Aggregation control optimization method and system suitable for distributed energy

Abstract

The invention discloses an aggregation control optimization method and system suitable for distributed energy, and relates to the technical field of distributed energy aggregation optimization control, and the method comprises the steps: collecting first data, and carrying out the preprocessing of the first data; establishing a mathematical model of the distributed energy according to the preprocessed first data; performing short-term prediction on load demand and renewable energy output by using a hybrid deep learning model; establishing a multi-objective optimization model and solving to obtain a multi-objective optimization result; selecting an optimal control decision from a multi-objective optimization result, and generating a specific control strategy; a hierarchical control mechanism and a self-adaptive control mechanism are introduced, the running state of the system is monitored in real time, and faults are detected and processed; according to the invention, flexible and efficient aggregation control of multiple types of distributed energy is realized; the method can effectively balance multi-dimensional indexes such as economy, reliability and environmental protection, has high adaptability and robustness, and provides important support for large-scale application and intelligent operation of distributed energy.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed energy aggregation optimization control, and in particular to an aggregation control optimization method and system suitable for distributed energy. Background Art

[0002] With the transformation of the global energy structure and the deepening of power system reform, the proportion of distributed energy resources (DERs) in the power system has increased significantly. Distributed energy mainly includes solar photovoltaic, wind power, small hydropower and energy storage systems, which have the advantages of flexible deployment, low carbon emissions and energy self-sufficiency. However, the intermittent, volatile and decentralized characteristics of distributed energy have brought new challenges to the safe and stable operation of the power system.

[0003] Most of the existing aggregation control methods are designed for specific distributed energy types or specific application scenarios, such as the optimization scheduling method of wind power or solar photovoltaic, or the energy management system for a specific area. These methods are usually limited to a single type of energy management or a specific application scenario, lacking sufficient flexibility and adaptability, and are difficult to cope with the diverse distributed energy combinations and the complex and changing power market environment; traditional aggregation control strategies often focus on a single performance indicator, such as economy, reliability or environmental protection, and it is difficult to achieve a comprehensive balance between multi-dimensional performance indicators. In addition, the existing aggregation control methods are often inefficient when dealing with large-scale distributed energy access and cannot meet the real-time and high-precision control requirements. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to achieve flexible and efficient aggregation control of multiple types of distributed energy sources to adapt to diverse energy combinations and complex and changeable electricity market environment.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides an aggregation control optimization method applicable to distributed energy, including:

[0008] Collecting first data and performing preprocessing;

[0009] Establishing a mathematical model of distributed energy according to the preprocessed first data;

[0010] Use hybrid deep learning models to make short-term forecasts of load demand and renewable energy output;

[0011] Based on the mathematical model of distributed energy and the results of short-term forecasting, a multi-objective optimization model is established and solved to obtain the multi-objective optimization results;

[0012] Select the optimal control decision from the multi-objective optimization results and generate a specific control strategy;

[0013] Based on the control strategy, hierarchical control mechanism and adaptive control mechanism are introduced to monitor the system operation status in real time and detect and handle faults.

[0014] As a preferred solution for the aggregation control optimization method applicable to distributed energy, wherein:

[0015] The step of establishing a mathematical model of distributed energy according to the preprocessed first data includes:

[0016] Establish a photovoltaic power generation model based on the physical characteristics of the photovoltaic power generation system;

[0017] Establish a wind power generation model based on the operating characteristics of wind turbines;

[0018] A hydrogen power generation model is established based on the working principle of proton exchange membrane fuel cells.

[0019] As a preferred solution for the aggregation control optimization method applicable to distributed energy, wherein:

[0020] The short-term prediction of load demand and renewable energy output using a hybrid deep learning model includes:

[0021] By combining different deep learning components, the input time series data is processed to capture temporal dependencies and generate prediction outputs;

[0022] Using the rolling horizon optimization method, control inputs are generated according to the prediction results, and the control strategy is adjusted through weight coefficients.

[0023] As a preferred solution for the aggregation control optimization method applicable to distributed energy, wherein:

[0024] The mathematical model based on distributed energy and the results of short-term prediction, establishing a multi-objective optimization model and solving it include:

[0025] Construct a multi-objective optimization model that takes into account economy, reliability and environmental protection:

[0026] Establish an economic objective function based on the total economic cost of the power system;

[0027] Based on the probability of load loss, the reliability index of the power system is measured and the reliability objective function is established;

[0028] Establish an environmental protection objective function based on the total carbon emissions of the power system;

[0029] Establishing constraints includes power balance constraints, output upper and lower limit constraints, and climbing rate constraints.

