Virtual power plant scheduling method based on multi-objective optimization
Through the combination of the improved whale optimization algorithm and the sequence minimum quadratic planning algorithm, a multi-objective optimization model is built, which solves the problem of slow convergence and easy to fall into local optimality in virtual power plant scheduling, and realizes efficient absorption of new energy and the stability of the power grid, and improves the operating reliability and robustness of the power grid.
Patent Information
- Application Number
- CN202511045696.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-29
- Publication Date
- 2025-08-29
AI Technical Summary
The existing virtual power plant scheduling methods have problems such as slow convergence or easy to fall into local optimality during the multi-objective optimization process, and it is difficult to adapt to the fluctuations in new energy output and load changes in real time, affecting the reliability and robustness of power grid operation.
The improved whale optimization algorithm combined with the sequence minimum quadratic planning algorithm is used to build a multi-objective optimization model through a combination of global search and local optimization, including the objective function with maximum new energy consumption rate, the highest voltage stability and the lowest operating cost, and overcome the limitations of a single algorithm through real-time data acquisition and dynamic optimization.
It improves the solution efficiency of virtual power plant scheduling, can adapt to the output fluctuations and load changes of new energy, improves the operating reliability and robustness of the power grid, and achieves efficient absorption of new energy and the stability of the power grid.
Smart Images

Figure CN120566428A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of virtual power plant scheduling, and in particular relates to a virtual power plant scheduling method based on multi-objective optimization. Background Art
[0002] As a new type of power dispatching method, virtual power plants (VPPs) have become an important tool for distributed energy management. By integrating distributed energy resources (DERs), such as wind power, photovoltaics, energy storage systems, and adjustable loads, VPPs achieve coordinated and optimized dispatching, thereby increasing the utilization of renewable energy and reducing grid operating costs.
[0003] When scheduling virtual power plants, multiple objectives are more effective than a single one. However, the choice of objectives also impacts scheduling. Current VPP scheduling models still face numerous challenges in terms of renewable energy generation forecast accuracy, grid voltage stability, and operational economics. Therefore, objective selection should prioritize these considerations.
[0004] Once the number and types of targets are determined, existing technologies primarily focus on VPP scheduling based on optimization algorithms. However, these methods are prone to falling into local optima when solving complex nonlinear problems. Some existing methods employ VPP scheduling strategies based on reinforcement learning. While these strategies offer some adaptability, their training convergence is slow, making them difficult to implement in real time. Summary of the Invention
[0005] In view of the shortcomings of existing virtual power plant scheduling methods, such as slow convergence or easy falling into local optimality when using a single optimization algorithm in the process of multi-objective optimization, the present invention provides a virtual power plant scheduling method based on multi-objective optimization, which selects the number and types of objectives, and then overcomes the defects of the single algorithm by combining the global search algorithm with the local search algorithm.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: a virtual power plant scheduling method based on multi-objective optimization, the virtual power plant scheduling method based on multi-objective optimization comprising: Step S1: Collect relevant data required for virtual power plant scheduling; Step S2: construct a multi-objective optimization model and set relevant constraints; Step S3: using the improved whale optimization algorithm combined with the sequential minimum quadratic programming algorithm to optimize and solve the multi-objective optimization model; Step S4: Generate device control instructions based on the optimization results and perform scheduling.
[0007] As an improvement, in step S1, the relevant data required for virtual power plant scheduling includes real-time operating data, meteorological data and load data of each power generation equipment in the virtual power plant.
[0008] As an improvement, step S1 also includes data preprocessing of relevant data, and the data preprocessing step includes data cleaning and predictive modeling.
[0009] As an improvement, data cleaning includes: using IQR or Z-score to eliminate outliers; using linear interpolation or machine learning methods to supplement missing data; using moving average or Kalman filter to smooth data; extracting features and standardizing meteorological data; Predictive modeling includes: establishing a new energy power generation prediction model based on meteorological data and historical data, and outputting the predicted load for the next 24 hours.
[0010] As an improvement, in step S2, the objectives in the multi-objective optimization model include maximizing the new energy absorption rate, achieving the highest voltage stability, and minimizing the operating cost.
[0011] As an improvement, the objective function of maximizing the new energy consumption rate is expressed as:
[0012] in, No. i New energy equipment in time t of power generation, It is I Power generation equipment at time t The total power generation, T is the scheduling period, n is the total number of new energy equipment, N is the total number of power generation equipment; The objective function for the highest voltage stability is expressed as:
[0013] in, is a node j In time t The voltage, is the reference voltage, M is the total number of nodes; The objective function with the lowest operating cost is expressed as: , Among them, T is the scheduling period, is the cost of electricity generation, is the transmission loss cost, It is the equipment maintenance cost.
