 34.0%, 12.3%, 10.2%, 30.1%, and 13.4%, respectively. As can be seen from the figure, scenario 1 has the highest probability of occurrence among the typical source-load scenarios. Therefore, this invention takes scenario 1 as an example to analyze the source-load collaborative scheduling results.

[0087] (3) PSMORUN-HECM performance test 1) PSMORUN-LADR analysis This invention uses fuzzy membership functions to calculate the peak and valley membership values ​​of the load at each time point, and combines this with the peak and valley time period division principle to determine the final PL time period division result. The statistical results of the peak and valley membership calculations for the 24-hour load in typical probability scenario 1 are shown in Table 4: Table 4. Peak-valley membership statistics at different times under typical probability scenario 1

[0088] According to the peak-valley membership statistics in Table 4, 12 o'clock is the point of maximum load and is assigned to the PL peak period. 7 o'clock is the point of minimum PL and is assigned to the PL valley period. According to the principle of dividing peak and valley periods of power load

[43] , the PL 24-hour peak-valley period division results are shown in Table 5. The same principle can be used to obtain the PL 24-hour peak-valley period division results for probability PL scenarios 2 to 5, and the statistical results are shown in Table 6.

[0089] Table 5. Results of 24-hour peak-shaving-valley time division under power load probability scenario 1

[0090] Table 6. Results of 24-hour peak-shaving-valley period division for probability scenarios 2 to 5 of electricity load.

[0091] Based on the precise segmentation of peak, flat, and valley periods (PL), this invention verifies the performance of the LADR electricity price-load optimization model. Using peak, flat, and valley electricity prices as decision variables, the PSMORUN algorithm is employed to solve the LADR model. The selection of the PF nondominated solution for the PSMORUN-LADR electricity price-load optimization model and the comparison of electricity price-load before and after optimization are shown below. Figure 17 As shown.

[0092] from Figure 17 (a) It can be seen that the shape of the PF distribution obtained by PSMORUN-LADR is close to a straight line, and the non-dominated solutions are uniformly distributed with a high distribution density. This indicates that PSMORUN has a good global exploration capability in handling LADR problems and can solve LADR problems well. From Figure 17 (b) It can be seen that PSMORUN-LADR effectively improves users' electricity consumption habits, makes PL fluctuations more stable, and further enhances the stability of the power system. The comparison results of PSMORUN-LADR before and after optimization are shown in Table 7.

[0093] Table 7 Comparison of PSMORUN-LADR optimization results before and after optimization

[0094] As shown in Table 7, the proposed PSMORUN-LADR electricity price-load optimization model significantly reduces PL fluctuations and improves the electricity economy for users, thereby incentivizing users to actively participate in load-side demand response and further enhancing dispatch flexibility. The electricity price-load comparison results after PSMORUN-LADR under typical probability scenarios 2-5 are as follows: Figure 18 As shown.

[0095] 2) Sensitivity analysis of source-load matching coefficient This invention is from arrive A total of 19 comparative experiments were set up to explore different , and Based on the weighting ratio, the variation pattern of SLDI is analyzed, and the optimal SLDI value is selected. , and The weight ratios were used as parameters in the PSMORUN-HECM model. The experimental comparison results are shown in Table 8.

[0096] Table 8 α 1. α 2 and α 3 Statistical Results of SLDI Optimization with Different Weight Ratio Combinations

[0097] As can be seen from Table 8, the SLDI value varies with the weighting coefficient. , and It changes with the changes. When At that time, the SLDI value was at its minimum of 23.810. , and The values ​​are 0.1, 0.1 and 0.8 respectively.

[0098] 3) Analysis of scheduling results for typical probabilistic scenarios To verify the effectiveness of the proposed PSMORUN-HECM, this invention uses five typical probabilistic scenarios (WP, PVP, and PL) as inputs to the HECM model, with SLDI and comprehensive economic cost as objective functions. PSMORUN is used to solve the HECM model, and the weighted sum of the probabilities of each source-load scenario is used as the overall expected result of the system. Table 9 shows the statistical results of the system's operating cost and source-load difference index under different typical probabilistic scenarios.