[0030] As a preferred solution for the aggregation control optimization method applicable to distributed energy, wherein:

[0031] The real-time monitoring system operating status, detecting and handling faults includes:

[0032] Monitor the system operating status in real time. When a fault is detected, perform the following steps:

[0033] Fault location: Use deep learning methods to quickly locate faults;

[0034] Topology reconstruction: Reconstruct the network topology structure based on graph theory algorithms;

[0035] Control reconfiguration: reconfigure control parameters according to the new topology;

[0036] Collaborative decision-making: Multiple agents reach a globally consistent control decision through a consensus algorithm.

[0037] As a preferred solution for the aggregation control optimization method applicable to distributed energy, wherein:

[0038] The hierarchical control mechanism includes:

[0039] Design a three-layer control architecture: energy unit layer, regional aggregation layer, and global coordination layer;

[0040] The controllers at each layer adopt the distributed model predictive control method, and the control law is expressed as:

[0041] u i (k) = argmin∑J i (x i (k+j|k),u i (k+j|k))

[0042] constraint:

[0043] x i (k+j+1|k)=f i (x i (k+j|k),u i (k+j|k))

[0044] g i (x i (k+j|k),u i (k+j|k))≤0

[0045] hi (∑x i (k+j|k),∑u i (k+j|k))≤0

[0046] Among them, J i is the local objective function, x i and u i are local states and control inputs, respectively, and f i is the system dynamic equation, g i and h i They are local constraints and coupling constraints respectively.

[0047] As a preferred solution for the aggregation control optimization method applicable to distributed energy, wherein:

[0048] The adaptive control mechanism includes:

[0049] Adopt DDPG algorithm, including Actor network and Critic network:

[0050] Actor: a = μ(s|θ μ )

[0051] Critic:Q(s,a|θ Q )

[0052] Among them, s is the current state, a is the action taken in a given state, μ is the Actor network; θ μ is the parameter of the Actor network; Q is the Critic network; θ Q are the parameters of the Critic network;

[0053] The update strategy of the Actor network is:

[0054]

[0055] Among them, ▽θ μ J is the objective function J with respect to the Actor network parameter θ μ The gradient of ; E is the expected value operator; Indicates that in state s t Next, when action a is generated by the Actor network μ(s t ) is decided, the gradient of the Critic network about action a; Indicates that in state s t The Actor network has parameters θ μ The gradient of

[0056] The update strategy of the Critic network is:

[0057] L(θ Q)=Ε[(r+γQ′(s′,u′(s′|θ μ′ )|θ Q )-Q(s,a|θ Q )) 2 ]

[0058] Among them, L(θ Q ) is the loss function of the Critic network, which is used to measure the accuracy of the value function prediction; r is the reward, γ is the discount factor, Q′(s′,u′(s′|θ μ′ )|θ Q ) is the next state s′ according to the updated Actor network u′(s′|θ μ′ ) is the value estimate of the action selected; Q(s,a|θ Q ) is the estimated value of taking action a in the current state s.

[0059] In a second aspect, an embodiment of the present invention provides an aggregated control optimization system applicable to distributed energy, including:

[0060] A data collection and preprocessing module, used for collecting first data and performing preprocessing;

[0061] A distributed energy modeling module, used to establish a mathematical model of distributed energy according to the preprocessed first data;

[0062] The short-term forecasting module is used to make short-term forecasts of load demand and renewable energy output using a hybrid deep learning model;

[0063] The multi-objective optimization module is used to establish and solve the multi-objective optimization model based on the mathematical model of distributed energy and the results of short-term prediction to obtain the multi-objective optimization results;

[0064] The decision generation module is used to select the optimal control decision from the multi-objective optimization results and generate a specific control strategy;

[0065] The real-time monitoring and fault handling module is used to introduce hierarchical control mechanism and adaptive control mechanism based on control strategy, monitor the system operation status in real time, detect and handle faults.

[0066] In a third aspect, an embodiment of the present invention provides a computing device, including:

[0067] Memory and processor;

[0068] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the aggregation control optimization method applicable to distributed energy as described in any embodiment of the present invention.

[0069] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the aggregation control optimization method applicable to distributed energy.