[0014] As an improvement, in step S2, the relevant constraints include power balance constraints, equipment capacity constraints, and grid security constraints.
[0015] As an improvement, in step S2, The power balance constraint is expressed as:
[0016] in, It is I Power generation equipment at time t of power generation, It is j Nodes at time t The demand, M is the total number of nodes, It's time t transmission loss; The equipment capacity constraint is expressed as:
[0017] in, and They are I The minimum and maximum power generation of each power generation equipment; The grid security constraint is expressed as: ,
[0018] in, and Node j The minimum and maximum voltages, and The lines k The minimum and maximum current.
[0019] As an improvement, in step S3, the improved whale optimization algorithm includes: Initialize the population: Use chaos mapping to initialize the population. The mapping formula is:
[0020] Generate the initial chaotic sequence according to the mapping formula , and iteratively generate chaotic sequences ; Mapping chaotic sequences to the solution space of practical problems[ , ], then the mapping is: ; Fitness evaluation: Calculate the fitness of each individual and evaluate the quality of each solution based on the objective function; Surround the Prey Update: ,in, , , , is the convergence factor, r is a random number in the interval [0, 1], ; Spiral Update: ,in, is a random number between [-1, 1], is a constant; Global Search: ,in, is the position of a randomly selected individual; Stop condition setting: Set the maximum number of iterations or the condition for fitness convergence. If the stop condition is met, exit the algorithm; otherwise, continue to execute the update step.
[0021] As an improvement, in step S3, the improved whale optimization algorithm also includes convergence factor adjustment: using a nonlinear convergence factor, expressed as:
[0022] in, is the maximum number of iterations.
[0023] As an improvement, in step S3, the sequential least quadratic programming algorithm includes: Input: The global optimal solution obtained from the improved whale optimization algorithm As the initial solution; Local optimization: Perform local refined search under the following constraints: power balance constraints, equipment capacity constraints, and grid security constraints. Continuous verification and adjustment are made during the optimization process to ensure that all constraints are met. Output: local optimal solution after optimization , according to the local optimal solution Perform scheduling.
[0024] The beneficial effects of the multi-objective optimization-based virtual power plant scheduling method presented in this paper are: by combining the global search capabilities of WOA and the local convergence efficiency of SLSQP, it overcomes the limitations of a single optimization algorithm. This ensures high-quality solutions (close to the global optimum or a high-quality Pareto solution) while improving solution efficiency and making it more suitable for actual VPP scheduling requirements. Furthermore, through real-time data collection and dynamic optimization, it can adapt to fluctuations in renewable energy output and load changes, improving the reliability and robustness of VPP operations. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 It is a flow chart of a virtual power plant scheduling method according to an embodiment of the present invention.
[0026] Figure 2 4 is an architectural diagram of a virtual power plant dispatching system according to an embodiment of the present invention.
[0027] Figure 3 It is a flow chart of an optimization algorithm of a virtual power plant scheduling method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0028] The technical solutions of the embodiments of the present invention are explained and described below, but the following embodiments are only preferred embodiments of the present invention and are not exhaustive. Based on the embodiments in the embodiments, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention.
[0029] See also Figure 1 The virtual power plant scheduling method based on multi-objective optimization in an embodiment of the present invention includes: Step S1: Collect relevant data required for virtual power plant scheduling; Step S2: construct a multi-objective optimization model and set relevant constraints; Step S3: using the improved whale optimization algorithm combined with the sequential minimum quadratic programming algorithm to optimize and solve the multi-objective optimization model; Step S4: Generate device control instructions based on the optimization results and perform scheduling.
[0030] See also Figure 2 In this embodiment, the virtual power plant dispatching system includes the following four levels: Data collection layer: Utilizes sensor networks and communication equipment to monitor in real time the voltage, power flow, load demand, and other data of physical equipment such as wind power, solar power, energy storage systems, and conventional generators; Optimization calculation layer: Based on the multi-objective optimization model, the optimization algorithm is used to calculate the optimal scheduling strategy according to the collected data, involving the definition of objective functions, setting of constraints, etc. Control execution layer: adjusts the operating status of each energy device based on the optimization calculation results to ensure grid stability, minimize costs, and maximize the absorption of new energy; Result feedback layer: monitors system execution, dynamically adjusts through real-time feedback, and continuously optimizes scheduling results.
