Virtual power plant source load interaction optimization scheduling model based on low-carbon response and solving algorithm

Abstract

The invention discloses a virtual power plant source load interaction optimization scheduling model based on low-carbon response and a solving algorithm, and belongs to the technical field of power system optimization scheduling. A low-carbon scheduling framework containing a distributed power supply, energy storage, a flexible load and a carbon transaction mechanism is constructed, the carbon emission intensity of each link is quantified to form a carbon flow scheduling signal, and a dynamic carbon emission factor and energy cost are coupled. A multi-objective optimization model is established, a complex function is processed by piecewise linearization, and a hybrid algorithm of an improved genetic algorithm and a commercial solver is designed to improve the solving efficiency. The prediction error is dynamically corrected through a'prediction-optimization-feedback 'closed loop, and the strategy is adjusted. According to the scheme, low-carbon and economic collaborative optimization is realized, renewable energy consumption and system stability are enhanced, user satisfaction and real-time scheduling are considered, and a solution is provided for low-carbon intelligent operation of the power distribution network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system optimal scheduling, in particular to a virtual power plant source-load interaction optimal scheduling model and solving algorithm fusing a low-carbon response mechanism. BACKGROUND

[0002] In the traditional distribution network scheduling model, distributed power sources (such as photovoltaic, wind power), energy storage systems and loads are usually regulated as independent units, lacking a collaborative optimization mechanism based on carbon emissions and power fluctuations. For example, CN110323785B points out that the interactive load in the prior art can only participate in scheduling through simple peak-valley transfer, and is not linked with the charging and discharging strategy of energy storage and carbon trading cost, resulting in that when photovoltaic output is excessive, the load cannot be effectively guided to be consumed or the energy storage cannot be effectively charged, and the light rejection rate is as high as 15%-20%. The traditional model does not introduce a dynamic carbon emission factor, and cannot reflect the real-time influence of energy structure changes on carbon emissions. For example, when the proportion of renewable energy in a region increases, the carbon emission intensity per unit of electricity should decrease, but the existing model usually uses a fixed carbon emission coefficient, resulting in that the scheduling result deviates from the actual low-carbon target.

[0003] The carbon trading mechanism is separated from the scheduling decision, and the carbon quota allocation and price fluctuation are not included in the optimization target. As described in CN110323785B, the existing method does not consider the influence of carbon cost on power generation combination, and high-carbon thermal power may still be preferentially scheduled under the economic target, which cannot achieve the "double carbon" target.

[0004] Non-convex non-linear functions (such as energy storage life attenuation model, carbon trading cost function) in the distribution network scheduling model cause the traditional algorithms (such as particle swarm optimization, genetic algorithm) to easily fall into local optimum, and the solving time is long. For example, when CN110323785B uses a multi-objective particle swarm optimization algorithm to solve a distribution network model containing 19 nodes, the time consumption of a single iteration is more than 30 minutes, which is difficult to meet the real-time scheduling demand.

[0005] The coding mode of mixed variables (continuous power value and discrete device switching state) is single, and the traditional binary coding cannot efficiently process continuous variables (such as wind power, energy storage SOC), resulting in that the algorithm search space is redundant and the convergence speed is slow. The prediction error of photovoltaic / wind power output usually reaches 10%-15%, and the existing model relies on fixed parameters to correct the error without establishing a "data acquisition-prediction-optimization-feedback" closed loop. For example, when the average photovoltaic prediction error in CN110323785B is more than 5%, the LSTM model parameters are not dynamically adjusted, resulting in that the subsequent scheduling strategy continuously deviates from the actual situation.

[0006] The load side response lags behind, and the regulation strategy of industrial and residential loads is not dynamically adjusted according to the real-time carbon emission intensity. For example, commercial loads may still use high-carbon thermal power during the photovoltaic low valley period, and cannot fully utilize low-carbon power. SUMMARY

[0007] In view of the deficiencies in the prior art above, the purpose of the present application is to propose a virtual power plant source-load interaction optimization scheduling model and solving algorithm based on low-carbon response, aiming to solve the blank of the prior art in low-carbon scheduling, source-load-storage depth interaction, algorithm efficiency and dynamic response, and to provide a systematic solution for the low-carbonization and intelligent operation of the distribution network.

[0008] In order to achieve the above purpose, the summary of the application adopts the following technical solutions:

[0009] A virtual power plant source-load interaction optimization scheduling model and solving algorithm based on low-carbon response, comprising the following steps:

[0010] S1. Construct a multi-energy collaborative virtual power plant low-carbon scheduling framework, establish a virtual power plant architecture, and design a dynamic carbon emission factor and energy cost coupled response mechanism; by quantifying the carbon emission intensity of each link of source-grid-load-storage, form a scheduling signal with carbon flow as the core, and realize the collaborative optimization of low-carbon target and economy;

[0011] S2. Establish a multi-objective optimization model and a constraint system, based on the established framework, construct a multi-objective optimization model, and the non-convex nonlinear function in the multi-objective optimization model is processed by using the piecewise linearization method;

[0012] S3. Design an improved hybrid solving algorithm, for the established multi-objective optimization model, combine the hybrid solving strategy of improved genetic algorithm and commercial solver, adopt hybrid coding mode, convert multiple objectives into single objective by weighting with preference factor, and introduce elite reservation strategy and adaptive mutation operator to improve convergence;

[0013] S4. Design an improved hybrid solving algorithm, update the dynamic carbon emission factor and load response plan in real time, construct a "prediction-optimization-feedback" closed loop, correct the prediction error of photovoltaic / wind power output based on historical data, and adjust the next period scheduling strategy based on actual energy consumption data of users.

[0014] The method for establishing a virtual power plant architecture comprising distributed power sources, energy storage systems, flexible loads and a carbon trading mechanism comprises the following steps:

[0015] S111. Integrate distributed power sources, integrate distributed power sources comprehensively, and distinguish different types such as wind power generation and photovoltaic power generation in detail, and establish accurate power output models according to their physical characteristics;

[0016] For wind power generation, accurately measure parameters such as wind speed and wind direction, and calculate the power generation power in real time through aerodynamic formula and equipment characteristic parameters; the wind power generation power calculation formula is:

[0017] ;

[0018] Pw is the wind power; Vw is the wind speed, Vci is the cut-in wind speed, Vr is the rated wind speed, Vco is the cut-out wind speed, Pr is the rated power;

[0019] For photovoltaic power generation, the comprehensive light intensity, temperature, photovoltaic cell conversion efficiency and other factors are used to calculate the power output by using the physical model of photovoltaic cell. The photovoltaic power calculation formula is:

[0020] ;

[0021] Ppv is the photovoltaic power; I is the comprehensive light intensity; T is the temperature; A is the area of the photovoltaic cell, is the temperature coefficient of the photovoltaic cell, is the reference temperature;

[0022] S112. Build the energy storage system, deeply analyze its charging and discharging characteristics, self-discharge rate, life attenuation law, etc., and plan the capacity configuration and charging and discharging strategy;

[0023] The state of charge (SOC) calculation formula of the energy storage system is:

[0024] ;

[0025] and are the charging and discharging efficiencies, respectively; and are the charging and discharging powers, respectively; C is the energy storage capacity, is the time interval; when charging, the SOC increases; when discharging, the SOC decreases;

[0026] The life attenuation of the energy storage system is measured by the number of charging and discharging cycles. With the increase of the number of charging and discharging cycles, the capacity will gradually attenuate, and the life attenuation model is: ;

[0027] wherein is the capacity after cycles; is the initial capacity; is the capacity attenuation coefficient, which is different for different types of energy storage systems The values are different; when planning the capacity configuration, consider the life attenuation to ensure that the energy storage system meets the regulation requirements of the virtual power plant within its service life;

