Regional-level water-wind-solar power generation power multi-target collaborative optimization prediction system and method

Abstract

The invention, which relates to the technical field of energy management, discloses a multi-target collaborative optimization prediction system and method for regional-level water-wind-light power generation power, and the system comprises a data collection module which is used for obtaining power generation power data, meteorological environment data and power grid load demand data in real time; the optimization processor is used for receiving the data fed back by the data acquisition module, and constructing a multi-target optimization model based on spatial-temporal feature extraction to generate a plurality of candidate optimization paths; the screening module is used for screening the candidate paths through a dynamic weight distribution algorithm and screening out an optimal collaborative optimization path; and the dynamic feedback module is used for carrying out online correction on the optimal collaborative optimization path in the real-time optimization process according to the power grid frequency deviation, the energy storage charge state and the meteorological abrupt change signal. The optimization processor constructs the multi-target optimization model based on the spatial-temporal characteristics, generates a plurality of candidate optimization paths and screens out the optimal path, and compared with a traditional single energy prediction and simple scheduling mode, the precision of power generation power prediction is improved.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, and in particular to a regional-level multi-objective collaborative optimization prediction system and method for hydropower, wind power, and solar power generation. Background Technology

[0002] With the continued growth in global demand for clean energy, the proportion of renewable energy sources such as hydropower, wind power, and solar power in the electricity supply is constantly increasing. Hydropower, wind power, and solar power, with their clean and sustainable advantages, have become an important way to alleviate the energy crisis and environmental pressures. However, these energy sources are characterized by significant intermittency, volatility, and randomness, posing a huge challenge to the stable operation of the power system and the balance between electricity supply and demand.

[0003] From the perspective of power generation characteristics, hydropower generation is affected by the inflow of water, and river flow varies with seasons and climate. Power generation capacity is limited during the dry season, while during the wet season, there is a risk of water wastage. Wind farms rely on wind speed, which is unstable and difficult to predict accurately, often resulting in turbine shutdowns or low power generation efficiency due to excessively high or low wind speeds. Photovoltaic power plants are constrained by sunlight intensity, unable to generate electricity at night, and their power output fluctuates significantly under cloudy or overcast weather conditions. For example, in some regions, wind farms may experience significantly lower-than-expected power generation in summer due to persistent light winds, affecting the stability of the power supply.

[0004] In terms of grid operation, the connection of these unstable power generation sources to the grid can cause problems such as grid frequency fluctuations and voltage instability. When the power generation of wind farms or photovoltaic power plants suddenly decreases, if the grid cannot adjust the output of other power sources in time, it may lead to a decrease in grid frequency, affecting the normal operation of power equipment, and in severe cases, even causing grid failures. Moreover, due to the uncertainty of hydropower, wind power, and solar power generation, grid dispatching departments find it difficult to formulate reasonable power generation plans, resulting in energy waste and a decline in the reliability of power supply.

[0005] Furthermore, traditional power generation forecasting methods often only predict power from a single energy source, lacking consideration for the coordinated operation of multiple energy sources. These methods have limited forecasting accuracy when dealing with complex meteorological conditions and grid operation constraints, failing to meet practical needs. When faced with comprehensive forecasting and coordinated optimal scheduling of power generation from multiple energy sources within a region, traditional methods prove particularly inadequate.

[0006] To address these issues, researchers have attempted various technical approaches. Some studies utilize numerical weather prediction models to forecast the power generation of wind farms and photovoltaic power plants, but the uncertainty of meteorological data and model errors limit prediction accuracy. Other studies employ statistical models to predict power generation, but these struggle to adapt to complex and ever-changing environmental factors. Currently, while some optimized scheduling methods exist for hydropower, wind power, and solar power generation, most fail to adequately consider multi-objective collaborative optimization, thus failing to achieve both stable power supply and efficient energy utilization while minimizing equipment losses. Summary of the Invention

[0007] In view of the above-mentioned problems, the present invention is proposed.

[0008] Therefore, the problem that this invention aims to solve is that traditional power generation prediction methods often only predict a single energy source and lack synergistic consideration of multiple energy sources. When dealing with complex meteorological conditions and grid operation constraints, the prediction accuracy is limited and cannot meet actual needs.