[0099] Table 9. Scheduling results of the PSMORUN-HECM model in probabilistic scenarios 1-5

[0100] Table 9 shows that the comprehensive economic cost and source-load difference index of the PSMORUN-HECM scheduling model differ under different scenarios. This indicates that the PSMORUN-HECM scheduling results are closely related to WP, PVP, and PL load scenarios. It also suggests that power system optimization scheduling research in single-day scenarios is somewhat accidental, while source-load matching-based optimization scheduling research in multiple scenarios is more universally applicable. The power factor (PF) and optimal system scheduling scheme of PSMORUN-HECM, as well as the source-load power balance verification under typical probabilistic scenario 1, are shown below. Figure 19 As shown.

[0101] As shown in Figure 19, the zero-carbon output of the power generation unit (WF), power generation unit (PP), and power supply unit (CSP) coordinated by the power source, along with the low-carbon output of the IFCCPP, accounts for the vast majority of the total output, effectively reducing the output of high-carbon units and decreasing system carbon emissions. Furthermore, the scheduling results obtained from the PSMORUN-HECM scheduling model showed renewable energy curtailment at 5:00, 6:00, 7:00, 13:00, and 16:00, but no load shedding occurred during the entire scheduling period. Moreover, compared to the load before LADR, the load fluctuations after PSMORUN-LADR optimization are smoother, further enhancing the stability of the power system. In summary, the PSMORUN-HECM scheduling model proposed in this invention, while ensuring the safety and stability of the power system, can also achieve a high proportion of renewable energy utilization, reduce system carbon emissions, and realize safe and low-carbon scheduling of the power system.

[0102] 4) Comparative Analysis of PSMORUN-HECM Multi-Scheduling Scenarios To verify the effectiveness of the PSMORUN-HECM model proposed in this invention, five scheduling scenarios were divided for comparative analysis experiments. Taking typical probabilistic scenario 1 (source-load WP, ​​PV, and EL) as an example, the analysis verified the low-carbon scheduling performance of source-load under different mechanisms. The detailed scheduling scenario division is as follows: Table 10 Comparison of settings for multiple scheduling scenarios

[0103] Under typical probabilistic scenario 1, the optimal scheduling scheme and source-load power balance test results of the PSMORUN-HECM model for scheduling scenarios one through four are as follows: Figure 20 As shown.

[0104] from Figure 20 As can be seen, renewable energy curtailment occurred in dispatch scenario 1 at 5:00, 6:00, 13:00, and 14:00, while renewable energy curtailment occurred in dispatch scenario 2 at 6:00, 13:00, and 14:00. This indicates that IFCCPP has a lower net output limit and can absorb more renewable energy output. Furthermore, the power pulsation (PL) fluctuations in dispatch scenarios 4 and 5 are smoother compared to those in dispatch scenarios 1-3. LADR shifts a portion of peak load to off-peak periods. This load shifting deficit is supplied by the output of low-carbon, low-cost IFCCPP units and the coordinated output of low-cost, zero-carbon WF, PP, and CSP units, effectively reducing system carbon emissions. Table 11 shows the statistical results of system operating costs and source-load difference indices for the five dispatch scenarios under typical probability scenario 1. Table 11 Statistical results of system operating costs and source-load difference index under five scheduling scenarios

[0105] As shown in Table 11, none of the five dispatching scenarios incurred load shedding costs, indicating that all five scenarios can achieve safe power system dispatching. Compared to dispatching scenario 1, dispatching scenario 2 reduced CHES curtailment costs by 18.3% and carbon emissions by 3.3%, demonstrating that compared to FSCCPP, IFCCPP can effectively increase the absorption of renewable energy and reduce system carbon emissions.

[0106] In scheduling scenario 3, the LCTM strategy was used to control the carbon emissions of the system. Compared with scheduling scenario 1, the overall economic cost was reduced by 5.8% and the carbon emissions were reduced by 29.5%. This shows that LCTM can effectively control the carbon emissions of high-carbon units and reduce the overall economic cost of the system.

