A virtual power plant regulation method and system for realizing new energy consumption

Abstract

The application relates to the technical field of virtual power plants, and discloses a virtual power plant regulation method and system for realizing new energy consumption, which comprises a multi-source heterogeneous data acquisition module, an intelligent prediction analysis module, a resource aggregation modeling module, an optimization decision module, a blockchain cooperation module and a digital twin evaluation module; the new energy output and load demand are coupled and predicted in multiple time scales through a deep space-time convolutional neural network, the fluctuation and intermittence characteristics of the new energy can be described, the short-term and ultra-short-term prediction accuracy is improved, the wind and light curtailment rate and load reduction risk are reduced, the virtual power plant dispatching plan and the actual operation condition are highly matched, the flexible resource feature matrix is constructed, the distributed energy storage, interruptible load and electric vehicle multi-resource are finely modeled and aggregated, the virtual unit capable of being efficiently dispatched is formed, the resource utilization efficiency is improved and the overall dispatching cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual power plants, in particular to a virtual power plant regulation method and system for realizing new energy consumption. BACKGROUND

[0002] Under the background of energy internet, new renewable energy generation will be connected to the power grid on a large scale to reduce the proportion of traditional fossil fuel power generation and achieve low-carbon and clean development of the power grid. With the increasing proportion of renewable energy generation, the capacity of power grid regulation demand and the climbing rate of regulation units must be improved, and the problem of new energy consumption arises. Virtual power plants can aggregate controllable loads, energy storage and other demand-side resources to participate in peak regulation, frequency regulation and new energy consumption in multiple scenarios, and currently have broad development potential.

[0003] Currently, in the field of new energy consumption, the regulation process of virtual power plants faces multiple technical bottlenecks. Due to the volatility and intermittency of new energy output, existing prediction methods cannot accurately depict the dynamic characteristics of multiple time scales, resulting in large deviations in short-term and ultra-short-term prediction, so that the dispatching plan formulated by the virtual power plant does not match the actual operating conditions well, causing wind and light curtailment and insufficient load supply.

[0004] Therefore, the present application provides a virtual power plant regulation method and system for realizing new energy consumption to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a virtual power plant regulation method and system for realizing new energy consumption to solve the problem of low matching degree between the dispatching plan formulated by the virtual power plant and the actual operating conditions, causing wind and light curtailment and insufficient load supply.

[0006] To achieve the above purpose, the present application provides the following technical solutions: a virtual power plant regulation method and system for realizing new energy consumption, comprising:

[0007] S1, real-time collection of new energy output data, load demand data and power grid dispatching instructions to generate new energy dynamic consumption data set;

[0008] S2, multi-time scale prediction processing of the new energy dynamic consumption data set based on a deep spatio-temporal convolutional neural network, outputting a new energy short-term output prediction curve, an ultra-short-term fluctuation characteristic spectrum and a load elasticity response interval;

[0009] S3, constructing a virtual power plant resource pool model, mapping distributed energy storage, interruptible load and electric vehicle clusters into virtual machine group parameters to generate a flexibility resource feature matrix;

[0010] S4, multi-objective optimization matching is performed based on the new energy short-term output prediction curve and the flexible resource characteristic matrix, an improved genetic algorithm is used to solve the optimal scheduling strategy, and a day-ahead regulation plan is generated;

[0011] S5, according to the ultra-short-term fluctuation characteristic spectrum and the real-time instruction, the day-ahead regulation plan is dynamically adjusted through a fuzzy adaptive controller, and a minute-level regulation instruction sequence is output;

[0012] S6, a risk assessment model is established, and the dynamic climbing margin and risk index are calculated based on the load elastic response interval and the minute-level regulation instruction sequence;

[0013] S7, when the risk index exceeds the threshold value, the blockchain mechanism is activated to call the cross-regional standby resource, and a collaborative consumption compensation scheme is generated;

[0014] S8, the regulation instruction sequence and the collaborative consumption compensation scheme are executed through the smart contract, and the regulation effect is verified;

[0015] S9, based on the verification of the regulation effect, a digital twin evaluation system is constructed to generate a consumption efficiency report and optimization suggestions.

[0016] Preferably, the new energy dynamic consumption data set in S1 includes new energy generation potential data, multi-element user load data, and market and grid constraint data, and the construction process includes:

[0017] S11, the new energy generation potential data of photovoltaic power stations and wind farms are collected through meteorological satellite remote sensing data and field station power prediction systems;

[0018] S12, the power consumption characteristic data of industrial, commercial and residential loads are collected through intelligent electric meter measurement systems and demand response management platforms;

[0019] S13, the day-ahead and real-time electricity price signals and cross-section transmission constraint data are obtained through the power market trading system and the grid energy management system.