[0031] In this embodiment, in step S1, in order to achieve accurate scheduling, the system monitors all equipment in the virtual power plant in real time, and the collected data includes: New energy power generation: real-time power generation of wind and solar power generation equipment; Load demand: Obtain the load demand of different nodes based on load forecasting models, historical data and real-time measurements; Grid voltage and power flow: The voltage, current, and power flow of each node are collected through the SCADA system (Supervisory Control And Data Acquisition system) and real-time sensors; Equipment status: working status, performance and maintenance information of each power generation equipment.
[0032] In this embodiment, in step S1, before performing the optimization calculation, the collected data needs to be preprocessed, which mainly includes: Data cleaning: Remove erroneous data, supplement missing data, and perform data smoothing to ensure data quality. Specifically, use statistical methods (such as the Interquartile Range (IQR) and Z-score) to detect outliers, use linear interpolation or machine learning techniques to estimate missing data, and use moving averages or Kalman filters to smooth data. The IQR stands for interquartile range. The z-score is also called the standard score.
[0033] Load forecasting: Leveraging historical load data and weather forecast information, time series analysis or machine learning models such as LSTM (Long Short-Term Memory) are used to perform short-term load forecasting, providing basic data for dispatch. The model consists of an LSTM layer (50 neurons) and a fully connected output layer (outputting a 24-hour load forecast). The LSTM uses the past 24 hours' load and weather data as input to predict the load for the next 24 hours.
[0034] Renewable energy power generation forecasting: This approach combines weather forecasts and historical power generation data, using machine learning methods to predict renewable energy power generation capacity, taking into account the impact of weather changes. Load data and weather data are combined to form the input data for the LSTM model. Each set of input data contains load and weather data for the past 24 hours, while the target data is the load value for the next 24 hours. This approach provides the LSTM model with time series input, helping the model capture time series characteristics and perform load forecasting.
[0035] See also Figure 3 In this embodiment, in step S2, the multi-objective optimization process includes establishing an objective function and constraints. The objective function includes: (1) Maximize the new energy consumption rate. By rationally dispatching new energy sources such as photovoltaic (PV) and wind power, we can make full use of renewable energy, maximize the proportion of new energy generation in total power generation, and improve the new energy consumption rate. The maximum new energy consumption rate is expressed as follows:
[0036] in, No. i New energy equipment in time t of power generation, It is I Power generation equipment at time t The total power generation, T is the scheduling period, n is the total number of new energy equipment, N is the total number of power generation equipment. Power generation equipment includes new energy equipment and traditional power generation equipment such as thermal power.
[0037] (2) Minimize voltage deviation to ensure that the voltage in the grid is as close to the reference voltage as possible to maintain the stability of the power system. Minimizing voltage deviation is expressed as follows:
[0038] in, is a node j In time t The voltage, is the reference voltage, M is the total number of nodes.
[0039] (3) Minimize operating costs. The operating costs of the power grid include power generation costs, transmission loss costs, and equipment maintenance costs. Minimizing operating costs is expressed as follows:
[0040] in, T is the scheduling period, is the cost of electricity generation, is the transmission loss cost, It is the equipment maintenance cost.
[0041] A comprehensive optimization algorithm is used to achieve these goals, balancing renewable energy consumption, operating costs, and grid stability. The multi-objective model in this embodiment directly addresses the conflicting requirements of VPP scheduling: maximizing renewable energy utilization, maintaining grid voltage stability, and reducing operating costs. Compared to existing technologies that focus solely on costs or tracking plans, this provides a more comprehensive and balanced scheduling solution.
[0042] In this embodiment, the constraints include: (1) power balance constraint, the power grid must meet the power balance at every moment. The power balance constraint is expressed as:
[0043] in, It is I Power generation equipment at time t of power generation, It is j Nodes at time t The demand, M is the total number of nodes, It's time t transmission loss.
[0044] (2) Equipment capacity constraint: The power generation of power generation equipment is limited by its capacity. The equipment capacity constraint is expressed as:
[0045] in, and They are I The minimum and maximum power generation of each power generation equipment.
[0046] (3) Grid safety constraints, which ensure that the voltage and current of the grid do not exceed their safety limits. The grid safety constraints are expressed as:
[0047]
[0048] in, and Node j The minimum and maximum voltages, and The lines k The minimum and maximum current.
[0049] The Whale Optimization Algorithm (WOA) is a heuristic optimization algorithm inspired by whale hunting behavior. It possesses strong global search capabilities, effectively exploring the search space and finding optimal solutions. The algorithm's update mechanisms include prey swarming and spiral updates. The existing standard WOA's random initialization can lead to uneven population distribution in the search space, making it prone to falling into local optima. It is also slow when faced with complex constraints and nonlinear problems such as VPP scheduling.