[0070] Beneficial effects of the invention: The invention realizes flexible and efficient aggregation control of multiple types of distributed energy through advanced technologies such as multi-model fusion, multi-objective optimization, reinforcement learning, hierarchical distributed architecture, model predictive control and multi-agent system; it can effectively balance multi-dimensional indicators such as economy, reliability and environmental protection, has strong adaptability and robustness, and provides important support for the large-scale application and intelligent operation of distributed energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0072] Figure 1 It is an overall flow chart of the aggregation control optimization method applicable to distributed energy as described in the first embodiment of the present invention. DETAILED DESCRIPTION

[0073] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0074] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0075] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0076] Example 1

[0077] Reference Figure 1 , which is the first embodiment of the present invention, and provides an aggregation control optimization method applicable to distributed energy, including:

[0078] S1: collect first data and perform preprocessing;

[0079] S2: establishing a mathematical model of distributed energy according to the preprocessed first data;

[0080] S3: Short-term forecasting of load demand and renewable energy output using a hybrid deep learning model;

[0081] S4: Based on the mathematical model of distributed energy and the results of short-term forecasting, a multi-objective optimization model is established and solved to obtain the multi-objective optimization results;

[0082] S5: Select the optimal control decision from the multi-objective optimization results and generate a specific control strategy;

[0083] S6: Based on the control strategy, a hierarchical control mechanism and an adaptive control mechanism are introduced to monitor the system operation status in real time and detect and handle faults.

[0084] It should be noted that, through S1-S6, this method can realize intelligent and adaptive aggregation control of distributed energy, effectively improving the economy, reliability and environmental protection of the system. This method has strong universality and scalability, and can be applied to distributed energy systems of various types and sizes. Regularly evaluate the economy, reliability and environmental protection indicators of the system to provide a basis for long-term optimization; according to the performance evaluation results, adjust the optimization objectives and constraints, repeat steps S3-S6, and achieve continuous optimization.

[0085] Example 2

[0086] Reference Figure 1 , which is an embodiment of the present invention, provides an aggregation control optimization method applicable to distributed energy based on the previous embodiment, including:

[0087] In an embodiment of the present application, the first data in the above step S1 includes historical operation data, load data and environmental data of distributed energy; and the preprocessing includes cleaning and normalizing the data.

[0088] In another possible implementation, the first data also includes real-time data: in addition to historical operation data, load data and environmental data, it may also include real-time operation data of distributed energy (such as current power generation, real-time status of equipment, etc.).

[0089] In another possible implementation, the first data also includes forecast data: the forecast data may be a forecast of future energy production and consumption based on historical data and current conditions. This may include weather forecast data, load forecast data, etc., which are important references to the future operating status of the energy system.

[0090] In another possible implementation, the preprocessing further includes:

[0091] Data interpolation: When data is missing or incomplete, data interpolation techniques are used to fill in the missing data points to ensure the continuity and integrity of the data.

[0092] Outlier detection and processing: Detect outliers in the data and process them, such as removing outliers, replacing outliers with neighboring values, or smoothing outliers using statistical methods.

[0093] Feature selection and dimensionality reduction: Feature selection technology can be used to remove redundant and irrelevant features, reduce the dimension of data, and improve the computational efficiency and accuracy of the model.

[0094] Data balancing: For unbalanced data sets, oversampling, undersampling, or synthetic data generation techniques can be used to balance data of different categories to avoid the model being biased towards categories with larger data volumes.

[0095] In the embodiment of the present application, in the above step S2, establishing a mathematical model of distributed energy according to the preprocessed first data includes:

[0096] Photovoltaic power generation model:

[0097] The output power of photovoltaic power generation can be expressed as:

[0098] P pv =η pv *S*G*(1-0.005(T-25))

[0099] Where P pv is the photovoltaic power output power (kW), η pv is the photovoltaic panel efficiency, S is the photovoltaic panel area (m 2 ), G is the solar radiation intensity (kW / m 2 ), T is the ambient temperature (℃).

[0100] Wind power generation model:

[0101] The output power of a wind turbine can be expressed as:

[0102]

[0103] Where P w (v) is the wind power output power (kW), Pr is the rated power (kW), v is the wind speed (m / s), v ci ,v r ,v co They are cut-in wind speed, rated wind speed and cut-out wind speed (m / s) respectively.