[0020] Preferably, the deep spatio-temporal convolutional neural network in S2 adopts a three-layer heterogeneous structure:

[0021] The first layer of spatio-temporal convolution module extracts the geographical distribution characteristics of new energy output;

[0022] The second layer of gate recurrent unit captures the output fluctuation time sequence correlation;

[0023] The third layer of self-attention mechanism fuses meteorological mutation factors and load transfer characteristics.

[0024] Preferably, the flexible resource characteristic matrix in S3 contains four-dimensional parameters:

[0025] The first dimension represents the charge-discharge power interval and the capacity state of the energy storage system;

[0026] The second dimension quantifies the duration of interruptible load reduction and the compensation cost;

[0027] The third dimension aggregates the V2G response rate and the power boundary of the electric vehicle cluster;

[0028] The fourth dimension integrates the rapid start-stop characteristics and the carbon emission coefficient of the small gas turbine;

[0029] The flexible resource feature matrix is represented by an 8xm matrix as follows:

[0030] ;

[0031] wherein represents the flexible resource feature matrix of the virtual power plant, represents the total number of time periods in the future scheduling period, 、 respectively represent the minimum and maximum charge-discharge power of the energy storage system at time period , represent the remaining capacity state of the energy storage system at time period , represent the compensation cost required for reducing one unit of interruptible load at time period , represent the maximum discharge power that the electric vehicle cluster can provide at time period , represent the upward ramping rate of the small gas turbine at time period , which represents the maximum value of the output that can be increased per minute, represent the downward ramping rate of the small gas turbine at time period , which represents the maximum value of the output that can be reduced per minute, represent the carbon emission coefficient generated by the small gas turbine at time period when emitting one unit of power.

[0032] Preferably, the improved genetic algorithm in S4 comprises a triple optimization mechanism:

[0033] The first optimization objective is to minimize the new energy curtailment rate and the load reduction amount;

[0034] The second optimization objective is to balance the calling cost of each flexible resource;

[0035] The third optimization objective is to meet the power grid safety constraint equation.

[0036] Preferably, the fuzzy adaptive controller of S5 adopts a double closed-loop control structure:

[0037] The outer ring corrects the power allocation proportion based on the ultra-short-term prediction error;

[0038] The inner ring dynamically adjusts the control parameter setting interval according to the frequency deviation signal.

[0039] Preferably, the calculation method of the dynamic ramping margin index in S6 is:

[0040] In a 15-minute time window, the matching degree difference between the actual output fluctuation rate of the new energy and the maximum ramping capacity of the virtual power plant is calculated, and the transmission margin of the power grid section is superimposed to form a three-dimensional evaluation vector, and the mathematical expression is:

[0041] ;

[0042] Wherein represents the dynamic ramping margin index at time , represents the maximum upward ramping power that the virtual power plant can provide at time , represents the maximum downward ramping power that the virtual power plant can provide at time , represents the change rate of the actual output of the new energy in the short time window near time , represents the transmission power margin of the key section of the power grid at time .

[0043] Preferably, the blockchain mechanism in S7 adopts the DPoS optimization algorithm:

[0044] Select the regional energy hub node as the accountant to verify the credibility of the standby resource;

[0045] Through the cross-chain communication protocol, the smart contract triggering of multi-virtual power plant collaborative consumption is realized.

[0046] Preferably, the consumption efficiency report and optimization suggestion in S9 are generated through a multi-dimensional weighted evaluation algorithm, including five core indicators:

[0047] New energy consumption completion rate, power grid regulation support contribution degree, carbon emission intensity reduction, resource calling cost benefit ratio, and user power consumption satisfaction.

[0048] Preferably, the system comprises:

[0049] A multi-source heterogeneous data acquisition module acquires new energy generation potential data through a meteorological satellite receiving unit, collects multi-element user load data using a wideband measurement unit, and receives market and power grid constraint data through a market interface protocol converter, and outputs new energy dynamic consumption data set;

[0050] The intelligent prediction analysis module receives the new energy dynamic consumption data set, generates a new energy short-term output prediction curve through a deep spatio-temporal convolutional neural network calculation engine, and outputs an ultra-short-term fluctuation characteristic spectrum and a load elasticity response interval using a fluctuation characteristic extraction algorithm library.

[0051] The resource aggregation modeling module receives the multi-element user load data and new energy power generation potential data, generates schedulable virtual machine group parameters through a virtual machine group equivalent converter, and constructs a flexibility resource feature matrix using a flexibility resource digital twin.

[0052] The optimization decision module receives the new energy short-term output prediction curve and the flexibility resource feature matrix, performs optimization matching processing through a multi-objective solver, generates a day-ahead regulation plan and a minute-level regulation instruction sequence using a rolling correction controller.