[0050] In this embodiment, in step S3, the existing whale optimization algorithm is improved to adapt to the uncertainty of the data and avoid falling into the local optimal solution. The improved whale optimization algorithm includes: (1) Population initialization: Use Tent chaos mapping to generate population positions. The mapping formula is:
[0051] Generate the initial chaotic sequence according to the Tent mapping formula (random number, range [0, 1]), and iteratively generate chaotic sequences ,N is the population size; Mapping chaotic sequences to the solution space of practical problems[ , ], then the mapping is: .
[0052] (2) Parameter initialization: set the initial maximum number of iterations , population size, convergence factor and control parameter B.
[0053] Convergence Factor The initial value is set to 2; The control parameter B is randomly generated using the formula: ,in, r A random number in the interval [0, 1].
[0054] (3) Iteration, including fitness evaluation and position update.
[0055] Fitness evaluation: The objective function is calculated using the weighted sum method: , in, is the weight coefficient, satisfying , P is the new energy consumption rate (i.e. ), V is the voltage deviation, and C is the operating cost.
[0056] Position update: including surround prey update and spiral update: Surround the Prey Update: Calculate distance: ; Update location; , in: , is the convergence factor, r is a random number in the interval [0, 1], ; Spiral Update: , in, is a random number between [-1, 1], is a constant.
[0057] Global Search: , in, is the position of a randomly selected individual.
[0058] (4) Convergence factor adjustment: Use a nonlinear convergence factor, expressed as:
[0059] is the maximum number of iterations.
[0060] (5) Boundary processing: Check whether the individual position exceeds the feasible space [a, b] of the decision variable. If so, pull the individual back to the boundary: , ensuring that each individual does not exceed the feasible space of decision variables during the optimization process.
[0061] In this embodiment, in step S3, after the improved whale optimization algorithm, the local optimization phase is entered. The local optimization phase adopts the SLSQP algorithm, which includes: Input: The global optimal solution obtained by the improved whale optimization algorithm As the initial solution of SLSQP; Constraints: SLSQP is used to optimize under the following constraints: power balance constraints, equipment capacity constraints, and grid security constraints (voltage and current limits). SLSQP optimization: Use the minimize function in SciPy (an open-source mathematical, scientific, and engineering computing package) with method = 'SLSQP'. The objective function is fitness calculated using the weighted sum method. Constraints include power balance, equipment capacity, and grid security constraints to ensure the optimal solution is obtained within these constraints. Termination condition: When the maximum number of iterations is reached Or when the fitness change is less than the set threshold, the algorithm terminates.
[0062] In this embodiment, the local optimal solution is obtained after SLSQP optimization. , according to the local optimal solution Perform scheduling.
[0063] In this embodiment, through real-time data collection, preprocessing and dynamic optimization, it is possible to adapt to fluctuations in renewable energy output and load changes, thereby improving the reliability and robustness of VPP operation.
[0064] For multi-objective optimization, when only the existing (standard) WOA is used, it is prone to falling into local optimal solutions due to the problems of imbalanced exploration and utilization and premature convergence. Especially when facing complex constraints and nonlinear problems such as VPP scheduling, the solutions it finds may not be very accurate and the speed of convergence to a high-quality solution may be slow, making it difficult to obtain the optimal balance point that fully satisfies all constraints.
[0065] When using only SLSQP, its performance is highly dependent on the choice of initial point, as it is a gradient-based algorithm. Without good initial guidance, it is easily trapped in the local optimum closest to the initial point, making it impossible to find the global optimal solution or a solution close to the global optimal solution. This is especially true when the VPP scheduling problem has multiple local extreme points, making it almost impossible to independently find the global optimal scheduling strategy.
[0066] When using the optimized WOA plus SLSQP in this embodiment, this hybrid strategy first leverages the improved WOA's strong global search capabilities to overcome SLSQP's vulnerability to local optimality. This strategy effectively locates potential regions within the entire solution space containing high-quality solutions (close to the global optimum) and provides a high-quality initial solution. Subsequently, leveraging SLSQP's powerful local search capabilities and high efficiency, starting from the excellent initial point found by WOA, it conducts rapid and accurate local optimization, effectively addressing various constraints. This combination leverages the advantages of both global exploration and local development, significantly increasing the probability of finding a high-quality global optimal solution (or one very close to the global optimal solution) while ensuring high solution accuracy and convergence efficiency. This provides VPPs with a scheduling solution that strikes a better balance between renewable energy consumption, voltage stability, and operating costs.