[0104] Hydrogen power generation model:

[0105] The output voltage of the proton exchange membrane fuel cell (PEMFC) can be expressed as:

[0106] V fc =E 0 -blog(i)-Ri-mexp(ni)

[0107] Where V fc is the fuel cell output voltage (V), E 0 is the open circuit voltage (V), i is the current density (A / cm 2 ), b is the Tafel slope, a constant related to electrode kinetics. R is the ohmic resistance (Ω·cm 2 ), m and n are mass transfer overvoltage coefficients, constants related to the mass transfer process.

[0108] In the embodiment of the present application, the short-term prediction of load demand and renewable energy output using the hybrid deep learning model in the above step S3 includes:

[0109] Combining the long short-term memory network (LSTM) and the temporal convolutional network (TCN) to build a hybrid prediction model:

[0110] y t =LSTM(x t ,h t-1 ,c t-1 )+TCN(x t-k :x t )

[0111] Among them, y t is the predicted output, x t is the input feature, h t-1 and c t-1 are the hidden state and unit state of LSTM respectively, x t-k :x t represents the input sequence for the past k time steps.

[0112] Based on the prediction results, the model predictive control method of the rolling horizon is adopted:

[0113] minJ=∑w 1 ||y(k+i|k)-r(k+i)|| 2 +w 2||Δu(k+i-1|k)|| 2

[0114] constraint:

[0115] y min ≤y(k+i|k)≤y max

[0116] u min ≤u(k+i-1|k)≤u max

[0117] |Δu(k+i-1|k)|≤Δu max

[0118] Among them, y(k+i|k) is the i-step ahead prediction output, r(k+i) is the reference trajectory, u(k+i-1|k) is the control input, and w 1 and w 2 is the weight coefficient.

[0119] In the embodiment of the present application, in the above step S4, based on the mathematical model of distributed energy and the result of short-term prediction, establishing and solving a multi-objective optimization model includes:

[0120] Construct a multi-objective optimization model that takes into account economy, reliability and environmental protection:

[0121] minF=[f 1 ,f 2 ,f 3 ]

[0122] Among them, f 1 ,f 2 ,f 3 They represent the economic, reliability and environmental protection objective functions respectively.

[0123] Establish the economic objective function:

[0124] f 1 =∑(C op +C main -R grid )

[0125] In the formula, f 1 is the first objective function, which usually represents the total economic cost of the power system; C op The operating cost includes energy consumption cost and labor cost during the operation of the power system; C main is the maintenance cost, including the cost of power system maintenance and repair; R grid The revenue from selling electricity to the power grid refers to the income obtained by the power system from selling excess electricity to the power grid.

[0126] Establish reliability objective function:

[0127] f 2 =LOLP = P(∑P i <L)

[0128] In the formula, f 2 is the second objective function, representing the load loss probability. LOLP is the load loss probability, which is an indicator to measure the reliability of the power system and indicates the probability that the system cannot meet the load demand. i is the output of the ith distributed energy source, and L is the load demand, that is, the total power demand that the power system needs to meet.

[0129] Establish environmental protection objective function:

[0130] f 3 =∑(α i *E i )

[0131] In the formula, f 3 is the third objective function, representing the total carbon emissions of the power system; α i is the carbon emission coefficient of the i-th energy source, indicating the carbon emissions generated per unit of electricity generation. i is the power generation of the i-th energy source.

[0132] Establishing constraints includes:

[0133] Power balance constraints:

[0134] ∑P i =L

[0135] In the formula, ∑P i is the total output of all generators; L is the load demand, that is, the total power demand that the power system needs to meet. This constraint ensures that the total power generation of the power system matches the load demand.

[0136] Output upper and lower limits:

[0137] P i,min ≤P i ≤P i,max

[0138] Where P i is the output of the i-th generator set; P i,min is the minimum output of the ith generator set; P i,max is the maximum output of the ith generator set. This constraint ensures that the output of each generator set is within its safe and technically acceptable range.

[0139] Ramp rate constraint:

[0140] |Pi (t)-P i (t-1)|≤ΔP i,max

[0141] Where P i is the output of the i-th generator set at time t; P i (t-1) is the output of the i-th generator set at time t-1; ΔP i,max is the maximum ramp rate of the ith generator set, i.e., the maximum change in output per unit time. This constraint ensures that the output change of the generator set will not be too drastic to avoid damage to the equipment or affect the stability of the power grid.