[0053] The blockchain collaboration module receives the regulation instruction sequence, performs instruction verification through a cross-region consensus verification node, activates cross-region backup resource calls using an intelligent contract execution engine, and outputs a collaborative consumption compensation scheme.

[0054] The digital twin evaluation module receives execution effect data of the minute-level regulation instruction sequence and collaborative consumption compensation scheme data, displays the regulation process through a three-dimensional visualization platform, and generates a consumption efficiency evaluation report and a strategy optimization suggestion using a strategy self-optimization feedback unit.

[0055] The strategy optimization suggestion is fed back to the rolling correction controller in the optimization decision module to adaptively adjust the generation logic of subsequent regulation instructions. Advantages

[0056] Compared with the prior art, the present application provides a virtual power plant regulation method and system for new energy consumption, which has the following advantages:

[0057] 1. In the present application, the deep spatio-temporal convolutional neural network is used to perform multi-time scale coupled prediction on new energy output and load demand, which can depict the volatility and intermittency characteristics of new energy, improve the short-term and ultra-short-term prediction accuracy, reduce the wind and light curtailment rate and load reduction risk, and ensure the high matching of virtual power plant dispatching plan and actual operation condition.

[0058] 2. In the present application, the flexibility resource feature matrix is constructed, and the distributed energy storage, interruptible load, and electric vehicle multi-element resources are finely modeled and aggregated, which can form a virtual machine group that can be efficiently dispatched, fully tap and coordinate the response potential of various flexible resources, improve resource utilization efficiency and reduce overall dispatching cost.

[0059] 3. In the present application, by introducing the rolling optimization correction mechanism and the blockchain collaborative accommodation scheme, the power grid instruction and the new energy power fluctuation can be responded in real time, the scheduling strategy can be adaptively adjusted, the cross-regional standby resources can be quickly and reliably called, and the multi-dimensional efficiency feedback and closed-loop optimization can be formed relying on the digital twin evaluation system, thereby enhancing the safety and stability of power grid operation and the overall new energy accommodation level. BRIEF DESCRIPTION OF DRAWINGS

[0060] Fig. 1 A flowchart of a virtual power plant regulation method for realizing new energy accommodation according to the present application;

[0061] Fig. 2 A framework diagram of a virtual power plant regulation system for realizing new energy accommodation according to the present application. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0063] Specific embodiments: please refer to Figs. 1-2 A virtual power plant regulation method and system for realizing new energy accommodation, comprising:

[0064] S1, real-time collection of new energy output data, load demand data and power grid dispatching instructions to generate new energy dynamic accommodation data set;

[0065] S2, multi-time scale prediction processing of the new energy dynamic accommodation data set based on a deep spatio-temporal convolutional neural network, outputting a new energy short-term output prediction curve, an ultra-short-term fluctuation characteristic spectrum and a load elastic response interval, the specific operation comprising: the network adopts a three-layer heterogeneous structure, the first layer of spatio-temporal convolution module extracts the spatial geographical distribution characteristics of new energy output, the second layer of gate recurrent unit captures the time sequence correlation of the output fluctuation, and the third layer of self-attention mechanism fuses meteorological mutation factors and load transfer characteristics, through network training and reasoning, outputting a new energy short-term output prediction curve, an ultra-short-term fluctuation characteristic spectrum and a load elastic response interval;

[0066] S3, constructing a virtual power plant resource pool model, mapping distributed energy storage, interruptible load and electric vehicle cluster into virtual machine group parameters to generate a flexibility resource characteristic matrix;

[0067] S4, multi-objective optimization matching based on new energy short-term output prediction curve and flexible resource characteristic matrix, taking the new energy short-term output prediction curve as the basic prediction data and the flexible resource characteristic matrix as the optimization adjustment boundary, constructing a multi-objective optimization model containing three optimization objectives and power grid safety constraints, solving the optimal scheduling strategy by using an improved genetic algorithm, and generating a day-ahead regulation plan;

[0068] S5, rolling correction according to the ultra-short-term fluctuation characteristic spectrum and real-time instructions, dynamically adjusting the day-ahead regulation plan through a fuzzy adaptive controller, and outputting a minute-level regulation instruction sequence, the specific operation contents of which include:

[0069] S51, input and error calculation: real-time reading of the new energy power minute-level fluctuation prediction value represented by the ultra-short-term fluctuation characteristic spectrum, and collecting the real-time scheduling instruction of the power grid, calculating the error between the new energy power prediction value and the actual value, and the deviation amount of the current regulation plan and the real-time instruction;

[0070] S52, fuzzy processing: taking the prediction error and instruction deviation amount calculated above as input variables, mapping them into corresponding fuzzy language variables and membership degrees according to the preset fuzzy rule base;