[0107] Compared to scheduling scenario 1, scheduling scenario 4 reduced CHES curtailment costs by 26.1%, carbon emissions by 1.0%, overall economic costs by 3.2%, and SLDI index by 23.5%. This indicates that the scheduling strategy based on the complementary low-carbon characteristics of source and load using IFCCPP, CHES, and LADR can effectively reduce the overall economic cost of the system and improve system stability.

[0108] Compared to scheduling scenario 1, the PSMORUN-HECM scheduling model proposed in this invention reduces CHES curtailment cost by 33.5%, carbon emissions by 42.4%, overall economic cost by 7.7%, and SLDI index by 34.3%. This indicates that the scheduling model proposed in this invention can fully exploit low-carbon resources on both the source and load sides, leverage the complementary advantages of low-carbon characteristics on both sides, and maximize the utilization rate of renewable energy while reducing system carbon emissions.

[0109] (4) Comparison Experiment of Multi-Objective Scheduling Models To verify the effectiveness of PSMORUN in optimizing HECM, a comparative experiment was conducted with NSGAIII, MODA, MOSSA, and MORUN algorithms under a typical source-load scheduling scenario 1. The optimal scheduling scheme and source-load power balance verification of the PSMORUN-HECM scheduling model are as follows: Figure 17 As shown. The optimal scheduling scheme and source-load power balance verification for the other four multi-objective comparison models are as follows. Figure 21 As shown in Table 12, the statistical results comparing the operating costs and SLDI index of the five multi-objective scheduling models are presented.

[0110] Table 12. Comparison of Operating Costs and SLDI Indicators of Five Multi-Objective Scheduling Models

[0111] from Figure 21 As shown in Table 12, compared with the comparative models, the PSMORUN-HECM scheduling model proposed in this invention achieves optimal values ​​in carbon emissions, overall economic cost, and SLDI index. This indicates that the PSMORUN-HECM scheduling model can better formulate scheduling instructions that take into account the system's low-carbon, economic, safe, and flexible aspects. This invention provides a complete technical solution from source-load data prediction, typical probabilistic scenario construction, source-load low-carbon characteristic complementary mechanism to low-carbon-economic-safety scheduling, which can effectively reduce the uncertainty on both the source and load sides and effectively improve the economic, safe, flexible, and low-carbon scheduling performance of the power system.

[0112] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for coordinated prediction and scheduling of heterogeneous energy sources considering the complementary and matching low-carbon characteristics of source and load, characterized in that, Includes the following steps: S1: Based on Laguerre polynomials, pseudo-inverse learning and ensemble learning, robust local mean decomposition (RLMD) and weighted permutation entropy (WPE), a prediction model for wind power (WP), photovoltaic power (PVP) and power load (PL) is constructed, which takes into account both prediction accuracy and prediction stability. S2: Construct the HECM mathematical model. The HECM mathematical model framework includes the Integrated Flexible Operation Mode of Carbon Capture Power Plant (IFCCPP), Wind Farm (WF), Photovoltaic Power Plant (PP), Concentrated Solar Power (CSP), Active Demand Response (LADR), Tiered Carbon Trading Mechanism (LCTM), objective function, and related constraints. The load in the HECM mathematical model is jointly undertaken by high-carbon emission coal-fired power plant (CFPP), low-carbon emission IFCCPP, and zero-carbon emission multi-heterogeneous complementary energy generation (CHES). LADR is used to change users' electricity consumption behavior to optimize PL through peak-valley operation, and LCTM is introduced to control system carbon emissions and reduce overall economic costs. S3: Introduce the Source-Load Difference Index (SLDI) to measure the degree of matching between power output and grid load, so as to achieve a balance between the economic cost and safety performance of system dispatch results; S4: The multi-objective Runge-Kutta algorithm PSMORUN, based on a population parallel search mechanism, is used to solve HECM and obtain the scheduling prediction results.