[0142] In the embodiment of the present application, the improved NSGA-III algorithm is used to solve the Pareto optimal solution set to obtain a multi-objective optimization result;

[0143] It should be noted that NSGA-III (Nondominated Sorting Genetic Algorithm III) is an evolutionary algorithm for multi-objective optimization problems, especially suitable for dealing with multi-objective optimization problems with a large number of objective functions and complex solution spaces. The improved NSGA-III algorithm is optimized on this basis to improve the solution efficiency and solution quality.

[0144] Specifically, the improved NSGA-III algorithm steps are as follows:

[0145] Initialize the population:

[0146] The initial population is randomly generated, and each individual in the population represents a potential solution (corresponding to the decision variable of the multi-objective optimization problem).

[0147] Evaluate fitness:

[0148] Calculate the objective function value of each individual, including economic, reliability and environmental protection objective functions.

[0149] Non-dominated sorting:

[0150] The population is non-dominated and sorted, and the individuals are divided into different fronts according to their non-dominated relationships in the target space. That is, the individuals on the Pareto front are non-dominated, while the other individuals are dominated.

[0151] Diversity maintenance:

[0152] Based on NSGA-III, the improved algorithm introduces a variety of diversity maintenance mechanisms, such as the Reference Points Method and Crowding Distance, to ensure the diversity and uniformity of solutions.

[0153] Select an action:

[0154] According to the results of non-dominated sorting and diversity maintenance, individuals of the next generation population are selected. The selection mechanism includes tournament selection and reference point-based selection.

[0155] Crossover and mutation:

[0156] Perform crossover and mutation operations on the selected individuals to generate new individuals and enrich the diversity of the population.

[0157] Update population:

[0158] Use the newly generated individuals to update the population, usually using an elite retention strategy to retain the best individuals in history.

[0159] Convergence check:

[0160] Check whether the algorithm meets the convergence conditions (such as the maximum number of iterations or the convergence criteria of the objective function). If not, return to non-dominated sorting for the next round of iteration; otherwise, output the Pareto optimal solution set.

[0161] Output the Pareto optimal solution set:

[0162] The individuals on the final Pareto frontier are output as the Pareto optimal solution set for decision makers to choose the optimal compromise solution.

[0163] In another possible implementation, MOEA / D (Multi-Objective Evolutionary Algorithm based on Decomposition) can also be used to solve the multi-objective optimization model:

[0164] MOEA / D is a decomposition-based multi-objective optimization algorithm that decomposes the multi-objective optimization problem into multiple scalar optimization sub-problems. Each sub-problem corresponds to a weight vector and is solved through collaborative evolution.

[0165] Here are the steps:

[0166] Initialize the weight vector: Decompose the multi-objective optimization problem into multiple scalar optimization sub-problems, each sub-problem corresponds to a weight vector.

[0167] Initialize population: Randomly generate the initial population.

[0168] Subproblem collaboration: The solution of each subproblem is evolved collaboratively with the solutions of neighboring subproblems, using genetic operators (crossover and mutation) to generate new solutions.

[0169] Update population: Update the individuals in the population and retain the historical optimal solution.

[0170] Convergence check: Check whether the algorithm meets the convergence conditions, such as the maximum number of iterations or the convergence criteria of the objective function. If not, return to the subproblem to collaborate on the next round of iteration; otherwise, output the Pareto optimal solution set.

[0171] In the embodiment of the present application, the above step S5 selects the optimal control decision from the multi-objective optimization results and generates a specific control strategy including:

[0172] According to the decision maker's preference, the optimal compromise solution is selected from the Pareto solution set as the control decision.

[0173] Specifically, Pareto solution set analysis:

[0174] Pareto solution set: Multi-objective optimization problems usually have multiple Pareto optimal solutions, which constitute the Pareto solution set. Each Pareto optimal solution has no superiority or inferiority relationship with each other in the objective space, and decision makers need to choose the optimal compromise solution according to specific needs.

[0175] Target space visualization: Visualize the Pareto solution set in the target space to help decision makers intuitively understand the relationship between solutions. Common visualization methods include scatter plots, parallel coordinates plots, etc.

[0176] Weight determination:

[0177] Decision maker preference input: Decision makers input weights based on the system's economic, reliability, and environmental priorities. Weights can be dynamically adjusted based on actual needs and business goals.

[0178] Weighted summation: Use the weighted summation method to integrate multiple objective functions into a single objective function.