[0071] S53, fuzzy reasoning and decision-making: outer loop control: based on the fuzzy amount of the ultra-short-term prediction error, applying fuzzy reasoning rules to dynamically correct the priority and proportion of various flexible resources in power distribution;

[0072] Inner loop control: according to the fuzzy amount of the power grid frequency deviation signal, applying another set of fuzzy reasoning rules to adaptively adjust the parameter setting interval of the controller to optimize the response speed and stability of the control system;

[0073] S54, defuzzification and instruction generation: converting the fuzzy amount output after fuzzy reasoning and decision-making into an accurate power adjustment amount through a defuzzification algorithm, and applying this adjustment amount to the day-ahead regulation plan to generate accurate power control instructions for the next few minutes to tens of minutes;

[0074] S55, instruction sequence output: arranging the generated accurate power control instructions in chronological order and packaging them into a minute-level regulation instruction sequence that can be issued, and outputting them to the execution unit;

[0075] S6, establishing a risk assessment model, calculating the dynamic ramping margin and risk index based on the load elasticity response interval and the minute-level regulation instruction sequence;

[0076] S7, when the risk index exceeds the threshold, activating the blockchain mechanism to call cross-regional backup resources and generate a collaborative consumption compensation scheme;

[0077] S8, execute the regulation instruction sequence and the coordinated consumption compensation scheme through the smart contract, and verify the regulation effect;

[0078] S9, based on the verification of the regulation effect, build a digital twin evaluation system to generate a consumption efficiency report and optimization suggestions.

[0079] The new energy dynamic consumption data set in S1 includes new energy generation potential data, multi-user load data, and market and grid constraint data, and the construction process includes:

[0080] S11, collect new energy generation potential data of photovoltaic power stations and wind farms through meteorological satellite remote sensing data and field station power prediction system;

[0081] S12, collect the power consumption characteristic data of industrial, commercial and residential loads through the intelligent electric meter measurement system and the demand response management platform;

[0082] S13, obtain the day-ahead and real-time price signals and cross-section transmission constraint data through the power market trading system and the grid energy management system.

[0083] The deep spatio-temporal convolutional neural network in S2 adopts a three-layer heterogeneous structure:

[0084] The first layer of spatio-temporal convolution module extracts the geographical distribution characteristics of new energy output;

[0085] The second layer of gate recurrent unit captures the time series correlation of output fluctuations;

[0086] The third layer of self-attention mechanism integrates meteorological mutation factors and load transfer characteristics.

[0087] The flexibility resource feature matrix in S3 contains four-dimensional parameters:

[0088] The first dimension represents the charge and discharge power interval and capacity state of the energy storage system;

[0089] The second dimension quantifies the duration of interruptible load reduction and compensation cost;

[0090] The third dimension aggregates the V2G response rate and power boundary of the electric vehicle cluster;

[0091] The fourth dimension integrates the rapid start-stop characteristics and carbon emission coefficient of small gas turbines;

[0092] Among them, the flexibility resource feature matrix is represented by an 8xm matrix:

[0093] ;

[0094] Among them represents the flexibility resource feature matrix of the virtual power plant, Total time period of future scheduling period, 、 Total time period of future scheduling period, Minimum and maximum charge-discharge power of energy storage system, Total time period of future scheduling period, Residual capacity state of energy storage system, Total time period of future scheduling period, Compensation cost required for curtailed interruptible load, Total time period of future scheduling period, Maximum discharge power that electric vehicle cluster can provide, Total time period of future scheduling period, Upward ramping rate of small gas turbine, indicating the maximum value that it can increase output per minute, Total time period of future scheduling period, Downward ramping rate of small gas turbine, indicating the maximum value that it can reduce output per minute, Total time period of future scheduling period, Carbon emission coefficient generated by small gas turbine for unit electric quantity.

[0095] The improved genetic algorithm in S4 includes a triple optimization mechanism:

[0096] The first optimization objective is to minimize new energy curtailment rate and load curtailment;

[0097] The second optimization objective is to balance the calling cost of each flexible resource;

[0098] The third optimization objective is to meet the power grid safety constraint equation;

[0099] The specific operation steps of the improved genetic algorithm include:

[0100] S41, chromosome coding and population initialization: adopt real number coding method to construct chromosome, each chromosome contains scheduling plan sequence of flexible resource, and initial population is randomly generated;

[0101] S42, multi-objective fitness function calculation: according to the triple optimization mechanism, the fitness function is constructed, and its expression is:

[0102] ;

[0103] Wherein is the individual fitness value, the larger the value, the better the solution, 、 、 is the weight coefficient, represents new energy curtailment rate, represents load curtailment, represents resource calling cost standard deviation, represents power grid constraint violation degree;

[0104] Weighting coefficient , , Dynamically adjust based on the real-time operating status of the power grid:

[0105] Increase when the predicted output of new energy sources is higher. When the cost of resource allocation differs significantly, it increases. Increase when the power grid operation approaches the safety boundary ;

[0106] S43. Non-dominated ranking and crowding calculation: Perform non-dominated ranking on individuals in the population, divide them into Pareto ranks, and calculate the crowding distance between individuals within the same rank.