[0028] S113. Classified management of flexible load, according to the power consumption characteristics of industrial, commercial and residential users, establish detailed load regulation, clear adjustable range, adjustment speed and response time;

[0029] Industrial adjustable load combines production process and production plan, optimizes power consumption period and power under the premise of not affecting production;

[0030] Commercial adjustable load adjusts the running state of air conditioning, lighting and other equipment according to business hours and power consumption peak and valley;

[0031] Residential intelligent load through smart home system, within the range acceptable to users, intelligent control of water heater, electric vehicle charging and other equipment, realize peak clipping and valley filling, improve the stability of power system;

[0032] S114. Introduce carbon trading mechanism, real-time track carbon market price fluctuation, policy dynamics, develop flexible carbon trading strategy; carbon trading mechanism provides economic incentive for low-carbon operation of virtual power plant, virtual power plant obtains certain carbon quota, in the operation process, if the actual carbon emission is lower than the quota, the remaining quota can be sold in the carbon market to obtain income; if it exceeds the quota, additional carbon emission rights need to be purchased to avoid punishment; the calculation formula of carbon trading cost is:

[0033] ;

[0034] Wherein, The carbon trading cost of virtual power plant; Real-time tracking of carbon market price fluctuation; The total amount of actual carbon emission of virtual power plant; The allocated carbon quota;

[0035] When the actual carbon emission exceeds the carbon quota, the virtual power plant needs to buy additional carbon quota in the carbon market, resulting in carbon trading cost; if the actual carbon emission is lower than the carbon quota, the excess carbon quota can be sold to obtain income;

[0036] By predicting the price trend of carbon market, buy carbon quota in advance when the price is low, reserve certain carbon emission rights; when the price is high, reduce carbon emission by optimizing power generation combination, increasing the proportion of renewable energy power generation, etc., sell the excess carbon quota to obtain maximum economic benefit; At the same time, pay attention to policy dynamics, adjust the operation strategy of virtual power plant in time to ensure that it meets the carbon emission supervision requirements.

[0037] The method of designing a dynamic carbon emission factor coupled with an energy cost response mechanism comprises:

[0038] S121. Calculate the energy cost, and the energy cost calculation formula is: ;

[0039] wherein, is the power generation of the first energy, is the unit power generation cost of the first energy;

[0040] The cost of different energy sources is different, such as the cost of thermal power generation covering fuel cost, equipment maintenance cost, environmental protection cost, etc.; wind power generation mainly involves equipment investment and operation and maintenance cost; the cost of photovoltaic power generation is related to photovoltaic equipment cost, conversion efficiency and light resources;

[0041] S122. Determine the dynamic carbon emission factor, and the dynamic carbon emission factor reflects the carbon emission of unit power generation, and changes with time, energy structure, technological progress, etc.; when calculating, the energy type, power generation technology, regional environmental policy, etc. are comprehensively considered, and the dynamic carbon emission factor is determined by a multiple linear regression model, and the multiple linear regression model is represented as:

[0042] ;

[0043] wherein, is the dynamic carbon emission factor; is a variable affecting the carbon emission factor, such as the proportion of a certain energy in the total energy, the application degree of carbon capture and storage technology, etc., is a regression coefficient, is an error term; when a region vigorously promotes renewable energy, the proportion of renewable energy increases, if is negative, then will decrease, indicating that the carbon emission per unit power generation is reduced;

[0044] S123. Coupling carbon emission and energy cost, coupling the dynamic carbon emission factor and the energy cost, constructing a response mechanism, and obtaining a comprehensive cost index :

[0045] ;

[0046] wherein and are weight coefficients, reflecting the importance of energy cost and carbon emission, and the virtual power plant takes as the optimization target, considering energy cost and carbon emission in power generation dispatching decision-making, realizing the collaborative optimization of low carbon and economy.

[0047] The method of forming a dispatch signal with carbon flow as the core by quantifying the carbon emission intensity of each link of source-network-load-storage includes:

[0048] S131. Distributed power carbon emission intensity quantification, distributed power covers multiple types, accurate quantification of the carbon emission intensity of distributed power can provide basic data for subsequent dispatching decisions; the carbon emission intensity of distributed power is relatively high, and the carbon emission intensity is determined according to the carbon content of coal, the combustion efficiency, and the application of carbon capture and storage (CCS) technology in the power generation process, and the formula is: ;

[0049] is the carbon content of coal; is the proportion converted into carbon dioxide after combustion; is the power generation efficiency, is the proportion of carbon captured by carbon capture and storage technology to the total carbon emission; is the carbon emission intensity of coal-fired power generation; is the conversion factor of carbon to carbon dioxide;

[0050] For wind power, there is almost no carbon emission in the power generation operation stage, but a certain amount of carbon emission will be generated in the manufacturing, installation, maintenance and decommissioning process of the whole life cycle of the equipment, and the carbon emission intensity of the wind power is ;

[0051] is the carbon emission intensity of wind power; is the total carbon emission of the whole life cycle of wind power equipment; is the expected total power generation;

[0052] Photovoltaic power generation has little carbon emission in the operation stage, and the carbon emission of the manufacturing, transportation, installation and recycling process of photovoltaic components needs to be considered, and the carbon emission intensity of the photovoltaic power generation is: ;

[0053] is the carbon emission intensity of photovoltaic power generation, is the total carbon emission of the whole life cycle of photovoltaic power generation, is the total power generation;

[0054] S132. Quantification of carbon emission intensity of energy storage system, the carbon emission intensity of energy storage system is closely related to the power source used for charging and discharging, and the carbon emission intensity during charging of energy storage system is:

[0055] ;

[0056] is the carbon emission intensity during charging of energy storage system; is the total carbon emission intensity, for charging efficiency;

[0057] In the discharging phase, its carbon intensity is: ;

[0058] for the carbon intensity of the energy storage system in the discharging phase, for considering the discharging efficiency;

[0059] By accurately quantifying the carbon intensity of the energy storage system in different phases, it helps to optimize its charging and discharging strategy and reduce overall carbon emissions;

[0060] S133. Quantify the carbon intensity of the power grid. The carbon emissions of the power grid in the process of power transmission and distribution are mainly caused by line loss and auxiliary equipment power consumption. The carbon intensity of the power grid is ;

[0061] for the carbon intensity of the power grid; for the total carbon emissions of the power grid; for the total power transmission capacity;

[0062] S134. Quantify the carbon intensity of the load side. The carbon intensity of the load side depends on the composition of the power consumed, so the carbon intensity of the load side is: ;

[0063] for the total power consumption of a certain area load; for the power from distributed power sources;

[0064] S135. Form a dispatch signal centered on carbon flow. Integrate the carbon intensity and power changes of each link to build a dispatch signal index centered on carbon flow, with the formula:

[0065] ;

[0066] where, is the dispatch signal index; is the total power generation of various distributed power sources, is the power loss of the power grid, and are the charging and discharging power of the energy storage system, is the load power;

[0067] The virtual power plant makes power generation and power consumption scheduling decisions according to the size of the dispatch signal index . When is larger, it indicates that the current combination of power generation and power consumption results in higher carbon emissions, and the virtual power plant will take a series of adjustment measures.

[0068] The multi-objective optimization model in S2 includes:

[0069] The operation cost includes distributed power generation cost, energy storage system charging and discharging cost, carbon trading cost, etc., and the calculation formula is: ;

[0070] is the operation cost; is the energy procurement cost; is the equipment maintenance cost;

[0071] The abandoned wind and light phenomenon is caused by unbalanced power supply and demand, limited grid accommodation capacity or insufficient energy storage, etc. The calculation formula of abandoned wind and light quantity is:

[0072] ;

[0073] is the abandoned wind / light quantity; is the wind power generation capacity; is the photovoltaic power generation capacity; is the actual utilization of wind power generation; is the actual utilization of photovoltaic power generation;

[0074] The user satisfaction is determined by the analytic hierarchy process (AHP), which divides the user electricity demand into different levels of indicators, constructs a judgment matrix to calculate the weight of each indicator, and calculates the comprehensive satisfaction through the user's satisfaction score of each indicator: ;

[0075] is the comprehensive satisfaction; is the satisfaction score of each indicator ; is the user satisfaction index weight of each indicator .