[0179] The best compromise solution is selected:

[0180] Single objective optimization: After determining the weights, the multi-objective optimization problem is transformed into a single objective optimization problem. Through the weighted summation method, the solution that makes the single objective function optimal, that is, the optimal compromise solution, is found.

[0181] Constraint check: Ensure that the selected optimal compromise solution satisfies all constraints, such as power balance constraints, output upper and lower limit constraints, and climbing rate constraints.

[0182] Control strategy generation:

[0183] Generation of power generation plan: Based on the selected optimal compromise solution, the power generation plan of each distributed energy source is determined, including the charging and discharging plans of photovoltaic power generation, wind power generation and energy storage systems.

[0184] Load distribution: Based on load forecast results and energy output forecast, reasonably distribute load to ensure the economy and reliability of the power system.

[0185] Real-time adjustment strategy: Generate real-time adjustment strategy to deal with emergencies in system operation, such as sudden load changes, equipment failures, etc. Real-time adjustment strategy includes equipment status monitoring, power adjustment instructions, etc.

[0186] Control instructions are issued:

[0187] Instruction generation: Convert specific control strategies into control instructions, including power generation instructions for each distributed energy source, charging and discharging instructions for the energy storage system, etc.

[0188] Instruction issuance: Control instructions are issued to various controllers and devices through the communication network for execution, ensuring the real-time and accuracy of instructions.

[0189] Feedback and Adjustment:

[0190] Real-time feedback: Obtain the actual operation data of the system and provide real-time feedback. Based on the feedback data, evaluate the execution effect of the control strategy.

[0191] Strategy adjustment: Adjust the control strategy based on real-time feedback data, such as adjusting the power generation plan, load distribution strategy, etc., to ensure the optimal state of system operation.

[0192] In the embodiment of the present application, real-time monitoring of the system operation status in the above step S6, detecting and handling faults includes:

[0193] Monitor the system operation status in real time and immediately start the fault self-healing mechanism once a fault is detected:

[0194] When a fault is detected, the system performs the following steps:

[0195] Fault location: Use deep learning methods to quickly locate faults.

[0196] Topology reconstruction: Reconstruct the network topology based on graph theory algorithms.

[0197] Control reconfiguration: Reconfigure control parameters according to the new topology.

[0198] Collaborative decision-making: Multiple agents reach a globally consistent control decision through a consensus algorithm.

[0199] The consensus algorithm can be expressed as:

[0200]

[0201] Among them, x i is the state of Agent i, ε is the step size, a ij is the adjacency matrix element.

[0202] It should be noted that the present invention constructs a distributed control system based on multi-agent technology, each agent represents a control unit, and the agents achieve fault self-healing through information sharing and collaborative decision-making.

[0203] In the embodiment of the present application, the hierarchical control mechanism in the above step S6 includes:

[0204] Design a three-layer control architecture:

[0205] Energy unit layer: responsible for the local control of a single distributed energy source, such as MPPT control, voltage regulation, etc.

[0206] Regional aggregation layer: responsible for the coordinated control of multiple distributed energy sources in the region, such as power distribution, voltage and reactive power coordination, etc.

[0207] Global coordination layer: responsible for the coordination and control among multiple regions, such as cross-regional power scheduling, system-level auxiliary services, etc.

[0208] The controllers at each layer adopt the distributed model predictive control (DMPC) method, and the control law is expressed as:

[0209] u i (k)=argminΣJ i (x i (k+j|k),u i (k+j|k))

[0210] constraint:

[0211] x i (k+j+1|k)=f i (x i (k+j|k),u i (k+j|k))

[0212] g i (x i (k+j|k),u i (k+j|k))≤0

[0213] h i (Σx i (k+j|k),∑u i (k+j|k))≤0

[0214] Among them, J iis the local objective function, x i and u i are local states and control inputs, respectively, and f i is the system dynamic equation, g i and h i They are local constraints and coupling constraints respectively.

[0215] In the embodiment of the present application, the adaptive control mechanism in the above step S6 includes:

[0216] Adopt DDPG algorithm, including Actor network and Critic network:

[0217] Actor: a = μ(s|θ μ )

[0218] Critic:Q(s,a|θ Q )

[0219] Among them, s is the current state, a is the action taken in a given state, μ is the Actor network, which is a policy function that determines what action to take in a given state; |θ μ θ is the parameter of the Actor network; Q is the Critic network, which is a value function that estimates the long-term benefit of taking an action in a given state. Q are the parameters of the Critic network.