[0107] S44. Elite Retention and Selection Operation: A binary tournament selection method is adopted, with the tournament size set to 2. The Pareto level of an individual is the primary selection criterion, and crowding distance is the secondary criterion. Excellent individuals are retained to the next generation of the population.

[0108] S45. Crossover and mutation operations: A new offspring population is generated using a simulated binary crossover operator and a polynomial mutation operator, where the crossover distribution index is set to 20 and the mutation distribution index is set to 20.

[0109] S46. Iteration Termination Judgment: When the maximum number of iterations is reached, output the Pareto optimal solution set; otherwise, return to S42 to continue iterating.

[0110] The S5's fuzzy adaptive controller employs a dual closed-loop control structure:

[0111] The outer ring adjusts the power allocation ratio based on ultra-short-term prediction errors;

[0112] The inner loop dynamically adjusts the control parameter tuning range based on the frequency deviation signal.

[0113] The calculation method for the dynamic ramp margin index in S6 is as follows:

[0114] Within a 15-minute time window, the difference in matching degree between the actual output fluctuation rate of new energy and the maximum ramping capacity of the virtual power plant is calculated, and the transmission margin of the grid section is superimposed to form a three-dimensional evaluation vector, the mathematical expression of which is:

[0115] ;

[0116] in Indicates at time The dynamic ramp margin index, Indicates the virtual power plant at any time The maximum uphill power that can be provided represents the maximum downward ramping power that the virtual power plant can provide at time represents the change rate of the actual output of new energy within a short time window represents the transmission power margin of the key section of the power grid at time

[0117] The blockchain mechanism in S7 adopts the DPoS optimization algorithm:

[0118] Select regional energy hub nodes as accountants to verify the credibility of backup resources;

[0119] Through cross-chain communication protocol, the smart contract triggering of multi-virtual power plant collaborative consumption is realized;

[0120] The specific operation steps of the DPoS optimization algorithm include:

[0121] S71, account node election: based on the historical scheduling performance data and communication reliability index of regional energy hub nodes, the comprehensive reputation value of each node is calculated, and the top N nodes with comprehensive reputation value are selected as the accountant node set of the current consensus cycle;

[0122] S72, resource verification and block generation: the calling capacity, response rate and contract validity of cross-regional backup resources are verified by the accountant nodes through multiple rounds of voting, and the verified resource information and scheduling instructions are packaged to generate a new block;

[0123] S73, cross-chain communication triggering: through cross-chain communication relay bridge, the on-chain state of the target virtual power plant is obtained, and when the local blockchain generates a new block containing backup resource calling instructions, the cross-chain smart contract is triggered automatically, and the collaborative consumption request is initiated to the target virtual power plant;

[0124] S74, state synchronization and confirmation: the smart contract executor on the chain of the target virtual power plant receives the request and verifies its legality, and after execution, the confirmation result is fed back to the source blockchain through the cross-chain protocol, and the global state synchronization is completed.

[0125] The consumption efficiency report and optimization suggestion in S9 is generated through a multi-dimensional weighted evaluation algorithm, which includes five core indicators:

[0126] New energy consumption completion rate, power grid regulation support contribution, carbon emission intensity reduction, resource calling cost benefit ratio, and user electricity satisfaction;

[0127] The specific operation steps of the multi-dimensional weighted evaluation algorithm include:

[0128] ​​​​S91, index data standardization processing: the original data of the five core indicators are obtained respectively, the extreme value standardization method is used to normalize each index data to the interval [0, 1], and a standardized index vector is formed;

[0129] S92, combination weight calculation: the subjective weight of each index is calculated by using the analytic hierarchy process, the objective weight of each index is calculated by combining the entropy weight method, and the comprehensive weight coefficient is generated by multiplying the subjective and objective weights through the multiplication integration method;

[0130] S93, comprehensive score calculation: the standardized index vector and the comprehensive weight coefficient are weighted and summed to generate a new energy consumption efficiency comprehensive score, and the calculation formula is:

[0131] ;

[0132] Wherein is the weight of the th index, is the standardized value of the th index;

[0133] S94, evaluation report generation: according to the comprehensive score, the efficiency level is divided, and based on the comparison and analysis of each sub-index score and weight coefficient, optimization suggestions for prediction accuracy, resource scheduling strategy and response instruction execution are generated.