[0009] To address the aforementioned technical problems, this invention provides the following technical solution: a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation, comprising: a data acquisition module, used to acquire real-time power generation data, meteorological environment data, and grid load demand data of hydropower stations, wind farms, and photovoltaic power stations, and feed the data back to an optimization processor; an optimization processor, used to receive the data fed back by the data acquisition module, construct a multi-objective optimization model based on spatiotemporal feature extraction, and generate several candidate optimization paths based on historical and real-time data; a screening module, used to calculate the power balance, equipment loss rate, and energy utilization rate under each candidate path, screen the candidate paths through a dynamic weight allocation algorithm, select the optimal collaborative optimization path, and control the output power of the power generation unit according to the optimal collaborative optimization path; and a dynamic feedback module, used to correct the optimal collaborative optimization path online during real-time optimization based on grid frequency deviation, energy storage state of charge, and meteorological abrupt change signals.

[0010] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the optimization processor includes: a spatiotemporal feature extraction unit, used to perform multi-scale decomposition of the spatiotemporal distribution of the collected data to extract power fluctuation features, meteorological correlation features, and load cycle features; a model fusion unit, used to input all features into a pre-trained reservoir inflow prediction model, wind power prediction model, and photovoltaic power output model to generate initial power output prediction results; and a path generation unit, used to generate the candidate optimization path based on the initial power output prediction results and grid constraints.

[0011] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the spatiotemporal feature extraction unit includes: a decomposition subunit, which uses an empirical mode decomposition algorithm to perform time-frequency decomposition on power generation data to obtain intrinsic mode components; an association subunit, which constructs a spatiotemporal association matrix between meteorological data and power generation based on a graph convolutional network; and a periodic identification subunit, which uses Fourier transform to extract periodic features of load data and constructs a load fluctuation spectrum.

[0012] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the path generation unit includes a multi-objective optimization calculation module and an extreme scenario contingency plan module. The multi-objective optimization calculation module is used to solve the Pareto front of the candidate optimization paths using an improved non-dominated sorting genetic algorithm, and to select the optimal solution from the Pareto solution set based on the entropy weight method. The extreme scenario contingency plan module pre-stores emergency optimization strategies for typhoons, droughts, and extreme temperatures.

[0013] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the multi-objective optimization calculation module includes: a constraint processing subunit, which transforms grid security constraints, equipment physical limits, and environmental limitations into multi-objective optimization boundary conditions; a population initialization subunit, which generates an initial population based on the distribution of historical best solutions; a crossover and mutation subunit, which introduces adaptive crossover probability and Cauchy mutation operator for global search; a reference point generation subunit, which dynamically adjusts the distribution of reference points based on hypervolume index; and a solution set screening subunit, which uses a clustering algorithm to reduce the dimensionality of the Pareto solution set and retains representative solutions.

[0014] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the extreme scenario contingency plan module includes: a scenario matching unit, which triggers corresponding contingency plans through satellite remote sensing data and meteorological warning signals; and a strategy migration unit, which adapts historical extreme scenario handling schemes to the current scenario based on case reasoning.

[0015] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the candidate optimization path includes a power output allocation strategy for hydropower, wind power, and solar power generation equipment, an energy storage scheduling scheme, and a grid interaction strategy; the dynamic weight allocation algorithm includes weight adjustment based on game theory, weight iteration based on Markov chains, and weight fusion based on fuzzy logic.

[0016] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the dynamic feedback module includes: a deviation detection unit, which monitors the deviation between the grid frequency and the planned output in real time; a state assessment unit, which predicts the remaining available capacity of the energy storage device based on a Kalman filter; and a path adjustment unit, which generates a corrected optimized path using a reinforcement learning algorithm.

[0017] As a preferred embodiment of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation described in this invention, the path adjustment unit includes: a state space construction subunit, which defines a joint space of grid state, energy storage state, and meteorological state; an action strategy subunit, which generates output adjustment actions based on a deep deterministic strategy gradient algorithm; and a reward function subunit, which uses a weighted sum of power balance, equipment loss rate, and energy utilization rate as the reward signal.