[0076] The constraint conditions of the multi-objective optimization model include power balance constraint, unit ramp rate constraint, energy storage SOC limit, etc.

[0077] The power balance constraint is:

[0078] is the power generation of the th distributed power source, is the adjustment power of the flexible load;

[0079] The unit ramp rate constraint is: ;

[0080] and are the power generation of the th Upward and downward ramp rate limits for the unit; For the first Taiwanese unit The rate of increase at any given moment; For the first Taiwanese unit The rate of increase at any given moment;

[0081] The energy storage SOC constraint is: ;

[0082] This represents the minimum state of charge for energy storage. This represents the maximum state of charge of the energy storage.

[0083] and This refers to the state of charge range that ensures the safe and stable operation of the energy storage system. During the dispatching process, it is determined based on the energy storage system's... By assessing the energy storage system's condition and rationally arranging charging and discharging strategies, overcharging or over-discharging can be avoided, thus extending the system's lifespan.

[0084] The non-convex nonlinear functions in the multi-objective optimization model in S2 are rendered using a piecewise linearization method.

[0085] Non-convex nonlinear functions in multi-objective optimization models are handled using a piecewise linearization method. This involves dividing the function's domain into multiple small intervals and approximating the original function with a linear function within each interval. For a cost function... Define its domain Divided into:

[0086] ;

[0087] In each small interval Internally, using linear functions approximate ;

[0088] in, , ;

[0089] By reasonably dividing the intervals, while ensuring accuracy, the difficulty of solving the model can be simplified, and optimization problems can be solved using methods such as linear programming.

[0090] S3 includes:

[0091] S311. Hybrid solution strategy, a hybrid solution strategy combining improved genetic algorithm and commercial solver, the improved genetic algorithm has global search ability, which is used for global rough search to quickly find the approximate feasible solution region; then the commercial solver has efficient local search ability, which is used for local accurate search in the feasible solution region found by the improved genetic algorithm to improve the solution accuracy.

[0092] S312. Hybrid coding strategy, a hybrid coding strategy is adopted for decision variables, real number coding is adopted for continuous variables such as power generation of distributed power supply and charging and discharging power of energy storage; binary coding is adopted for discrete variables such as switching state of equipment; meanwhile, corresponding coding conversion and genetic operation rules are designed to ensure that continuous variables and discrete variables can evolve cooperatively in the operation process of genetic algorithm.

[0093] S313. Multi-objective conversion into single objective, multi-objective is converted into single objective through preference factor weighting, and the single objective function after conversion is:

[0094] ;

[0095] for operation cost, for wind / photovoltaic curtailment amount, for user satisfaction, 、 、 , for preference factor respectively, and ;

[0096] S314. Convergence improvement, an elite reservation strategy is introduced, in the evolution process of each generation, the optimal individual in the current population is reserved to directly enter the next generation, so that the loss of optimal solution is avoided; meanwhile, an adaptive mutation operator is adopted, and the mutation probability is adjusted according to the fitness variance of the population, and the formula is: wherein is the basic mutation probability, is the adjustment coefficient, is the maximum fitness variance;

[0097] When the fitness variance of the population is small, the mutation probability is increased, the search range is expanded, and the algorithm is prevented from falling into local optimum; when the fitness variance is large, the mutation probability is reduced, the convergence speed is accelerated, and the algorithm efficiency is improved.

[0098] The S4 comprises:

[0099] S411. Real-time updating and closed-loop construction;

[0100] The dynamic carbon emission factor is updated, the distributed power output, the energy storage system SOC, the flexible load power and the carbon market price are collected in real time through the smart meter and the sensor, the collected data is brought into the multiple linear regression model in S122, the model parameters are updated in real time through the real-time data, and the carbon emission intensity of unit power generation is dynamically adjusted;

[0101] The closed-loop optimization process is adopted. An LSTM neural network is used to take historical light intensity, wind speed and load curve as input, and output wind power / photovoltaic power output prediction value and load demand prediction value in future period. Based on the prediction result, a mixed solution algorithm of S3 is called to generate a scheduling plan. The objective function is a multi-objective optimization model of S2. The constraint conditions include power balance, unit climbing rate, energy storage SOC limit, etc. The actual output is compared with the prediction value, and the error is calculated.

[0102] S412. Prediction error correction, statistical photovoltaic / wind power prediction error, if the average of photovoltaic / wind power prediction error is equal to 0, it indicates that the model has deviation; the historical error data is trained by using the random forest algorithm to generate a compensation term to correct the systematic error, the error compensation term is added to the original prediction value to obtain the corrected output prediction correction, and the deviation tolerance is set in the scheduling model to avoid constraint violation caused by prediction error;

[0103] S413. Scheduling strategy adjustment:

[0104] S4131. User energy consumption analysis and load regulation, based on K-means clustering, users are divided into industrial, commercial and residential three categories, and the power consumption mode is extracted, the user satisfaction index weight of each index is calculated by the analytic hierarchy process (AHP) of S2 , dynamically adjust the flexible load strategy;

[0105] The flexible load strategy includes:

[0106] Industrial load, reduce the power of non-critical process in peak period, the constraint is the adjustment range allowed by production process;

[0107] Residential load, delay electric vehicle charging to low valley period through smart home to meet user acceptable response time;

[0108] S4132. Energy storage and power coordination optimization, based on power correction charging and discharging plan, if photovoltaic power is excessive, increase the energy storage charging power; if the carbon market price rises, preferentially replace thermal power by energy storage discharging to reduce carbon trading cost;

[0109] S414. Real-time error correction based, based on the latest data update short-term scheduling, retrain the LSTM model using all-day error data after each day ends, adjust network parameters, when the single objective function of continuous When the change rate is less than 1%, the determination algorithm converges, and the optimal scheduling scheme is output.

[0110] In summary, due to the adoption of the technical solutions described above, the beneficial technical effects of the present application are:

[0111] The deep synergy between low-carbon transformation and economic operation can automatically prioritize the scheduling of low-carbon power (such as wind power and photovoltaic power) by quantifying the carbon emission intensity of the virtual power plant at all stages and constructing a carbon flow scheduling signal, thereby reducing the dependence on high-carbon thermal power. The coupling mechanism of dynamic carbon emission factors and energy costs, combined with the economic incentives of carbon trading, enables the virtual power plant to dynamically optimize power generation combinations and load strategies in response to carbon price fluctuations and changes in energy structure, thereby achieving a simultaneous reduction in operating costs while achieving the "double carbon" goal and economic benefits.

[0112] The bidirectional reinforcement of renewable energy consumption and system stability significantly improves the system's ability to accommodate intermittent power sources through a "source-storage-load" collaborative interaction mechanism: the energy storage system responds in real time to power fluctuations, charging when there is excess photovoltaic / wind power and discharging during peak load periods, thereby mitigating power supply and demand imbalances; industrial, commercial, and residential flexible loads are classified and regulated (such as industrial peak-shifting electricity and residential smart electricity control), forming a "peak clipping and valley filling" collaborative chain with energy storage, effectively reducing the phenomenon of wind and light curtailment, and enhancing the efficiency and stability of the power system in accommodating clean energy and reducing the risk of power grid fluctuations.