2. The heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 1, is characterized in that... The modeling process for the wind power (WP), photovoltaic power (PVP), and electricity load (PL) prediction models is as follows: S11: Robust Local Mean Decomposition Robust Local Mean Decomposition (RLMD) is used to decompose the time series of wind power, photovoltaic power, and electricity load into high-frequency and low-frequency components to reduce their volatility. The core formula of RLMD is as follows: (1) (2) (3) (4) (5) In the formula, It is the average of adjacent local extreme points. Local extreme points The corresponding signal value, This is a local envelope estimate. and These are the frequency modulation signal and the envelope signal, respectively. and These are the product function and the residual signal, respectively. For mirror expansion of time series signals, For component index, For discrete time series indexing, The first PF component A single component, The number of amplitude-modulated signals participating in the synthesis; S12: Weighted permutation entropy We use Weighted Permutation Entropy (WPE) to analyze the complexity of the subsequences after RLMD decomposition, and merge subsequences with similar complexities to reduce the model's prediction complexity. The core formula of WPE is as follows: Decomposition of subsequences conduct Phase space reconstruction: (6) In the formula, and These represent the embedding bit length and latency, respectively. , The total length of the original decomposed subsequence; Calculate subsequence WPE value: (7) S13: Hybrid Pseudo-Inverse Laguree Neural Network (1) Hybrid Laguree Neural Network (HLNN) The orthogonality of the two sets of Laguerre orthogonal polynomials with respect to the weighting function is described as follows: (8) (9) The recurrence equations for the two sets of Laguerre orthogonal polynomials are as follows: (10) (2) Pseudo-reverse learning Pseudo-inverse learning is used to train the Hybrid Laguerre Neural Network (HLNN) and establish the Hybrid Pseudo-Inverse Laguerre Neural Network (HPLNN) model. The specific process is as follows: When the input matrix of the training set is And the output matrix is At that time, the optimal weight matrix of HPLNN The calculation is as follows: (11) in: (12) in It is a matrix Moore Penrose pseudo-retrogression, It is the output of the hidden layer of HPLNN. These are Laguerre orthogonal basis functions; (3) AdaBoost.R2 To ensure strong stability of the prediction results, AdaBoost.R2 is used to further improve the prediction accuracy of HPLNN. AdaBoost.R2 trains multiple weak predictors, and the final strong predictor is constructed through adaptive correction and recombination of the prediction error weights of these multiple predictors. This assumes the training sample set... , It is the input vector. It is the output vector; the core mathematical formula of Adaboost.R2 is as follows: (13) (14) (15) In the formula, The initial weights of the samples, and These are the weak predictor and the relative error, respectively. The normalization constant is yes The weighted median.

3. The heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 1, is characterized in that... The mathematical model for IFCCPP is as follows: (16) in, and They are respectively Period IFCCPP Total output power and net output power; and Carbon capture equipment Fixed energy consumption and operating energy consumption; The energy consumption coefficient for capturing one unit of CO2; for Period IFCCPP The amount of CO2 captured; This refers to the maximum operating condition coefficient of the regeneration tower and compressor; , and IFCCPP Carbon capture efficiency, flue gas split ratio, and carbon emission intensity coefficient; For IFCCPP The upper limit of output power, For solution storage exist The amount of CO2 to be captured provided during the time period; for The total amount of CO2 generated by IFCCPP during the period; The formula for converting the amount of CO2 extracted from a solution storage device into solution volume is: (17) in, For IFCCPP solution storage exist The volume of solution required to release CO2 at any given time; and Let be the molar mass of ethanolamine and CO2, respectively; This refers to the regeneration tower's analytical value; and These represent the concentration and density of the alkanolamine solution, respectively.

4. A heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 3, is characterized in that... The CSP mathematical model is as follows: (18) (19) (20) in, Let t be the heat power absorbed by the heat collection device at time t. The thermal power generated directly by the thermal collector. To supply the thermal power of the thermal storage system to the thermal collector, Let be the thermal storage power of the thermal storage device at time t. For the heat storage system's charging efficiency, The heat energy transferred from the heat storage device to the heat transfer fluid at time t. Let be the heat release power of the thermal storage device at time t. A product of zero indicates that the thermal storage device cannot store and release heat simultaneously. This indicates the heat release efficiency of the thermal storage system. Let be the electrical power of the solar thermal power plant at time t. This refers to the thermoelectric conversion efficiency of a solar thermal power plant.