[0067] While WOA and SLSQP are both well-known optimization algorithms, this embodiment specifically combines them and applies them to solve multi-objective optimization problems in the VPP domain (considering absorption, voltage, and cost simultaneously). This unprecedented application leverages WOA's global search capabilities to identify potential optimal regions for multi-objective optimization problems, along with SLSQP's local convergence efficiency. This combination is not a simple algorithmic overlay; rather, it employs a collaborative strategy designed to address the unique characteristics of multi-objective optimization problems (such as the presence of multiple Pareto optimal solutions and potential conflicts between objectives) and the complexity of VPP scheduling (nonlinearity and multiple constraints). Furthermore, improvements to the existing WOA, such as chaotic initialization and adjustment of nonlinear convergence factors, further enhance the performance of the WOA algorithm.
[0068] The virtual power plant scheduling method based on multi-objective optimization in the embodiment of the present invention constructs a multi-objective optimization model including maximizing the new energy consumption rate, optimizing voltage stability (minimizing voltage deviation), and minimizing operating costs, thereby improving the new energy consumption, enhancing voltage stability, and reducing operating costs; adopts an improved Whale Optimization Algorithm (WOA) (including chaos map initialization and nonlinear convergence factor) and Sequential Least Squares Programming (SLSP) to optimize the power plant scheduling method based on multi-objective optimization. The solution is solved by a hybrid optimization algorithm that combines WOA with SLSQP (SLSQP), where WOA is responsible for global search and SLSQP is responsible for local fine search. The global search capability of WOA is used to find the potential optimal area of the multi-objective optimization problem, overcoming the defect that gradient methods such as SLSQP are prone to fall into local optimality. Then, SLSQP is used for local fine search to improve the accuracy and convergence efficiency of the solution, taking into account both global search and local convergence speed, and adapting to dynamic adjustment. Although WOA and SLSQP are known optimization algorithms themselves, combining them specifically and applying them to solve this novel multi-objective optimization problem in the VPP field (taking into account absorption, voltage, and cost at the same time) is not something that those skilled in the art can easily think of. This combination is not a simple algorithm superposition, but a targeted A collaborative strategy is designed to address the characteristics of multi-objective optimization problems (such as multiple Pareto optimal solutions and potential conflicts between objectives) and the complexity of VPP scheduling (nonlinearity and multiple constraints). Furthermore, it not only combines existing algorithms but also improves the WOA (such as chaotic initialization and possible nonlinear convergence factor adjustment) to further enhance algorithm performance. The improved WOA improves global search efficiency and the potential for finding better solutions (estimated improvement of 5%-15%) through chaotic initialization and nonlinear convergence factors. The hybrid strategy of the improved WOA and SLSQP combines the advantages of global search and local fine-tuning, achieving solutions with higher accuracy (estimated improvement of 2%-5%) and more reliability (avoiding local optimality, with performance improvements of up to 20%+) than single algorithms while maintaining global performance.
[0069] The method of the embodiment of the invention is described below based on application examples.
[0070] Application example 1: Multi-objective optimization and coordinated scheduling of a high-proportion new energy virtual power plant demonstration project.
[0071] 1. Project Background and Challenges Location and scale: A certain bay economic and technological development zone, covering 15.7 square kilometers, with a total virtual power plant capacity target of ≥100MW.
[0072] Resource composition: Rooftop photovoltaic capacity has been built: 48.11MW (self-generated and used, surplus power is connected to the grid).
[0073] Thermal power plant: 2 sets of 15MW steam turbine generators.
[0074] Demand-side load: Mainly industrial loads such as pharmaceutical chemicals, automobile manufacturing, etc., with the potential for interruption and adjustment.
[0075] Possible future connections: energy storage systems, more distributed photovoltaics, charging piles, etc.
[0076] Core goals: To achieve local consumption of a high proportion of new energy (mainly photovoltaics), improve the flexibility and stability of the regional power grid, and build a virtual power plant operation model with cloud-edge-end intelligent collaboration.
[0077] challenge: The impact of intermittency and volatility of photovoltaic power generation on grid stability.
[0078] The coordinated optimization and scheduling of multiple heterogeneous resources (photovoltaic, thermal power, adjustable loads, energy storage, etc.) is complex.
[0079] How to maximize the utilization of new energy and reduce operating costs while meeting grid security constraints.
[0080] 2. Application of the multi-objective collaborative scheduling solution based on the present invention.