[0113] The balanced optimization of intelligent scheduling and user experience significantly improves the efficiency and accuracy of multi-objective optimization (operating costs, wind and light curtailment, and user satisfaction) through a hybrid algorithm that combines genetic algorithms and commercial solvers, mixed coding, and "global-local" collaborative search, ensuring the real-time generation of high-quality scheduling schemes. The "prediction-optimization-feedback" closed-loop mechanism, combined with real-time data acquisition and error correction, enables the system to dynamically adapt to changes in power output, load demand, and carbon prices; the user satisfaction model based on the analytic hierarchy process considers both power supply reliability and user experience during load regulation, achieving a dynamic balance between system optimization and user demand, and promoting the upgrading of virtual power plants towards low-carbon, efficient, and user-friendly. BRIEF DESCRIPTION OF DRAWINGS

[0114] Figure 1 A flowchart of a virtual power plant source-load interaction optimization scheduling model and solution algorithm incorporating a low-carbon response mechanism;

[0115] Figure 2 A flowchart of a method for establishing a virtual power plant architecture;

[0116] Figure 3 A flowchart of a method for designing a dynamic carbon emission factor and energy cost coupling response mechanism;

[0117] Figure 4 S1. Constructing a low-carbon response-based virtual power plant source-load interaction optimization scheduling model, including the following steps:

[0118] Figure 5 S3. Designing an improved hybrid solving algorithm, for the established multi-objective optimization model, combining the improved genetic algorithm and the hybrid solving strategy of the commercial solver, adopting the mixed coding mode, converting the multi-objective into a single objective through the preference factor weighting, and introducing the elite reservation strategy and the adaptive mutation operator to improve the convergence;

[0119] Figure 6 S4. Dynamic feedback and iterative optimization, through real-time updating of dynamic carbon emission factor and load response plan, constructing a "prediction-optimization-feedback" closed loop, using historical data to correct the prediction error of photovoltaic / wind power output, and adjusting the next period scheduling strategy based on the actual energy consumption data of users. DETAILED DESCRIPTION

[0120] In order to make the purpose, technical scheme and advantages of the invention clearer, the invention will be further described in detail below with examples. It should be understood that the specific examples described herein are only used to explain the invention and not to limit the invention.

[0121] As shown in Figure 1 , a low-carbon response-based virtual power plant source-load interaction optimization scheduling model and solving algorithm, including the following steps:

[0122] S1. Constructing a low-carbon response-based virtual power plant source-load interaction optimization scheduling model, including the following steps:

[0123] S2. Establishing a multi-objective optimization model and a constraint system, based on the established framework, constructing a multi-objective optimization model, and the non-convex nonlinear function in the multi-objective optimization model is processed by the piecewise linearization method;

[0124] S3. Designing an improved hybrid solving algorithm, for the established multi-objective optimization model, combining the improved genetic algorithm and the hybrid solving strategy of the commercial solver, adopting the mixed coding mode, converting the multi-objective into a single objective through the preference factor weighting, and introducing the elite reservation strategy and the adaptive mutation operator to improve the convergence;

[0125] S4. Dynamic feedback and iterative optimization, through real-time updating of dynamic carbon emission factor and load response plan, constructing a "prediction-optimization-feedback" closed loop, using historical data to correct the prediction error of photovoltaic / wind power output, and adjusting the next period scheduling strategy based on the actual energy consumption data of users.

[0126] As shown in Figure 2 , the method for establishing the virtual power plant architecture is:

[0127] S111. Integrating distributed power sources, integrating distributed power sources comprehensively, and distinguishing different types such as wind power generation and photovoltaic power generation in detail, and establishing accurate power output models according to their physical characteristics;

[0128] For wind power generation, the parameters such as wind speed and wind direction are accurately measured, and the power generation power is calculated in real time through aerodynamic formula and equipment characteristic parameters. The wind power generation power calculation formula is:

[0129] ;

[0130] For wind power generation power; For wind speed, For cut-in wind speed, For rated wind speed, For cut-out wind speed, For rated power;

[0131] For photovoltaic power generation, factors such as light intensity, temperature, and photovoltaic cell conversion efficiency are integrated, and the power output is calculated by using the physical model of photovoltaic cell. The photovoltaic power generation power calculation formula is:

[0132] ;

[0133] For photovoltaic power generation power; For comprehensive light intensity; For temperature; For photovoltaic cell area, is the temperature coefficient of the photovoltaic cell, is the reference temperature;

[0134] S112. Constructing energy storage system, deeply analyzing its charging and discharging characteristics, self-discharge rate, life attenuation law, etc., planning capacity configuration and charging and discharging strategy;

[0135] The state of charge (SOC) calculation formula of the energy storage system is:

[0136] ;

[0137] and are the charging and discharging efficiencies; and are the charging and discharging powers; is the energy storage capacity, is the time interval; when charging, increases the SOC; when discharging, decreases the SOC;

[0138] The life attenuation of the energy storage system is measured by the number of charging and discharging cycles. With the increase of the number of charging and discharging cycles, the capacity will gradually attenuate, and the life attenuation model is: ;

[0139] wherein is the capacity after the nth cycle; is the initial capacity; is the initial capacity; is the capacity attenuation coefficient, different types of energy storage systems have different values; when planning the capacity configuration, consider the life attenuation to ensure that the energy storage system meets the regulation requirements of the virtual power plant within its service life;

[0140] S113. Flexible loads are classified and managed, and detailed load regulation is established according to the power consumption characteristics of industrial, commercial, and residential users, and the adjustable range, adjustment speed, and response time are clearly defined;

[0141] Industrial adjustable load combines production process and production plan to optimize power consumption period and power without affecting production;

[0142] Commercial adjustable load adjusts the operation state of air conditioning, lighting and other equipment according to business hours and power consumption peak and valley;

[0143] Residential intelligent load controls water heaters, electric vehicle charging and other equipment within the user's acceptable range through a smart home system to achieve peak clipping and valley filling and improve the stability of the power system;

[0144] S114. Introduce carbon trading mechanism, real-time track carbon market price fluctuation, policy dynamics, develop flexible carbon trading strategy; carbon trading mechanism provides economic incentive for low-carbon operation of virtual power plant, virtual power plant obtains certain carbon quota, in operation process, if actual carbon emission is lower than quota, can sell surplus quota to obtain income in carbon market; if exceeds quota, need to buy additional carbon emission right to avoid punishment; the calculation formula of carbon trading cost is:

[0145] ;

[0146] Among them, is the carbon trading cost of virtual power plant; is the real-time tracking of carbon market price fluctuation; is the total actual carbon emission of virtual power plant; is the allocated carbon quota;

[0147] When the actual carbon emission exceeds the carbon quota, the virtual power plant needs to buy additional carbon quota in the carbon market, resulting in carbon trading cost; if the actual carbon emission is lower than the carbon quota, the excess carbon quota can be sold to obtain income;

[0148] ​By predicting the trend of carbon market prices, purchasing carbon quotas in advance when prices are low, and reserving certain carbon emission rights, when prices are high, reducing carbon emissions by optimizing power generation combinations, increasing the proportion of renewable energy power generation, and selling excess carbon quotas to maximize economic benefits. At the same time, pay attention to policy dynamics and adjust the operation strategy of the virtual power plant in a timely manner to ensure compliance with carbon emission regulation requirements.