5. A heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 4, is characterized in that... The LADR model is as follows: (21) (22) In the formula, Indicates the actual load transfer rate. and These represent the load transfer rates of the pessimistic and optimistic response curves, respectively. These are the weighting coefficients. For the peak-valley electricity price difference, and The nodes are divided into dead zone, response zone, and saturation zone based on the electricity price difference. Using the same method, the actual load transfer rates from peak to flat and from flat to valley were calculated respectively. and The load transfer amount and load value after responding to user demands are calculated as follows: (23) (24) in, , , These represent peak, flat, and trough periods, respectively. , and This represents the average load for each time period before the implementation of peak-valley time-of-use pricing; This represents the amount of load transfer caused by demand response at time t; , These represent the load values ​​at time t before and after the implementation of peak-valley electricity pricing; fuzzy membership functions are used to divide the load into peak, valley, and flat loads. The LADR model is constructed with the objectives of minimizing load fluctuations and maximizing household electricity satisfaction. The LADR optimization objective function is as follows: (25) (26) (27) (28) Where T is the scheduling period. To accurately measure the average load under demand response, and These are respectively user power comfort and economy. and They are respectively Electricity prices are calculated for both the period before and after demand response.

6. A heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 2, is characterized in that... The overall economic cost of the HECM mathematical model is as follows: (29) in, To take into account the overall economic costs, To enable the combined output of high-carbon and low-carbon power units, and It is time Unit fuel cost and valve point effect VPE cost at that time For the overall cost of a solar thermal power plant, and They are time CHES curtailment costs and load shedding costs at that time For carbon trading costs, For the cost of CCPP modification and depreciation, The cost of solvent loss during the carbon capture process; (1) Fuel cost of thermal power units (30) in, , and It is the fuel cost coefficient. It is the sum of low-carbon units and high-carbon units. It is time Combined output power of IFCCPP and IFCCPP; (2) Valve point effect cost (31) in, and It is the VPE cost coefficient; (3) CSP Comprehensive Cost (32) in, This is the CSP configuration cost per unit of heat storage capacity. It is the CSP thermal storage capacity. Environmental benefit coefficient It is CSP in Output power over a time period; (4) CHES curtailment cost When the CHES (Current Equivalent to Spinning Reserve) exceeds the scheduling plan demand, and the system's negative spinning reserve capacity cannot offset this excess power, CHES power curtailment will occur. The mathematical formula for the cost of curtailment is as follows: (33) in, To underestimate the cost coefficient, and They are respectively Real-time system network loss and demand response load, and They are respectively Wind and solar power generation capacity at all times for The negative spinning reserve capacity provided by the combined high-carbon and low-carbon units at all times; (5) System load shedding cost When the actual output of wind-solar hybrid power generation is less than the demand in the dispatch plan, and the system's reserved forward rotation reserve capacity is insufficient to make up for the power shortage, load shedding will occur. The mathematical formula for the load shedding cost of HECM is as follows: (34) in, To prevent the system from overestimating cost coefficients, for The positive-rotation reserve capacity provided by the combined high-carbon and low-carbon units at all times; (6) Costs of CCPP modification and depreciation (35) This includes the depreciation costs of CCPP and solution storage devices. and For FSCCPP depreciation costs and depreciation periods, This is the discount rate for CCPP projects. and For the volume and depreciation period of the solution storage device, The unit cost of solution storage devices; (7) Solvent loss cost (36) in, and These are the ethanolamine solvent cost coefficient and the solvent operating loss coefficient, respectively. For IFCCPP The amount of CO2 captured; (8) LCTM cost The mathematical formula for tiered carbon trading is as follows: (37) (38) (39) in, and These represent the carbon emissions and carbon emission quotas of thermal power units, respectively. For the number of IFCCPPs, and These represent the quota factor and carbon emission intensity factor of thermal power units, respectively. This represents the total carbon trading cost of the system. This indicates the length of the carbon emission range in carbon trading costs. This indicates the growth rate of carbon trading costs.