[0220] The update strategy of the Actor network is:

[0221]

[0222] Among them, ▽θ μ J is the objective function J with respect to the Actor network parameter θ μ The gradient of ; E is the expected value operator, which means taking the average over all possible states and actions; In status t Next, when action a is generated by the Actor network μ(s t ) is decided, the gradient of the Critic network about action a; In status t The Actor network has parameters θ μ gradient.

[0223] The update strategy of the Critic network is:

[0224] L(θ Q )=Ε[(r+γQ′(s′,u′(s′|θ μ′ )|θ Q )-Q(s,a|θ Q ))2 ]

[0225] Among them, L(θ Q ) is the loss function of the Critic network, which is used to measure the accuracy of the value function prediction; r is the reward, γ is the discount factor, Q′(s′,u′(s′|θ μ′ )|θ Q ) is the next state s′ according to the updated Actor network u′(s′|θ μ′ ) The value of the action selected is estimated by the updated Critic network θ Q Provide; Q(s,a|θ Q ) is the estimated value of taking action a in the current state s, provided by the current Critic network.

[0226] It should be noted that the present invention adopts an adaptive control method based on the deep deterministic policy gradient (DDPG) algorithm. The DDPG algorithm combines the advantages of the deep Q network (DQN) and the policy gradient method, and can handle problems in continuous action space.

[0227] Example 3

[0228] The above is a schematic scheme of the aggregation control optimization method applicable to distributed energy in this embodiment. It should be noted that the technical scheme of the aggregation control optimization system applicable to distributed energy and the technical scheme of the aggregation control optimization method applicable to distributed energy belong to the same concept. For the details not described in detail in the technical scheme of the aggregation control optimization system applicable to distributed energy in this embodiment, please refer to the description of the technical scheme of the aggregation control optimization method applicable to distributed energy.

[0229] This embodiment also provides a system based on the aggregation control optimization method applicable to distributed energy, including:

[0230] A data collection and preprocessing module, used for collecting first data and performing preprocessing;

[0231] A distributed energy modeling module, used to establish a mathematical model of distributed energy according to the preprocessed first data;

[0232] The short-term forecasting module is used to make short-term forecasts of load demand and renewable energy output using a hybrid deep learning model;

[0233] The multi-objective optimization module is used to establish and solve the multi-objective optimization model based on the mathematical model of distributed energy and the results of short-term prediction to obtain the multi-objective optimization results;

[0234] The decision generation module is used to select the optimal control decision from the multi-objective optimization results and generate a specific control strategy;

[0235] The real-time monitoring and fault handling module is used to introduce hierarchical control mechanism and adaptive control mechanism based on control strategy, monitor the system operation status in real time, detect and handle faults.

[0236] This embodiment further provides a computing device, which is applicable to the case of the aggregation control optimization method applicable to distributed energy, including:

[0237] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the aggregation control optimization method applicable to distributed energy as proposed in the above embodiment.

[0238] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the aggregation control optimization method applicable to distributed energy sources proposed in the above embodiment is implemented.

[0239] The storage medium proposed in this embodiment and the aggregation control optimization method applicable to distributed energy proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0240] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for optimizing the aggregation control of distributed energy, characterized in that: include: Collecting first data and performing preprocessing; Establishing a mathematical model of distributed energy according to the preprocessed first data; Use hybrid deep learning models to make short-term forecasts of load demand and renewable energy output; Based on the mathematical model of distributed energy and the results of short-term forecasting, a multi-objective optimization model is established and solved to obtain the multi-objective optimization results; Select the optimal control decision from the multi-objective optimization results and generate a specific control strategy; Based on the control strategy, hierarchical control mechanism and adaptive control mechanism are introduced to monitor the system operation status in real time and detect and handle faults.

2. The aggregation control optimization method applicable to distributed energy sources according to claim 1, characterized in that: The step of establishing a mathematical model of distributed energy according to the preprocessed first data includes: Establish a photovoltaic power generation model based on the physical characteristics of the photovoltaic power generation system; Establish a wind power generation model based on the operating characteristics of wind turbines; A hydrogen power generation model is established based on the working principle of proton exchange membrane fuel cells.

3. The aggregation control optimization method applicable to distributed energy sources according to claim 2, characterized in that: The short-term prediction of load demand and renewable energy output using a hybrid deep learning model includes: By combining different deep learning components, the input time series data is processed to capture temporal dependencies and generate prediction outputs; Using the rolling horizon optimization method, control inputs are generated according to the prediction results, and the control strategy is adjusted through weight coefficients.