[0149] As Figure 3 shown, the method of designing a dynamic carbon emission factor coupled with an energy cost response mechanism includes:

[0150] S121. Calculate the energy cost, and the energy cost calculation formula is: ;

[0151] Wherein, is the power generation of the first energy, is the unit power generation cost of the first energy; The cost of different energies is different, such as the cost of thermal power generation, which includes fuel cost, equipment maintenance cost, environmental protection cost, etc.; wind power generation mainly involves equipment investment and operation and maintenance cost; the cost of photovoltaic power generation is related to photovoltaic equipment cost, conversion efficiency and light resources;

[0152]

[0153] S122. Determine the dynamic carbon emission factor, and the dynamic carbon emission factor reflects the carbon emission of unit power generation, and changes with time, energy structure, technological progress, etc.; when calculating, the energy type, power generation technology, regional environmental policy, etc. are considered comprehensively; the dynamic carbon emission factor is determined by a multiple linear regression model, and the multiple linear regression model is represented as:

[0154] ;

[0155] Wherein, is the dynamic carbon emission factor; is a variable that affects the carbon emission factor, such as the proportion of a certain energy in the total energy, the application degree of carbon capture and storage technology, is the regression coefficient, is the error term; when the region vigorously promotes renewable energy, the proportion of renewable energy increases, if is negative, will decrease, indicating that the carbon emission per unit of power generation is reduced;

[0156] S123. Coupling carbon emission and energy cost, coupling dynamic carbon emission factor and energy cost, building response mechanism, and getting comprehensive cost index : ​​

[0157] ;

[0158] wherein, and are weight coefficients, reflecting the importance of energy cost and carbon emission, and the virtual power plant takes as the optimization target, considering energy cost and carbon emission in power generation scheduling decision-making, and realizing the collaborative optimization of low carbon and economy.

[0159] At a certain moment, if the wind power cost is relatively low but the carbon emission factor is relatively high, and the photovoltaic power cost is slightly high but the carbon emission factor is extremely low, the virtual power plant can balance energy cost and carbon emission by adjusting the power generation combination, and realize the collaborative optimization of low carbon and economy. The weight coefficients and can be adjusted according to the operation target of the virtual power plant, market environment and policy requirements. For example, in the period when carbon emission reduction requirement is strict, the weight of is appropriately increased to strengthen the control of carbon emission; when the energy cost pressure is relatively large, the weight of is increased to preferentially reduce the energy cost.

[0160] As shown in Figure 4 , the method of forming a dispatching signal with carbon flow as the core by quantifying the carbon emission intensity of source-grid-load-storage includes:

[0161] S131. Distributed power carbon emission intensity quantification, distributed power covers multiple types (thermal power, wind power, photovoltaic power), accurate quantification of the carbon emission intensity of distributed power can provide basic data for subsequent scheduling decision-making; the carbon emission intensity of thermal power in distributed power is relatively high, which is determined according to the carbon content of coal, the combustion efficiency and the application of carbon capture and storage (CCS) technology in the power generation process, and the formula is:

[0162] ;

[0163] is the carbon content of coal; is the proportion of carbon dioxide converted after combustion; is the power generation efficiency, is the proportion of carbon captured by carbon capture and storage technology in the total carbon emission; is the carbon emission intensity of coal-fired power generation; is the conversion coefficient of carbon to carbon dioxide, generally takes ;

[0164] For wind power generation, there is almost no carbon emission in the power generation stage, but there is a certain amount of carbon emission in the manufacturing, installation, maintenance and decommissioning process of the equipment throughout its life cycle, so the carbon emission intensity of wind power generation is ;

[0165] the carbon emission intensity of wind power generation; the total carbon emission of wind power generation throughout its life cycle; the total expected power generation of the equipment;

[0166] Photovoltaic power generation has little carbon emission in the operation stage, and the carbon emission of photovoltaic component manufacturing, transportation, installation and recycling process needs to be considered, so the carbon emission intensity of photovoltaic power generation is ;

[0167] the carbon emission intensity of photovoltaic power generation, the total carbon emission of photovoltaic power generation throughout its life cycle, the total power generation;

[0168] S132. Quantify the carbon emission intensity of energy storage system, the carbon emission intensity of energy storage system is closely related to the power source used for its charging and discharging, and the carbon emission intensity during the charging process of energy storage is ;

[0169] the carbon emission intensity of energy storage system charging; the total carbon emission intensity (thermal power generation, wind power generation, photovoltaic power generation), the charging efficiency;

[0170] In the discharging stage, the carbon emission intensity is ;

[0171] the carbon emission intensity of energy storage system in the discharging stage, considering the discharging efficiency;

[0172] By accurately quantifying the carbon emission intensity of energy storage system in different stages, it is helpful to optimize its charging and discharging strategy and reduce the overall carbon emission;

[0173] S133. Quantify the carbon emission intensity of power grid, the carbon emission of power grid in the process of power transmission and distribution is mainly caused by line loss and auxiliary equipment power consumption, and the carbon emission intensity of power grid is ;

[0174] the carbon emission intensity of power grid; the total carbon emission of power grid; the total power transmission capacity;

[0175] the total carbon emission of power grid In relation to the grid loss, the grid loss is related to the line resistance, the current size, the power transmission distance and other factors. In combination with the carbon emission intensity in the power generation process, the carbon emission caused by the grid loss is calculated. At the same time, the power consumption of auxiliary equipment (such as the cooling system of the substation, communication equipment and the like) will also cause carbon emission. By comprehensively considering these factors, the carbon emission intensity of the grid can be more accurately quantified;

[0176] S134. Quantify the carbon emission intensity of the load side. The carbon emission intensity of the load side depends on the source composition of the consumed power. Therefore, the carbon emission intensity of the load side is:

[0177] is the total power consumed by the load of a certain region; is the power from the distributed power supply;

[0178] S135. Form a dispatch signal with carbon flow as the core. The carbon emission intensity and power change of each link are comprehensively considered to construct a dispatch signal index with carbon flow as the core, and the formula is:

[0179]

[0180] wherein, is the dispatch signal index; is the total power of various distributed power supplies, is the grid loss power, and are the charging and discharging power of the energy storage system, respectively, is the load power;

[0181] The virtual power plant makes power generation and power consumption scheduling decisions according to the size of the dispatch signal index . When is larger, it indicates that the current combination of power generation and power consumption leads to higher carbon emission, and the virtual power plant will take a series of adjustment measures. In practice, the distributed power supply with low carbon emission intensity (such as wind power and photovoltaic power) is arranged first to generate power, and the proportion of thermal power generation is reduced; the load is reasonably distributed to guide the flexible load to increase power consumption when the low-carbon power is sufficient, such as guiding the air conditioning, lighting and other equipment of commercial users to use photovoltaic power more during the day when the light is sufficient; the charging and discharging strategy of the energy storage system is optimized to charge during the period of low-carbon power surplus (such as the period of large wind power generation) and discharge during the peak power demand and high carbon emission (such as when the proportion of thermal power supply is large at night), so as to reduce the value of and realize the coordinated optimization scheduling of low carbon and economy. By continuously monitoring and adjusting the power distribution of each link, the value of is always kept at a low level, and the virtual power plant is promoted to run in the direction of low carbon and high efficiency.

[0182] ​​The multi-objective optimization model includes achieving the lowest virtual power plant operating cost, the minimum amount of wind / solar curtailment, and the highest user satisfaction.

[0183] Operating costs include distributed power generation costs, energy storage system charging and discharging costs, and carbon trading costs, and are calculated using the following formula: ;

[0184] Operating costs; For energy procurement costs; For equipment maintenance costs;

[0185] Wind and solar power curtailment is a waste of renewable energy caused by factors such as imbalance between power supply and demand, limited grid capacity, or insufficient energy storage. The formula for calculating the amount of wind and solar power curtailment is:

[0186] ;

[0187] This refers to the amount of wind / solar energy curtailed. The amount of electricity that can be generated by wind power; The amount of electricity that can be generated by photovoltaic power generation; This refers to the actual amount of wind power generated. This refers to the actual amount of photovoltaic power generated that is utilized.

[0188] User satisfaction was determined using the Analytic Hierarchy Process (AHP). User electricity needs were categorized into different levels of indicators, such as power supply reliability, power quality, and electricity cost. A judgment matrix was constructed to calculate the weight of each indicator. Overall satisfaction was calculated based on user ratings for each indicator. ;

[0189] For overall satisfaction; For each indicator Satisfaction rating; For each indicator The weight of user satisfaction metrics.