[0018] To address the aforementioned technical problems, this invention provides the following technical solution: a regional-level multi-objective collaborative optimization prediction method for hydropower, wind power, and solar power generation, comprising: real-time acquisition of power generation data, meteorological data, and grid load data from hydropower stations, wind farms, and photovoltaic power stations; extraction of spatiotemporal features from the acquired data to generate power fluctuation features, meteorological correlation features, and load cycle features; inputting the features into pre-trained reservoir inflow prediction models, wind power prediction models, and photovoltaic output models to generate initial output prediction results; generating several candidate optimization paths based on the initial prediction results and grid constraints; calculating the power balance, equipment loss rate, and energy utilization rate for each path; selecting the optimal collaborative optimization path using a dynamic weight allocation algorithm; controlling the output power of the power generation unit according to the optimal collaborative optimization path and providing real-time feedback correction.

[0019] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described above.

[0020] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described above.

[0021] The beneficial effects of this invention are as follows: This invention acquires rich real-time data on power generation, meteorological environment, and grid load through a data acquisition module. The optimization processor constructs a multi-objective optimization model based on spatiotemporal feature extraction, generating multiple candidate optimization paths and selecting the optimal path. Compared to traditional single-energy forecasting and simple scheduling methods, this significantly improves the accuracy of power generation forecasting, enabling more rational allocation of power output from hydropower, wind power, and solar power generation equipment, scheduling of energy storage, and arrangement of grid interaction strategies, thus achieving efficient allocation of power resources. For example, under different seasons and meteorological conditions, the proportion of hydropower, wind power, and solar power generation can be precisely adjusted according to actual conditions, avoiding energy waste.

[0022] The dynamic feedback module adjusts and optimizes the power path in real time based on grid frequency deviation, energy storage status of charge, and meteorological abrupt changes, ensuring a close match between the output power of the power generation unit and the grid load demand, effectively reducing grid frequency and voltage fluctuations. Taking sudden power changes in wind farms and photovoltaic power plants as examples, the system can quickly adjust the output of other power generation equipment or the charging and discharging status of energy storage to maintain stable grid operation, reduce the risk of grid failures caused by power generation fluctuations, and ensure the reliability of power supply. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a scene diagram of a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation in Example 1.

[0025] Figure 2 This is a structural diagram of a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation in Example 1.

[0026] Figure 3 This is a structural diagram of the first support unit of a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation in Example 2.

[0027] Figure 4 This is another perspective view of the first support unit of a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation in Example 2. Detailed Implementation

[0028] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0029] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0030] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation, comprising:

[0031] The data acquisition module is used to acquire real-time power generation data, meteorological environment data, and grid load demand data of hydropower stations, wind farms, and photovoltaic power stations, and feed the data back to the optimization processor;

[0032] An optimized processor is used to receive data from the data acquisition module, construct a multi-objective optimization model based on spatiotemporal feature extraction, and generate several candidate optimization paths based on historical and real-time data.

[0033] The filtering module is used to calculate the power balance, equipment loss rate and energy utilization rate of each candidate path, and to filter the candidate paths through a dynamic weight allocation algorithm to select the optimal collaborative optimization path, and control the output power of the power generation unit according to the optimal collaborative optimization path.

[0034] The dynamic feedback module is used to correct the optimal collaborative optimization path online based on grid frequency deviation, energy storage state of charge, and meteorological abrupt change signals during real-time optimization.

[0035] The workflow of the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation is as follows: Figure 1 As shown.

[0036] The data acquisition module acquires real-time power generation data, meteorological environmental data, and grid load demand data from hydropower stations, wind farms, and photovoltaic power stations, and feeds this data back to the optimization processor. This module achieves real-time data acquisition by setting up data acquisition devices at various power stations, meteorological monitoring stations, and grid data interfaces. For example, in hydropower stations, power generation data is acquired through sensors installed on turbines, generators, and other equipment, and water level and flow rate data are collected using hydrological monitoring equipment. In wind farms, wind speed meters, wind direction meters, and wind turbine power sensors are used to acquire wind condition and power generation data. Photovoltaic power stations rely on current and voltage sensors on photovoltaic panels to acquire power generation data. Meteorological environmental data, including temperature, humidity, light intensity, wind speed, and wind direction, is acquired through professional meteorological monitoring stations or meteorological data service platforms. Grid load demand data is acquired in real-time from the data interface of the grid dispatch center.

[0037] Further processor optimizations include:

[0038] The spatiotemporal feature extraction unit is used to decompose the spatiotemporal distribution of the collected data at multiple scales and extract power fluctuation features, meteorological correlation features, and load cycle features.