4. The aggregation control optimization method applicable to distributed energy sources according to claim 3, characterized in that: The mathematical model based on distributed energy and the results of short-term prediction, establishing a multi-objective optimization model and solving it include: Construct a multi-objective optimization model that takes into account economy, reliability and environmental protection: Establish an economic objective function based on the total economic cost of the power system; Based on the probability of load loss, the reliability index of the power system is measured and the reliability objective function is established; Establish an environmental protection objective function based on the total carbon emissions of the power system; Establishing constraints includes power balance constraints, output upper and lower limit constraints, and climbing rate constraints.

5. The aggregation control optimization method applicable to distributed energy sources according to claim 4, characterized in that: The real-time monitoring system operating status, detecting and handling faults includes: Monitor the system operating status in real time. When a fault is detected, perform the following steps: Fault location: Use deep learning methods to quickly locate faults; Topology reconstruction: Reconstruct the network topology structure based on graph theory algorithms; Control reconfiguration: reconfigure control parameters according to the new topology; Collaborative decision-making: Multiple agents reach a globally consistent control decision through a consensus algorithm.

6. The aggregation control optimization method applicable to distributed energy sources according to claim 5, characterized in that: The hierarchical control mechanism includes: Design a three-layer control architecture: energy unit layer, regional aggregation layer, and global coordination layer; The controllers at each layer adopt the distributed model predictive control method, and the control law is expressed as: u i (k)=argmin∑J i (x i (k+j|k),u i (k+j|k)) constraint: x i (k+j+1|k)=f i ( x i (k+j|k),u i (k+j|k)) g i (x i (k+j|k),u i (k+j|k))≤0 h i (∑x i (k+j|k),∑u i (k+j|k))≤0 Among them, J i is the local objective function, x i and u i are local states and control inputs, respectively, and f i is the system dynamic equation, g i and h i They are local constraints and coupling constraints respectively.

7. The aggregation control optimization method applicable to distributed energy sources according to claim 6, characterized in that: The adaptive control mechanism includes: Adopt DDPG algorithm, including Actor network and Critic network: Actor: a=μ(s|θ μ ) Critic:Q(s,a|θ Q ) Among them, s is the current state, a is the action taken in a given state, μ is the Actor network; |θ μ is the parameter of the Actor network; Q is the Critic network; θ Q is the parameter of the Critic network; The update strategy of the Actor network is: in, is the objective function J with respect to the Actor network parameter θ μ The gradient of ; E is the expected value operator; Indicates that in state s t Next, when action a is generated by the Actor network μ(s t ) is decided, the gradient of the Critic network about action a; Indicates that in state s t The Actor network has parameters θ μ The gradient of The update strategy of the Critic network is: L(θ Q )=Ε[(r+γQ′(s′,u′(s′|θ μ′ )|θ Q )-Q(s,a|θ Q )) 2 ] Among them, L(θ Q ) is the loss function of the Critic network, which is used to measure the accuracy of the value function prediction; r is the reward, γ is the discount factor, Q′(s′,u′(s′|θ μ′ )|θ Q ) is the next state s′ according to the updated Actor network u′(s′|θ μ′ ) is the value estimate of the action selected; Q(s,a|θ Q ) is the estimated value of taking action a in the current state s.

8. A system using the aggregation control optimization method applicable to distributed energy as claimed in any one of claims 1 to 7, characterized in that: include: A data collection and preprocessing module, used for collecting first data and performing preprocessing; A distributed energy modeling module, used to establish a mathematical model of distributed energy according to the preprocessed first data; The short-term forecasting module is used to make short-term forecasts of load demand and renewable energy output using a hybrid deep learning model; The multi-objective optimization module is used to establish and solve the multi-objective optimization model based on the mathematical model of distributed energy and the results of short-term prediction to obtain the multi-objective optimization results; The decision generation module is used to select the optimal control decision from the multi-objective optimization results and generate a specific control strategy; The real-time monitoring and fault handling module is used to introduce hierarchical control mechanism and adaptive control mechanism based on control strategy, monitor the system operation status in real time, detect and handle faults.

9. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the aggregation control optimization method applicable to distributed energy as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the aggregation control optimization method applicable to distributed energy as described in any one of claims 1 to 7.

Images