[0190] The constraints of the multi-objective optimization model include power balance constraints, unit ramp rate constraints, and energy storage SOC limits.

[0191] The power balance constraint is: ;

[0192] For the first The power generation capacity of this type of distributed power source The regulating power for flexible loads;

[0193] The unit's ramp rate constraint is: ;

[0194] and are the upward and downward ramp rate limits of the unit, respectively; are the upward and downward ramp rates of the unit at time t, respectively; are the upward and downward ramp rates of the unit at time t, respectively; are the upward and downward ramp rates of the unit at time t, respectively;

[0195] The energy storage SOC constraint is: ;

[0196] is the minimum state of charge of the energy storage; is the maximum state of charge of the energy storage;

[0197] and is the state of charge range that ensures safe and stable operation of the energy storage system, in the dispatch process, according to the state of charge of the energy storage system, the charging and discharging strategy is reasonably arranged to avoid overcharging or overdischarging, and the service life of the energy storage system is prolonged.

[0198] The non-convex nonlinear function in the multi-objective optimization model adopts a piecewise linearization method:

[0199] The non-convex nonlinear function existing in the multi-objective optimization model, such as part of the cost function, power-efficiency relationship function, etc., these functions increase the difficulty of solving, adopt piecewise linearization method to process, divide the function definition domain into multiple small intervals, and approximate the original function with a linear function in each small interval; for a cost function , its definition domain is divided into: ;

[0200] In each small interval , a linear function is used to approximate ;

[0201] wherein, , ;

[0202] By reasonably dividing the intervals, the model solving difficulty is simplified while the accuracy is ensured, and linear programming method can be used to solve the optimization problem.

[0203] Figure 5 As shown in the figure, the S3 includes:

[0204] ​​​​​S311. Hybrid solution strategy, a hybrid solution strategy combining improved genetic algorithm and commercial solvers (such as CPLEX, Gurobi), improved genetic algorithm has global search ability, first use it for global rough search, quickly find the approximate feasible solution area; then the commercial solver has high local search ability, in the feasible solution area found by the improved genetic algorithm for local accurate search, improve the solution accuracy. For example, in the initial stage, the improved genetic algorithm searches for potential optimal solutions in the solution space through selection, crossover, mutation and other operations; then these solutions are used as the initial solution of the commercial solver, and the commercial solver uses its optimization algorithm (such as simplex method, interior point method) to further optimize and obtain more accurate optimal solution.

[0205] S312. Hybrid encoding method, for decision variables, use hybrid encoding method, for continuous variables such as distributed power generation power, energy storage charging and discharging power, use real number encoding; for discrete variables such as device switching state, use binary encoding. For example, for wind power generation power, directly use real number to represent its size; for whether a device is put into operation, use 0 to represent off and 1 to represent on. This encoding method can better adapt to the characteristics of different types of variables and improve the search efficiency of the algorithm. At the same time, design corresponding encoding conversion and genetic operation rules to ensure that continuous variables and discrete variables can evolve together in the genetic algorithm operation process.

[0206] S313. Multi-objective conversion to single objective, convert multi-objective to single objective by weighting preference factors; for example, the objectives are (operation cost), (wind / photovoltaic curtailment amount), (user satisfaction), and the preference factors are , , , and , then the single objective function after conversion is ; users can adjust the preference factors according to actual needs, such as increasing the weight of carbon emission related objectives (affecting operation cost and wind / photovoltaic curtailment amount) when carbon emission reduction requirement is strict; increase the weight of user satisfaction objective when power supply is tight; through multiple tests and analysis, determine the appropriate combination of preference factors to obtain the optimization result that meets the actual needs.

[0207] S314. Improve convergence, introduce elite preservation strategy, in each generation evolution process, keep the best individual in the current population, which directly enters the next generation, to avoid loss of optimal solution; at the same time, use adaptive mutation operator, mutation probability is adjusted according to the fitness variance of the population, the formula is: where is the basic mutation probability, is the adjustment coefficient, is the maximum fitness variance;

[0208] When the fitness variance of the population is small, the mutation probability is increased to expand the search range and avoid the algorithm from falling into local optimum; when the fitness variance is large, the mutation probability is reduced to accelerate the convergence speed and improve the algorithm efficiency.

[0209] As shown in Figure 6 , the S4 comprises:

[0210] S411. Real-time updating and closed-loop construction;

[0211] The dynamic carbon emission factor is updated, the distributed power output, the energy storage system SOC, the flexible load power and the carbon market price are collected in real time through the smart meter and the sensor, the collected data is brought into the multiple linear regression model in S122, the model parameters are updated in real time, and the carbon emission intensity of unit power generation is dynamically adjusted;

[0212] The closed-loop optimization process adopts an LSTM neural network, historical light intensity, wind speed and load curve are used as input, and wind power / photovoltaic output prediction value and load demand prediction value in future period are output, based on the prediction results, the mixed solution algorithm of S3 is called to generate a scheduling plan, the objective function is the multi-objective optimization model of S2 (the lowest operation cost, the smallest abandoned wind / light, and the highest user satisfaction), the constraint conditions include power balance, unit climbing rate, energy storage SOC limit, etc., the actual output is compared with the prediction value, and the error is calculated;

[0213] S412. Prediction error correction, the photovoltaic / wind power prediction error is counted, if the average of the photovoltaic / wind power prediction error is equal to 0, it indicates that the model has deviation; the random forest algorithm is used to train the historical error data (input is meteorological factor and time, and output is error), a compensation term is generated to correct systematic error, the error compensation term is added to the original prediction value to obtain the corrected output prediction correction, and a deviation tolerance is set in the scheduling model to avoid constraint violation caused by prediction error;

[0214] S413. Scheduling strategy adjustment:

[0215] S4131. User energy consumption analysis and load regulation, based on K-means clustering, users are divided into industrial, commercial and residential categories, power consumption mode (such as industrial load peak period) is extracted, user satisfaction index weight of each index is calculated through the analytic hierarchy process (AHP) of S2 , and the flexible load strategy is dynamically adjusted;

[0216] The flexible load strategy includes:

[0217] Industrial load, reduce non-key process power in peak period, the constraint is the adjustment range allowed by production process;

[0218] Residential load, delay electric vehicle charging to off-peak period through smart home, meet user acceptable response time;

[0219] S4132. Energy storage and power supply optimization, based on power correction charging and discharging plan, if photovoltaic power is surplus, increase energy storage charging power; if carbon market price rises, preferentially replace thermal power through energy storage discharging to reduce carbon trading cost (based on S114 carbon trading cost formula);

[0220] S414. Based on real-time error correction, update short-term scheduling based on the latest data, retrain LSTM model using all-day error data after the end of the day, adjust network parameters, when the single objective function of continuous iteration Change rate less than 1%, determine that the algorithm converges, output optimal scheduling scheme.

[0221] The above is the preferred embodiment of the invention content, and does not limit the invention content, any modification, equivalent replacement and improvement within the spirit and principle of the invention content should be included in the protection scope of the invention content.

Claims

1. A virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response, characterized in that, Includes the following steps: S1. Construct a low-carbon dispatch framework for a virtual power plant with multi-energy synergy, establish a virtual power plant architecture, and design a response mechanism that couples dynamic carbon emission factors with energy costs; by quantifying the carbon emission intensity of each link from source to grid to load to storage, form a dispatch signal with carbon flow as the core, and achieve synergistic optimization of low-carbon goals and economic efficiency. S2 establishes a multi-objective optimization model and constraint system. Based on the established framework, a multi-objective optimization model is constructed, and the non-convex nonlinear functions in the multi-objective optimization model are processed using a piecewise linearization method. S3. Design an improved hybrid solution algorithm. For the established multi-objective optimization model, combine the improved genetic algorithm and commercial solver hybrid solution strategy, adopt a hybrid encoding method, transform multi-objective into single objective through preference factor weighting, and introduce an elite retention strategy and adaptive mutation operator to improve convergence. S4. Dynamic feedback and iterative optimization: By updating the dynamic carbon emission factor and load response plan in real time, a "prediction-optimization-feedback" closed loop is constructed. Historical data is used to correct the prediction error of photovoltaic / wind power output, and the scheduling strategy for the next cycle is adjusted based on the actual energy consumption data of users.

2. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, The method for establishing a virtual power plant architecture that includes distributed power sources, energy storage systems, flexible loads, and carbon trading mechanisms: S111. Integrate distributed power sources, comprehensively integrate distributed power sources, distinguish in detail between different types of wind power generation and photovoltaic power generation, and establish accurate power output models based on their physical characteristics; For wind power generation, precise measurements of wind speed and direction parameters are used, and the power generation is calculated in real time using aerodynamic formulas and equipment characteristic parameters. The formula for calculating wind power generation is: ； This refers to the power output of wind power generation. For wind speed, To cut into wind speed, Rated wind speed, To cut off the wind speed, Rated power; For photovoltaic (PV) power generation, considering factors such as irradiance, temperature, and PV cell conversion efficiency, the power output is calculated using a PV cell physical model. The formula for calculating PV power generation is: ； Photovoltaic power generation; To take into account the overall light intensity; For temperature; For the area of ​​photovoltaic cells, It is the temperature coefficient of photovoltaic cells. This is a reference temperature; S112. Construct an energy storage system, analyze its charging and discharging characteristics, self-discharge rate, and lifespan decay patterns in depth, and plan capacity configuration and charging and discharging strategies. The formula for calculating the state of charge (SOC) of an energy storage system is as follows: ； and These are charging and discharging efficiency, respectively. and It refers to charging and discharging power; It refers to energy storage capacity. It is a time interval; during charging, Increases SOC; during discharge, Reduce SOC; The lifespan degradation of an energy storage system is measured by the number of charge-discharge cycles. As the number of charge-discharge cycles increases, the capacity gradually decreases. The lifespan degradation model is as follows: ; in It was through The capacity after the next iteration; It is the initial capacity; It is the capacity decay coefficient, which varies depending on the type of energy storage system. The values ​​are different; when planning capacity configuration, lifetime degradation should be considered to ensure that the energy storage system meets the regulation needs of the virtual power plant throughout its lifespan; S113. Classify and manage flexible loads, and establish detailed load regulation based on the electricity consumption characteristics of industrial, commercial and residential users, specifying the adjustable range, regulation speed and response time; Industrial adjustable loads are combined with production processes and plans to optimize power consumption periods and power without affecting production. Commercial adjustable loads adjust the operating status of air conditioning and lighting equipment according to business hours and peak and off-peak electricity consumption. Residential smart loads, through smart home systems, can intelligently regulate water heaters and electric vehicle charging equipment within acceptable limits for users, thereby achieving peak shaving and valley filling and improving the stability of the power system. S114. Introduce a carbon trading mechanism to track carbon market price fluctuations and policy dynamics in real time, and formulate flexible carbon trading strategies; the carbon trading mechanism provides economic incentives for the low-carbon operation of virtual power plants. Virtual power plants obtain a certain amount of carbon allowances. During operation, if actual carbon emissions are lower than the allowances, the remaining allowances can be sold in the carbon market to generate revenue; if they exceed the allowances, additional carbon emission rights must be purchased to avoid penalties; the formula for calculating carbon trading costs is: ； in, Carbon trading costs for virtual power plants; To track carbon market price fluctuations in real time; This represents the actual total carbon emissions from the virtual power plant. The allocated carbon quota; When actual carbon emissions exceed carbon allowances, virtual power plants need to purchase additional carbon allowances in the carbon market, incurring carbon trading costs; if actual carbon emissions are lower than carbon allowances, they can sell the excess carbon allowances to generate revenue. By predicting carbon market price trends, companies can purchase carbon allowances in advance when prices are low to reserve a certain amount of carbon emission rights; when prices are high, they can reduce carbon emissions by optimizing power generation mix and increasing the proportion of renewable energy generation, and sell excess carbon allowances to obtain maximum economic benefits; at the same time, they should pay attention to policy dynamics and adjust the operation strategy of virtual power plants in a timely manner to ensure compliance with carbon emission regulatory requirements.

3. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, The method for designing a response mechanism that couples dynamic carbon emission factors with energy costs includes: S121. Calculate energy costs. The formula for calculating energy costs is: ； in, It is the first The power generation capacity of this energy source It is the first The unit cost of generating electricity from this type of energy source; S122. Determine the dynamic carbon emission factor, which reflects the carbon emissions per unit of electricity generated. The dynamic carbon emission factor is determined through a multiple linear regression model, which is expressed as follows: ； in, As a dynamic carbon emission factor; These are variables that influence carbon emission factors, such as the proportion of a certain energy source in total energy and the degree of application of carbon capture and storage technologies. It is the regression coefficient. This is an error term; when a region vigorously promotes renewable energy and the proportion of renewable energy increases, if If it is negative, then This will decrease, indicating a reduction in carbon emissions per unit of electricity generated; S123. Coupling carbon emissions and energy costs: Dynamic carbon emission factors are coupled with energy costs to construct a response mechanism, and a comprehensive cost index is obtained through a weighted summation method. : ； in, and These are weighting coefficients that reflect the degree of importance attached to energy costs and carbon emissions. Virtual power plants are based on... To optimize the objectives, energy costs and carbon emissions are considered simultaneously in power generation dispatch decisions, achieving synergistic optimization of low-carbon and economic goals.

4. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, The method for generating a scheduling signal centered on carbon flow by quantifying the carbon emission intensity of each link in the source-grid-load-storage system includes: S131. Quantification of Carbon Emission Intensity of Distributed Power Generation: Distributed power generation encompasses various types. Accurately quantifying the carbon emission intensity of distributed power generation can provide fundamental data for subsequent dispatch decisions. Among distributed power generation, thermal power generation has a relatively high carbon emission intensity. The carbon emission intensity is determined based on the carbon content of coal, combustion efficiency, and the application of carbon capture and storage technologies during the power generation process. The formula is: ； The carbon content of coal; This represents the proportion of carbon dioxide converted after combustion. For power generation efficiency, The proportion of carbon captured by carbon capture and storage technologies to total carbon emissions; Carbon emission intensity for coal-fired power generation; This is the conversion factor for converting carbon to carbon dioxide; While wind power generation produces almost no carbon emissions during the power generation and operation phase, the manufacturing, installation, maintenance, and decommissioning processes throughout the equipment's entire lifecycle generate certain carbon emissions. Therefore, its carbon emission intensity... ; Carbon emission intensity of wind power generation; This refers to the total carbon emissions throughout the entire lifecycle of wind power generation equipment. This represents the equipment's expected total power generation. Photovoltaic power generation has low carbon emissions during the operation phase. However, carbon emissions from the manufacturing, transportation, installation, and recycling of photovoltaic modules must be considered. Therefore, its carbon emission intensity is: ； Carbon emission intensity of photovoltaic power generation, This refers to the total carbon emissions throughout the entire lifecycle of photovoltaic power generation. Total power generation; S132. Quantify the carbon emission intensity of energy storage systems. The carbon emission intensity of energy storage systems is closely related to the power source used for charging and discharging. The carbon emission intensity during the energy storage charging process is: ； Carbon intensity when charging an energy storage system; Total carbon intensity, For charging efficiency; During the discharge phase, its carbon emission intensity is: ； The carbon emission intensity during the discharge phase of the energy storage system. To consider discharge efficiency; Precisely quantifying the carbon emission intensity of energy storage systems at different stages helps to optimize their charging and discharging strategies and reduce overall carbon emissions. S133. Quantifying the carbon emission intensity of the power grid: The carbon emissions of the power grid during power transmission and distribution mainly originate from line losses and auxiliary equipment power consumption. ； The carbon emission intensity of the power grid; Total carbon emissions from the power grid; This represents the total transmitted power. S134. Quantify the carbon emission intensity on the load side. The carbon emission intensity on the load side depends on the composition of the sources of electricity consumed. Therefore, the carbon emission intensity on the load side is: ； The total electricity consumed by the load in a certain area; The electricity comes from distributed power sources; S135. A scheduling signal centered on carbon flow is formed, integrating carbon emission intensity and power changes across various stages, to construct a scheduling signal index centered on carbon flow. The formula is as follows: ； in, For scheduling signal indicators; It is the sum of the generating power of all types of distributed power sources. It is the power loss of the power grid. and These are the charging and discharging power of the energy storage system. For load power; Virtual power plants based on dispatch signal indicators The size of the power generation and consumption dispatch decisions is used to determine the power supply and consumption schedule; when A higher value indicates that the current combination of power generation and consumption is causing high carbon emissions, and the virtual power plant will take a series of adjustment measures.

5. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, The multi-objective optimization model in S2 includes: Operating costs include distributed power generation costs, energy storage system charging and discharging costs, and carbon trading costs. The calculation formula is as follows: ； Operating costs; For energy procurement costs; For equipment maintenance costs; Wind and solar power curtailment is a waste of renewable energy caused by imbalances in power supply and demand, limited grid capacity, or insufficient energy storage. The formula for calculating the amount of wind and solar power curtailment is: ； This refers to the amount of wind / solar energy curtailed. The amount of electricity that can be generated by wind power; The amount of electricity that can be generated by photovoltaic power generation; This refers to the actual amount of wind power generated. This refers to the actual amount of photovoltaic power generated that is utilized. User satisfaction was determined using the analytic hierarchy process (AHP). User electricity demand was categorized into different levels of indicators, a judgment matrix was constructed to calculate the weight of each indicator, and overall satisfaction was calculated based on user ratings for each indicator. ； For overall satisfaction; For each indicator Satisfaction rating; For each indicator The weight of user satisfaction metrics.

6. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 5, characterized in that, The constraints of the multi-objective optimization model include power balance constraints, unit ramp rate constraints, and energy storage SOC limits. The power balance constraint is: ； For the first The power generation capacity of this type of distributed power source The regulating power for flexible loads; The unit's ramp rate constraint is: ； and The first Upward and downward ramp rate limits for the unit; For the first Taiwanese unit The rate of increase at any given moment; For the first Taiwanese unit The rate of increase at any given moment; The energy storage SOC constraint is: ； This represents the minimum state of charge for energy storage. This represents the maximum state of charge of the energy storage. and This refers to the state of charge range that ensures the safe and stable operation of the energy storage system. During the dispatching process, it is determined based on the energy storage system's... By assessing the energy storage system's condition and rationally arranging charging and discharging strategies, overcharging or over-discharging can be avoided, thus extending the system's lifespan.

7. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, The non-convex nonlinear functions in the multi-objective optimization model in S2 are rendered using a piecewise linearization method. Non-convex nonlinear functions in multi-objective optimization models are handled using a piecewise linearization method. This involves dividing the function's domain into multiple small intervals and approximating the original function with a linear function within each interval. For a cost function... Define its domain Divided into: ； In each small interval Internally, using linear functions approximate ; in, , ; By reasonably dividing the intervals, while ensuring accuracy, the difficulty of solving the model can be simplified, and the optimization problem can be solved using linear programming methods.

8. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, S3 includes: S311. Hybrid solution strategy: Combining the improved genetic algorithm with the commercial solver, the improved genetic algorithm has global search capability, which is first used to perform a global coarse search to quickly find the approximate feasible solution region; then the commercial solver has efficient local search capability, which performs a local precise search within the feasible solution region found by the improved genetic algorithm to improve the solution accuracy. S312. Hybrid coding method: A hybrid coding method is adopted for decision variables. For continuous variables such as the power generation of distributed power sources and the charging and discharging power of energy storage, real number coding is used; for discrete variables such as the switching status of equipment, binary coding is used. At the same time, corresponding coding conversion and genetic operation rules are designed to ensure that continuous variables and discrete variables can co-evolve during the operation of the genetic algorithm. S313. Multi-objective transformation into a single objective: Multi-objective functions are transformed into a single objective function by weighting with preference factors. The transformed single-objective function is: ； For operating costs, For wind / solar curtailment, For user satisfaction, , , , for the preference factors respectively, and ; S314. To improve convergence, an elite retention strategy is introduced. In each generation of evolution, the best individual in the current population is retained and directly enters the next generation, avoiding the loss of the optimal solution. Simultaneously, an adaptive mutation operator is used, with the mutation probability adjusted according to the population's fitness variance, as shown in the formula: ,in It is the basic mutation probability. It is an adjustment factor. It is the maximum fitness variance; When the fitness variance of the population is small, increase the mutation probability to expand the search range and avoid the algorithm getting stuck in local optima; when the fitness variance is large, decrease the mutation probability to speed up the convergence speed and improve the algorithm efficiency.

9. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 1, characterized in that, S4 includes: S411. Real-time updates and closed-loop construction; Update the dynamic carbon emission factor by collecting data in real time from distributed power generation output, energy storage system SOC, flexible load power and carbon market price through smart meters and sensors. Input the collected data into the multiple linear regression model in S122 and update the model parameters through real-time data to dynamically adjust the carbon emission intensity per unit of power generation. The closed-loop optimization process uses an LSTM neural network, taking historical light intensity, wind speed, and load curves as inputs, and outputs predicted wind / solar power output and load demand for future periods. Based on the prediction results, the hybrid solution algorithm of S3 is called to generate a scheduling plan. The objective function is the multi-objective optimization model of S2, and the constraints include power balance, unit ramp rate, and energy storage SOC limit. The actual output is compared with the predicted value to calculate the error. S412. Prediction Error Correction: Statistically analyze the prediction errors of photovoltaic / wind power. If the mean of the prediction error of photovoltaic / wind power is equal to 0, it indicates that there is a bias in the model. Use the random forest algorithm to train historical error data, generate compensation terms to correct systematic errors, and superimpose the error compensation terms onto the original prediction values ​​to obtain the corrected power output prediction. Set a deviation tolerance in the scheduling model to avoid constraint violations due to prediction errors. S413. Scheduling strategy adjustment; S414. Based on real-time error correction, the short-term schedule is updated based on the latest data. At the end of each day, the LSTM model is retrained using the error data from the entire day, and the network parameters are adjusted. When continuous... Single objective function in the next iteration When the rate of change is less than 1%, the algorithm is considered to have converged, and the optimal scheduling scheme is output.

10. The virtual power plant source-load interaction optimization scheduling model and solution method based on low-carbon response as described in claim 9, characterized in that, The scheduling strategy adjustment includes: S4131. User Energy Consumption Analysis and Load Control: Based on K-means clustering, users are divided into three categories: industrial, commercial, and residential. Electricity consumption patterns are extracted, and the weights of user satisfaction indicators for each indicator are calculated using the analytic hierarchy process (AHP) in S2. Dynamically adjust flexible load strategies; Flexible load strategies include: Industrial load, during peak hours, reduces non-critical process power, constrained to the adjustment range allowed by the production process; To reduce residential load, smart home technology can delay electric vehicle charging to off-peak hours, ensuring acceptable response times for users. S4132. Energy storage and power supply synergistic optimization: Based on power correction charging and discharging plans, if photovoltaic power is in excess, increase energy storage charging power; if carbon market prices rise, prioritize energy storage discharge to replace thermal power and reduce carbon trading costs.

Images