[0039] The model fusion unit is used to input all features into the pre-trained reservoir inflow prediction model, wind power prediction model, and photovoltaic output model to generate initial output prediction results.

[0040] The path generation unit is used to generate the candidate optimized path based on the initial output prediction results and grid constraints. The candidate optimized path includes the output allocation strategy of hydro, wind and solar power generation equipment, energy storage scheduling scheme and grid interaction strategy.

[0041] Among them, the specific workflow of optimizing the processor is as follows: Figure 2 As shown.

[0042] Specifically, the spatiotemporal feature extraction unit includes:

[0043] The sub-units are decomposed, and the empirical mode decomposition algorithm is used to perform time-frequency decomposition on the power generation data to obtain the intrinsic mode components. The empirical mode decomposition algorithm is an adaptive signal processing method that decomposes the power generation data into a finite number of intrinsic mode components. These intrinsic mode components reflect the fluctuation characteristics of power generation at different time scales, which helps to analyze the power change pattern in depth.

[0044] The correlation subunit constructs a spatiotemporal correlation matrix between meteorological data and power generation based on a graph convolutional network. The graph convolutional network treats meteorological data and power generation data as nodes and their spatiotemporal relationships as edges to construct a graph structure. By performing convolution operations on the graph structure data through the graph convolutional network, the spatiotemporal correlation matrix between meteorological data and power generation can be obtained. This matrix can reflect the degree of influence of meteorological factors on power generation and their spatiotemporal distribution relationship.

[0045] The system also includes a periodic identification subunit, which uses Fourier transform to extract the periodic characteristics of load data and constructs a load fluctuation spectrum. Through Fourier transform, the load data can be decomposed into different frequency components, thereby extracting the periodic characteristics of the load and constructing a load fluctuation spectrum, providing a basis for subsequent load forecasting and optimized scheduling.

[0046] Furthermore, in the model fusion unit, the pre-trained model can be trained using machine learning or deep learning algorithms, such as Support Vector Machine (SVM) and Long Short-Term Memory (LSTM). The extracted features are used as input to the model, and after calculation and training, the model outputs initial predicted values ​​for reservoir inflow, wind power, and photovoltaic output.

[0047] To further explain, the path generation unit includes a multi-objective optimization calculation module and an extreme scenario contingency plan module.

[0048] The multi-objective optimization calculation module is used to solve the Pareto front of the candidate optimization path using an improved non-dominated sorting genetic algorithm, and to select the optimal solution from the Pareto solution set based on the entropy weight method.

[0049] The extreme scenario contingency plan module stores emergency optimization strategies for typhoons, droughts, and extreme temperatures.

[0050] Specifically, the multi-objective optimization calculation module includes:

[0051] The constraint processing subunit transforms power grid security constraints, equipment physical limits, and environmental limitations into multi-objective optimization boundary conditions.

[0052] The population initialization subunit generates the initial population based on the distribution of historical best solutions.

[0053] The crossover mutation subunit introduces an adaptive crossover probability and a Cauchy mutation operator for global search.

[0054] Reference point generation sub-units are used, and the distribution of reference points is dynamically adjusted based on hypervolume indices.

[0055] The solution set is filtered into sub-units, and a clustering algorithm is used to reduce the dimensionality of the Pareto solution set while retaining representative solutions.

[0056] In the constraint processing subunit, grid security constraints include power balance constraints, voltage constraints, and frequency constraints. For example, the power balance constraint requires that the total output of hydropower, wind power, and solar power generation equipment be balanced with the grid load demand, which can be expressed as follows:

[0057]

[0058] Among them, P i Let P be the output of the i-th type of power generation equipment (hydropower, wind power, or photovoltaic), m be the total number of power generation equipment categories, and P be the output of the i-th type of power generation equipment. 储能 P represents the charging and discharging power of the energy storage device (negative during charging and positive during discharging). 电网交互For the interaction power with the power grid (positive for supplying power to the grid, negative for taking power from the grid), P 负荷 To meet the power grid load demand.