[0081] Step 1: Data collection and intelligent preprocessing Data collection: Source: Smart energy units (edge computing terminals) deployed in photovoltaic power plants and thermal power plants collect real-time data such as power generation power, voltage, current, and equipment status. PV data includes inverter output, irradiance, and temperature.
[0082] Load side: For key industrial users such as pharmaceutical chemicals, automobile manufacturing, etc., real-time power load, voltage, interruptible / adjustable load declared capacity and response characteristics data are collected through smart meters and load control terminals.
[0083] Storage end (if any): SOC, charge and discharge power, available capacity, etc. of the energy storage system.
[0084] Network side: Obtain grid operation data such as key node voltages and line flows from the regional SCADA (Supervisory Data Acquisition and Control System) / DMS (Distribution Management System).
[0085] External: weather forecast data (temperature, light intensity, etc.), electricity market price signals (if participating in the market).
[0086] Data transmission and preprocessing: The smart energy unit performs preliminary data cleaning and format conversion on the edge side, and uploads the data to the virtual power plant cloud platform via 4G / 5G or wired networks (such as MQTT, IEC104, and Modbus protocols).
[0087] The cloud platform further cleans the aggregated data (using IQR / Z-score to remove outliers), fills in missing data (using linear interpolation / machine learning), and smoothes the data (using moving average / Kalman filtering).
[0088] Intelligent Forecasting: The cloud platform uses machine learning algorithms such as LSTM (Long Short-Term Memory Network), combined with historical data and weather forecasts, to accurately predict photovoltaic output and various load types (including the baseline of adjustable loads) for the next 24 hours.
[0089] Step 2: Construct a multi-objective optimization model.
[0090] A mathematical model with the following optimization objectives and constraints is constructed on the virtual power plant cloud platform: Optimization goal: Maximize the new energy absorption rate, give priority to absorbing local photovoltaic power generation, and reduce abandoned light.
[0091] Optimize grid voltage stability, minimize voltage deviations at key nodes, and maintain voltage within a reasonable range.
[0092] Minimize the operating costs of virtual power plants.
[0093] Comprehensively consider the power generation cost of the thermal power plant, demand response compensation cost, energy storage operation cost, equipment maintenance cost and possible market transaction income.
[0094] Constraints: Power balance constraint: regional power generation (photovoltaic, thermal power) + demand response (load reduction / valley filling) + energy storage charging and discharging = total load + network losses + power exchanged with the main grid.
[0095] Equipment capacity constraints: output / adjustment upper and lower limits of each power generation equipment (photovoltaic inverter, thermal power unit), energy storage, and adjustable load.
[0096] Ramp rate constraint: limits the rate of power change of thermal power units and energy storage.
[0097] Grid safety constraints: voltage at key nodes and line currents do not exceed safety limits.
[0098] Demand response constraints: the number of responses, duration, minimum response interval, etc. that users can accept.
[0099] Energy storage constraints: SOC upper and lower limits, charge and discharge times limit, etc.
[0100] Step 3: Use the improved WOA combined with the SLSQP algorithm to optimize and solve the multi-objective optimization model.
[0101] Global Search (Improved WOA): The whale population is initialized using chaotic mapping (such as Tent mapping), and each whale individual represents a set of scheduling plans (output of various resources, load adjustment amount, etc.).
[0102] Through WOA's prey encirclement, spiral update, and global search mechanisms, while satisfying all constraints, it explores the multidimensional solution space to find the optimal or near-optimal solution that comprehensively balances the three objectives. The algorithm incorporates nonlinear convergence factors to improve optimization efficiency and avoid premature convergence.
[0103] Locally refined search (SLSQP): The high-quality solution output by WOA is used as the initial solution of the SLSQP algorithm.
[0104] SLSQP uses its gradient descent properties to perform fast local optimization within the constraint boundaries, further improving the accuracy of the solution and ensuring that all equality and inequality constraints are strictly satisfied.
[0105] Iteration and convergence: Set the maximum number of iterations or fitness convergence conditions, and output the optimal scheduling plan when the conditions are met.
[0106] Step 4: Scheduling instruction generation and execution.
[0107] Instruction generation: The cloud platform generates specific control instructions for each controllable resource based on the optimal scheduling solution obtained through optimization. For example: Photovoltaic power station: active / reactive power regulation instructions.
[0108] Thermal power plants: power generation plans.
[0109] Adjustable industrial load: reduce / increase load target value, respond to start and stop time.
[0110] Energy storage system: charging and discharging power and SOC management targets.