[0134] The system comprises:

[0135] A multi-source heterogeneous data acquisition module acquires new energy power generation potential data through a meteorological satellite receiving unit, collects multi-user load data using a wideband measurement unit, and receives market and power grid constraint data through a market interface protocol converter, and outputs a new energy dynamic consumption data set;

[0136] An intelligent prediction analysis module receives the new energy dynamic consumption data set, generates a new energy short-term output prediction curve through a deep spatio-temporal convolutional neural network calculation engine, and outputs an ultra-short-term fluctuation feature spectrum and a load elasticity response interval using a fluctuation feature extraction algorithm library. The operation process of the fluctuation feature extraction algorithm library comprises:

[0137] I. The new energy power time series data is decomposed by a wavelet packet decomposition algorithm to extract the energy distribution characteristics in different frequency bands;

[0138] II. The energy distribution characteristics are identified by using a spectral clustering algorithm to divide into stable fluctuation mode, gradual fluctuation mode and sudden fluctuation mode;

[0139] III. The duration, change slope and amplitude probability distribution of each fluctuation mode are calculated based on historical data to generate a feature spectrum for describing the fluctuation characteristics;

[0140] The resource aggregation modeling module receives multi-user load data and new energy power generation potential data, generates a dispatchable virtual machine group parameter through a virtual machine group equivalent converter, and constructs a flexibility resource feature matrix by using a flexibility resource digital twin. The specific operation process of the virtual machine group equivalent converter includes:

[0141] I. Feature vector extraction: For various flexibility resources, the upper limit of the adjustable power, the lower limit of the adjustable power, the rated capacity, the maximum ramp rate, the minimum continuous operation time and the response delay time are extracted as the equivalent clustering feature vector;

[0142] II. Dynamic weighted clustering: A k-means clustering algorithm based on Euclidean distance is adopted, and each dimension of the feature vector is dynamically weighted according to the importance and scheduling cost of the resource type, and the distributed resources with similar characteristics are aggregated into several virtual machine groups;

[0143] III. Equivalent parameter mapping: The upper and lower limits of the adjustable power of all resources in each virtual machine group are algebraically superimposed as the total adjustable power range of the virtual machine group; the maximum value of the minimum continuous operation time and the response delay time of the resources in the machine group is respectively taken as the minimum continuous operation time and the response delay time of the virtual machine group;

[0144] The optimization decision module receives the short-term output prediction curve of the new energy and the flexibility resource feature matrix, performs optimization matching processing through a multi-objective solver, generates a day-ahead regulation plan and a minute-level regulation instruction sequence by using a rolling correction controller, and the specific operation process of the rolling correction controller includes:

[0145] I. Error calculation: Receive ultra-short-term fluctuation characteristic spectrum and real-time dispatching instructions of the power grid, calculate the new energy power prediction error and the grid instruction deviation;

[0146] II. Parameter setting: Based on the new energy power prediction error and the grid instruction deviation, the proportional, integral and differential parameters of the fuzzy adaptive controller are dynamically adjusted by using fuzzy logic rules;

[0147] III. Instruction generation: The rolling optimization of the day-ahead regulation plan is performed by using the set fuzzy adaptive controller to generate a minute-level regulation instruction sequence;

[0148] IV. Safety check: The regulation instruction sequence is sent to the power grid safety constraint checking unit for out-of-limit detection to ensure that the instruction meets the system safety operation requirements;

[0149] The blockchain collaboration module receives the regulation instruction sequence, executes instruction verification through a cross-region consensus verification node, and activates a cross-region standby resource call by using a smart contract execution engine, and outputs a collaborative consumption compensation scheme;

[0150] The digital twin evaluation module receives execution effect data of the minute-level regulation instruction sequence and collaborative consumption compensation scheme data, displays the regulation process through a three-dimensional visualization platform, and generates a consumption efficiency evaluation report and a strategy optimization suggestion by using a strategy self-optimization feedback unit.

[0151] The strategy optimization suggestion is fed back to a rolling correction controller in the optimization decision module, and is used for adaptive adjustment of the generation logic of subsequent regulation instructions.

[0152] Based on the above disclosure, in combination with actual applications, the operation steps of the virtual power plant regulation method and system for realizing new energy consumption are as follows:

[0153] Step one, multi-source data sensing and convergence

[0154] The system collects and converges the output data of new energy power generation equipment, the load demand data of multi-element users and the dispatching instruction data issued by the power grid dispatching institution in real time through various sensing and communication units deployed, to jointly form a new energy dynamic consumption data set for subsequent in-depth analysis.

[0155] Step two, multi-time scale collaborative prediction

[0156] The deep spatio-temporal convolutional neural network is used for intelligent analysis of the new energy dynamic consumption data set, and a high-precision new energy short-term output prediction curve, an ultra-short-term fluctuation characteristic spectrum for depicting minute-level fluctuations and a load elasticity response interval are generated in parallel, to provide accurate input basis for optimization decision.