7. A heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 1, is characterized in that... The mathematical expression for the objective function of the Source-Load Difference Index (SLDI) is as follows: (40) This represents the variance of the source load, while , and These correspond to the standard deviations of power supply side output fluctuation, residual load, and curve inflection point, respectively; in addition, , and It is the weighting coefficient.

8. A heterogeneous energy collaborative prediction and scheduling method considering the complementary and matching low-carbon characteristics of source and load, as described in claim 1, is characterized in that... The constraints of the HECM mathematical model are as follows: (1) Power balance constraint In the HECM model, the source-side outputs are the combined outputs of high-carbon units, IFCCPP units, WF, PP, and CSP units. The load-side includes power system network losses and the power load after LFDR. Its constraints are equality constraints, and the mathematical expressions are as follows: (41) in, The mathematical formula is as follows: (42) In the formula, , and This is the network loss coefficient. and They are respectively Time of the first and the The output of the thermal power unit; (2) Upper and lower limits of unit output constraints (43) In the formula, and They are respectively The upper and lower limits of power generation for PCPP high-carbon units and IFCCPP units at all times; (3) Generator slope limit (44) In the formula, and These are the first in HECM The maximum power of each unit during the rise and fall per unit time; (4) Rotational spare constraint To address CHES uncertainties, IFCCPP and CCPP are used together to provide spinning reserve constraints; the mathematical expression for the spinning reserve capacity is as follows; (45) (46) In the formula, and These are the rated power of wind farms and photovoltaic power plants, respectively. To maximize the output of the solar thermal power plant, For the rotational backup response time, in the middle is Hour; (5) CSP output constraint The power generation capacity of a CSP must be within a certain range, and the mathematical formula is as follows: (47) In the formula, for At any given moment, the CSP's minimum power output is [not specified]. (6) CSP ramp constraint The power output of CSP power generation per unit time meets the following constraints: (48) In the formula, , The maximum power of the CSP during its rise and fall per unit time; (7) CSP charging and discharging thermal power constraint CSP thermal storage system in The heat storage and heat release power at any given time must meet the following constraints: (49) In the formula, and The lower limit for charging and releasing heat in CSP thermal storage systems; and The upper limit for the charging and releasing of heat in the thermal storage system; (8) CSP heat charging and heat dissipation constraints CSP The mathematical formula for the inability to simultaneously perform heat charging and heat release is as follows: (50) (9) IFCCPP alcoholamine solution volume and CO2 mass constraint CO2 exists in the form of a compound in the alcoholamine solution of the solution storage device. The relationship between the mass of CO2 extracted and the volume of the alcoholamine solution is as follows: (51) In the formula, for Time of the first The volume of amine solution required for one IFCCPP unit to release CO2; and These are the molar masses of the alkanolamine solution and CO2, respectively. , and These represent the regeneration tower desorption rate, alkanolamine solution concentration, and density, respectively. for Time of the first CO2 capture quality of the IFCCPP unit; (10) IFCCPP storage capacity constraint (52) in, and Power plants The installed rich liquid storage tank and poor liquid storage tank are in The volume of solution over a given time period; For power plants The capacity of the installed solution storage device; , , and Power plants The initial and final solution volumes of the installed rich and poor solution storage tanks; (11) FSCCPP constraint FSCCPP does not include a solution storage device, and its constraints are as follows: (53) (12) LADR related constraints 1) User battery level fluctuates within a certain range: (54) In the formula, It is the power fluctuation ratio; 2) To avoid excessive price differences between peak, off-peak, and valley periods, it is necessary to restrict electricity prices during these periods: (55) In the formula, , and These represent prices during peak load periods, normal load periods, and off-peak periods, respectively. 3) To ensure user participation in LADR, the user's electricity expenditure after LADR optimization should not be higher than the electricity expenditure before optimization: (56) 4) The average electricity price after LADR cannot be higher than before optimization: (57) In the formula, and These are the average electricity prices before and after LADR optimization, respectively.

Images