[0111] Instruction issuance and execution: Control instructions are sent to the smart energy units (edge computing terminals) at each resource site through the network.
[0112] The intelligent energy unit parses the instructions and converts them into control signals for the corresponding equipment, achieving precise control of photovoltaic inverters, thermal power plant DCS, industrial load controllers, energy storage PCS, etc.
[0113] Real-time monitoring and dynamic adjustment: The intelligent energy unit monitors the operating status of each resource in real time and feeds the data back to the cloud platform.
[0114] If the actual operating conditions (such as photovoltaic output and load demand) deviate significantly from the forecast, or a power grid emergency occurs, the cloud platform will trigger a re-optimization process and dynamically adjust the scheduling strategy.
[0115] 3. Application effect.
[0116] Improve the level of new energy consumption: Through multi-objective optimization, prioritize photovoltaic power generation, and coordinate thermal power, energy storage and adjustable loads, effectively solve the intermittent problem of 48.11MW of photovoltaic power in the demonstration area, maximize local consumption, reduce abandoned light, and significantly improve the utilization rate of renewable energy.
[0117] Enhance grid voltage stability: Taking voltage stability as the direct optimization target, through coordinated voltage regulation of the reactive power output of photovoltaic inverters, thermal power plants and energy storage, and load regulation when necessary, effectively smooth out voltage fluctuations caused by photovoltaic grid connection and ensure the voltage quality of the regional grid.
[0118] Reduce the overall operating costs of virtual power plants: While ensuring absorption and stability, the overall operating costs of virtual power plants can be minimized by optimizing the output of thermal power plants, energy storage charging and discharging strategies (off-peak charging, peak discharge or participation in frequency regulation), and economically and efficiently calling on demand response resources.
[0119] Achieving efficient collaboration of heterogeneous resources: The hybrid optimization algorithm can effectively handle the coordination of resources with multiple energy forms and different response characteristics, realize the intelligent interaction of "source, grid, load and storage", and give full play to the potential of various resources in this demonstration project.
[0120] Improve the scientific nature and real-time nature of dispatching decisions: Data-based intelligent prediction and rapid multi-objective optimization solutions make dispatching decisions more accurate and timely, enabling rapid response to grid changes and emergencies, and improving the overall operating efficiency and reliability of virtual power plants.
[0121] Promoting cloud-edge-end intelligent collaboration: The implementation of this solution relies on the centralized optimization decision-making of the cloud platform and the rapid response execution of the edge intelligent unit, which perfectly fits the technical route of "cloud-edge-end intelligent collaboration" of the demonstration project and promotes the implementation of advanced energy management technologies.
[0122] An embodiment of the present invention further provides a virtual power plant scheduling system for multi-objective optimization, comprising: Data collection layer: Utilizes sensor networks and communication equipment to monitor in real time the voltage, power flow, load demand, and other data of physical equipment such as wind power, solar power, energy storage systems, and conventional generators; Optimization calculation layer: Based on the multi-objective optimization model, the optimization algorithm is used to calculate the optimal scheduling strategy according to the collected data, involving the definition of objective functions, setting of constraints, etc. Control execution layer: adjusts the operating status of each energy device based on the optimization calculation results to ensure grid stability, minimize costs, and maximize the absorption of new energy; Result feedback layer: monitors system execution, dynamically adjusts through real-time feedback, and continuously optimizes scheduling results.
[0123] An embodiment of the present invention also provides a computer device, including a processor and a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by the processor, the aforementioned virtual power plant scheduling method based on multi-objective optimization is implemented.
[0124] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed, the aforementioned virtual power plant scheduling method based on multi-objective optimization is implemented.
[0125] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Those skilled in the art will understand that the present invention includes, but is not limited to, the contents described in the above specific embodiments. Any modifications that do not deviate from the functional and structural principles of the present invention are intended to be included within the scope of the claims.
Claims
1. A virtual power plant scheduling method based on multi-objective optimization, characterized by: The virtual power plant scheduling method based on multi-objective optimization includes: Step S1: Collect relevant data required for virtual power plant scheduling; Step S2: construct a multi-objective optimization model and set relevant constraints; Step S3: using the improved whale optimization algorithm combined with the sequential minimum quadratic programming algorithm to optimize and solve the multi-objective optimization model; Step S4: Generate device control instructions based on the optimization results and perform scheduling.
2. The virtual power plant scheduling method based on multi-objective optimization according to claim 1 is characterized in that: In step S1, the relevant data required for virtual power plant scheduling includes real-time operating data, meteorological data and load data of each power generation equipment in the virtual power plant; step S1 also includes data preprocessing of the relevant data, and the data preprocessing step includes data cleaning and predictive modeling.