[0157] Step three, flexible resource aggregation modeling

[0158] The virtual power plant resource pool dynamic aggregation model is constructed, the distributed energy storage, interruptible load and electric vehicle cluster heterogeneous resources are mapped into a dispatchable virtual machine group with standardized parameters through an equivalent conversion algorithm, and a flexible resource characteristic matrix fully characterizing the response capability is generated.

[0159] Step four, multi-objective optimization decision generation

[0160] Based on the new energy prediction curve and the flexible resource characteristic matrix, an improved genetic algorithm is used for solving, the algorithm simultaneously optimizes three objectives of consumption maximization, cost economization and grid safety, and finally generates a day-ahead regulation plan of the virtual power plant.

[0161] Step five, rolling correction and real-time instruction generation

[0162] According to the ultra-short-term fluctuation characteristic spectrum and the real-time scheduling instruction of the power grid, the day-ahead plan is dynamically rolled and optimized and corrected at the minute level through a fuzzy adaptive controller. The process specifically includes: first, calculating the new energy power prediction error and the instruction deviation; then, dynamically setting the parameters of the controller using fuzzy logic rules; subsequently, generating a minute-level real-time regulation instruction sequence using the set controller; and finally, sending the instruction sequence to a safety checking unit to ensure that it meets all safety constraint conditions of line power, node voltage and system frequency before output.

[0163] Step six, cross-regional coordinated emergency response

[0164] When the evaluation finds that the local regulation capacity is insufficient, a mechanism based on blockchain consensus is immediately activated to automatically and reliably verify and call cross-regional backup resources through a smart contract, generate and execute an emergency coordinated consumption compensation scheme to cope with extreme working conditions.

[0165] Step seven, whole-process digital twin evaluation and optimization

[0166] A digital twin evaluation system for regulation effect is constructed to comprehensively evaluate multi-dimensional indicators such as new energy consumption rate, power grid support degree, carbon emission reduction benefit, economy and user satisfaction in the whole regulation process, generate a quantitative evaluation report and strategy optimization suggestion, form a closed-loop feedback, and drive continuous optimization of the system.

[0167] It should be noted that in this text, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0168] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for realizing new energy consumption of a virtual power plant, characterized in that: The application relates to a new energy dynamic consumption method based on a deep spatio-temporal convolution neural network. S1, real-time collection of new energy output data, load demand data and power grid dispatching instructions to generate a new energy dynamic consumption data set; S2, multi-time scale prediction processing of the new energy dynamic consumption data set based on a deep spatio-temporal convolution neural network, output of a new energy short-term output prediction curve, an ultra-short-term fluctuation characteristic spectrum and a load elasticity response interval; S3, construction of a virtual power plant resource pool model, mapping of distributed energy storage, interruptible load and electric vehicle clusters into virtual machine group parameters to generate a flexible resource characteristic matrix, the flexible resource characteristic matrix containing four-dimensional parameters: The first dimension represents the charge-discharge power interval and capacity state of the energy storage system; The second dimension quantifies the reduction duration and compensation cost of the interruptible load; The third dimension aggregates the V2G response rate and power boundary of the electric vehicle cluster; The fourth dimension integrates the rapid start-stop characteristics and carbon emission coefficient of the small gas turbine; The flexible resource characteristic matrix is represented by an 8*m matrix as follows: ； wherein represents a flexibility resource feature matrix of the virtual power plant, represents the total number of time periods of the future dispatch cycle, , respectively represent the minimum and maximum charging / discharging power of the energy storage system at time period , represent the remaining capacity state of the energy storage system at time period , represent the compensation cost required for the curtailment of a unit interruptible load at time period , represent the maximum discharging power that the electric vehicle cluster can provide at time period , represent the upward ramping rate of the small gas turbine at time period , which represents the maximum value that the small gas turbine can increase the output per minute, represent the downward ramping rate of the small gas turbine at time period , which represents the maximum value that the small gas turbine can reduce the output per minute, represent the carbon emission coefficient generated by the small gas turbine for a unit of electricity at time period , S4, multi-objective optimization matching based on the new energy short-term output prediction curve and the flexible resource characteristic matrix, solving of an optimal dispatching strategy by using an improved genetic algorithm to generate a day-ahead regulation plan; S5, rolling correction according to the ultra-short-term fluctuation characteristic spectrum and real-time instructions, dynamic adjustment of the day-ahead regulation plan by a fuzzy adaptive controller to output a minute-level regulation instruction sequence; S6, establishment of a risk assessment model, calculation of a dynamic ramping margin and a risk index based on the load elasticity response interval and the minute-level regulation instruction sequence; S7, when the risk index exceeds a threshold value, a blockchain mechanism is activated to call cross-regional backup resources to generate a collaborative consumption compensation scheme; S8, execution of the regulation instruction sequence and the collaborative consumption compensation scheme by an intelligent contract and verification of the regulation effect; S9, based on the verified regulation effect, a digital twin evaluation system is constructed to generate a consumption efficiency report and optimization suggestions. 2.The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The new energy dynamic consumption data set in S1 includes new energy generation potential data, multi-element user load data and market and power grid constraint data, and the construction process comprises the following steps: S11, new energy generation potential data of photovoltaic power stations and wind farms are collected through meteorological satellite remote sensing data and field station power prediction systems; S12, power consumption characteristic data of industrial, commercial and residential loads are collected through intelligent electric meter measurement systems and demand response management platforms; S13, day-ahead and real-time electricity price signals and cross-section transmission constraint data are obtained through power market transaction systems and power grid energy management systems. 3.The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The deep spatio-temporal convolution neural network in S2 adopts a three-layer heterogeneous structure: The first layer of the spatio-temporal convolution module extracts new energy output geographical distribution characteristics; The second layer of the gating recurrent unit captures output fluctuation time sequence correlation; The third layer of the self-attention mechanism fuses meteorological mutation factors and load transfer characteristics. 4.The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The improved genetic algorithm in S4 contains a triple optimization mechanism: The first optimization target minimizes the new energy curtailment rate and load reduction; The second optimization target balances the calling cost of each flexible resource; The third optimization target meets the power grid safety constraint equation.