3. The virtual power plant scheduling method based on multi-objective optimization according to claim 2 is characterized in that: Data cleaning includes: removing outliers; supplementing missing data using linear interpolation or machine learning methods; smoothing data using moving average or Kalman filter; feature extraction and standardization of meteorological data; Predictive modeling includes: establishing a new energy power generation prediction model based on meteorological data and historical data, and outputting the predicted load for the next 24 hours.
4. The virtual power plant scheduling method based on multi-objective optimization according to claim 1 is characterized in that: In step S2, the objectives in the multi-objective optimization model include maximizing the new energy absorption rate, achieving the highest voltage stability, and minimizing the operating cost.
5. The virtual power plant scheduling method based on multi-objective optimization according to claim 4 is characterized in that: The objective function for maximizing the new energy consumption rate is expressed as: in, No. i New energy equipment in time t of power generation, It is I Power generation equipment at time t The total power generation, T is the scheduling period, n is the total number of new energy equipment, N is the total number of power generation equipment; The objective function for the highest voltage stability is expressed as: in, is a node j In time t The voltage, is the reference voltage, M is the total number of nodes; The objective function with the lowest operating cost is expressed as: in, T is the scheduling period, is the cost of electricity generation, is the transmission loss cost, It is the equipment maintenance cost.
6. The virtual power plant scheduling method based on multi-objective optimization according to any one of claims 1 to 5, characterized in that: In step S2, the relevant constraints include power balance constraints, equipment capacity constraints, and grid security constraints.
7. The virtual power plant scheduling method based on multi-objective optimization according to claim 6 is characterized in that: In step S2, The power balance constraint is expressed as: in, It is I Power generation equipment at time t of power generation, It is j Nodes at time t The demand, M is the total number of nodes, It's time t transmission loss; The equipment capacity constraint is expressed as: in, and They are I The minimum and maximum power generation of each power generation equipment; The grid security constraint is expressed as: , in, and Node j The minimum and maximum voltages, and The lines k The minimum and maximum current.
8. The virtual power plant scheduling method based on multi-objective optimization according to claim 7 is characterized in that: In step S3, the improved whale optimization algorithm includes: Initialize the population: Use chaos mapping to initialize the population. The mapping formula is: Generate the initial chaotic sequence according to the mapping formula , and iteratively generate chaotic sequences ; Mapping chaotic sequences to the solution space of practical problems[ , ], then the mapping is: ; Fitness evaluation: Calculate the fitness of each individual and evaluate the quality of each solution based on the objective function; Surround the Prey Update: ,in, , , , is the convergence factor, r is a random number in the interval [0, 1], ; Spiral Update: ,in, is a random number between [-1, 1], is a constant; Global Search: ,in, is the position of a randomly selected individual; Stop condition setting: Set the maximum number of iterations or the condition for fitness convergence. If the stop condition is met, exit the algorithm; otherwise, continue to execute the update step.
9. The virtual power plant scheduling method based on multi-objective optimization according to claim 8, characterized in that: In step S3, the improved whale optimization algorithm also includes convergence factor adjustment: using a nonlinear convergence factor, expressed as: in, is the maximum number of iterations.
10. The virtual power plant scheduling method based on multi-objective optimization according to claim 9, characterized in that: In step S3, the sequential least quadratic programming algorithm includes: Input: The global optimal solution obtained from the improved whale optimization algorithm As the initial solution; Local optimization: Perform local refined search under the following constraints: power balance constraints, equipment capacity constraints, and grid security constraints. Continuous verification and adjustment are made during the optimization process to ensure that all constraints are met. Output: local optimal solution after optimization , according to the local optimal solution Perform scheduling.
Citation Information
Patent Citations
Bionic strategy optimized hybrid whale optimization algorithm
CN107016436A
Micro-grid economic dispatching method based on cross variation whale optimization algorithm
CN110070292A
Structural damage identification method based on adaptive weight whale optimization algorithm
CN111507045A
Optimization problem solving method based on improved whale optimization algorithm
CN116070672A
Method for determining optimal test planning of 3D NoC based on improved whale optimization algorithm
CN116187383A
Cited By
Virtual power plant optimization scheduling system and method
CN121213294A
Distribution network user side resource optimization scheduling method for multi-source data fusion
CN121436590A
A method for optimizing and scheduling user-side resources in distribution networks based on multi-source data fusion
CN121436590B