5. The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The fuzzy adaptive controller in S5 adopts a double closed-loop control structure: The outer ring corrects the power distribution proportion based on the ultra-short-term prediction error; The inner ring dynamically adjusts a control parameter setting interval according to the frequency deviation signal. 6.The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The calculation method of the dynamic ramping margin index in S6 is: In a 15-minute time window, the difference between the actual output fluctuation rate of new energy and the matching degree of the maximum ramping capacity of the virtual power plant is calculated, and the transmission margin of the power grid section is superimposed to form a three-dimensional evaluation vector, and the mathematical expression is: ； wherein represents the dynamic ramping margin index at time , represents the maximum upward ramping power that the virtual power plant can provide at time , represents the maximum downward ramping power that the virtual power plant can provide at time , represents the change rate of the actual output of new energy within a short time window adjacent to time , represents the transmission power margin of the key section of the power grid at time .

7. The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The blockchain mechanism in S7 adopts a DPoS optimization algorithm: Select the regional energy hub node as the accountant to verify the credibility of the standby resource; Through the cross-chain communication protocol, the smart contract triggering of multi-virtual power plant collaborative consumption is realized. 8.The virtual power plant regulation method for new energy consumption according to claim 1, characterized in that: The consumption efficiency report and optimization suggestion in S9 are generated by a multi-dimensional weighted evaluation algorithm, including five core indicators: New energy consumption completion rate, power grid regulation support contribution, carbon emission intensity reduction, resource calling cost benefit ratio, and user electricity satisfaction. 9.A virtual power plant regulation system for new energy consumption, configured to implement the method of any one of claims 1-8. The system comprises: A multi-source heterogeneous data acquisition module acquires new energy generation potential data through a meteorological satellite receiving unit, collects multi-user load data using a wideband measurement unit, and receives market and grid constraint data through a market interface protocol converter, and outputs new energy dynamic consumption data set; An intelligent prediction analysis module receives the new energy dynamic consumption data set, generates a new energy short-term output prediction curve through a deep spatio-temporal convolutional neural network calculation engine, and outputs an ultra-short-term fluctuation feature spectrum and load elasticity response interval using a fluctuation feature extraction algorithm library; A resource aggregation modeling module receives the multi-user load data and new energy generation potential data, generates adjustable virtual unit parameters through a virtual unit equivalent converter, and constructs a flexible resource feature matrix using a flexible resource digital twin; An optimization decision module receives the new energy short-term output prediction curve and the flexible resource feature matrix, performs optimization matching processing through a multi-objective solver, and generates a day-ahead regulation plan and a minute-level regulation instruction sequence using a rolling correction controller; A blockchain collaboration module receives the regulation instruction sequence, performs instruction verification through a cross-regional consensus verification node, and activates cross-regional standby resource calling using a smart contract execution engine, and outputs a collaborative consumption compensation scheme; A digital twin evaluation module receives the execution effect data of the minute-level regulation instruction sequence and the collaborative consumption compensation scheme data, displays the regulation process through a three-dimensional visualization platform, and generates a consumption efficiency evaluation report and a strategy optimization suggestion using a strategy self-optimization feedback unit; The strategy optimization suggestion is fed back to the rolling correction controller in the optimization decision module to adaptively adjust the generation logic of subsequent regulation instructions.

Images