[0059] Furthermore, in the constraint processing subunit, the physical limits of the equipment include the maximum power limit of the power generation equipment, the charging and discharging power limit and capacity limit of the energy storage equipment, etc.; environmental constraints include the limitations of meteorological conditions such as light intensity and wind speed on the power generation. These constraints are transformed into boundary conditions in the optimization algorithm to ensure that the generated optimization path meets the actual operation requirements.

[0060] In the population initialization subunit, the initial population is generated by analyzing the distribution pattern of the optimal solution obtained from historical data and using a sampling method. For example, the Latin hypercube sampling method can be used to sample within the range of historical optimal solutions, so that the initial population can uniformly cover the solution space and improve the search efficiency of the algorithm.

[0061] In the crossover and mutation subunit, the adaptive crossover probability is adjusted based on the individual's fitness value. For individuals with better fitness values, their crossover probability is reduced to preserve superior genes; for individuals with poor fitness values, their crossover probability is increased to enhance gene diversity. Crossover operations can employ single-point or multi-point crossover methods, exchanging partial genes between two parent individuals to generate offspring. The Cauchy mutation operator is a non-Gaussian mutation operator that allows for broader exploration within the search space. Compared to the traditional Gaussian mutation operator, the Cauchy mutation operator has a higher probability of escaping local optima. In the mutation operation, the genes of an individual are randomly perturbed, resulting in the generation of new individuals in the solution space.

[0062] In the reference point generation sub-unit, the hypervolume index is an indicator for measuring the quality of the solution set in multi-objective optimization. It represents the size of the volume enclosed by the points in the solution set and the reference point. When the hypervolume index is large, it indicates that the quality of the current solution set is good, and the reference point can be appropriately moved closer to the solution set. When the hypervolume index is small, it indicates that the quality of the solution set needs to be improved, and the reference point can be appropriately moved away from the solution set to guide the algorithm to search for a better region.

[0063] In the solution set selection sub-unit, clustering algorithms such as K-Means clustering can be used to group similar solutions in the Pareto solution set into one class, and then a representative solution is selected from each class. This can reduce the number of solutions and improve decision-making efficiency without losing too much information. For example, by calculating the Euclidean distance between solutions, solutions that are close in distance are grouped into the same class, and then representative solutions (such as cluster centers) in each class are selected as the final candidate solutions, which facilitates the subsequent selection of the optimal solution.

[0064] Furthermore, the extreme scenario contingency plan module includes:

[0065] The scenario matching unit triggers corresponding contingency plans based on satellite remote sensing data and meteorological early warning signals.

[0066] The strategy migration unit adapts historical extreme scenario handling solutions to the current scenario based on case reasoning.

[0067] In the screening module, the power balance P is calculated. 平衡度 In this case, the ratio of the absolute value of the deviation between the actual power generation and the grid load demand to the load demand can be used as a measure, i.e.

[0068]

[0069] Equipment loss rate can be calculated based on the operating time, output, and loss characteristic curve of the power generation equipment; energy utilization rate can be obtained by calculating the ratio of actual utilization to available utilization of hydropower, wind power, and solar power.

[0070] In the screening module, the dynamic weight allocation algorithm includes game theory-based weight adjustment, Markov chain-based weight iteration, and fuzzy logic-based weight fusion.

[0071] It should be noted that the spatiotemporal feature extraction unit employs advanced technologies such as empirical mode decomposition algorithm, graph convolutional network, and Fourier transform to deeply mine the spatiotemporal features of the data, enabling the system to adapt to complex and ever-changing meteorological conditions and power grid operating environment. Whether it is short-term meteorological changes or long-term seasonal variations, it can accurately analyze their impact on power generation, providing a reliable basis for optimized scheduling and enhancing the system's adaptability.

[0072] Example 2, refer to Figure 3 and Figure 4 This is the second embodiment of the present invention, which differs from the first embodiment in that: a regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation further includes, in the dynamic feedback module, online correction of the optimal collaborative optimization path based on grid frequency deviation, energy storage state of charge, and meteorological abrupt change signals, specifically including:

[0073] The deviation detection unit monitors the deviation between the power grid frequency and the planned output in real time.

[0074] The condition assessment unit predicts the remaining available capacity of the energy storage device based on a Kalman filter.

[0075] The path adjustment unit uses a reinforcement learning algorithm to generate a corrected optimized path.

[0076] The specific workflow of the dynamic feedback module is as follows: Figure 3 As shown.

[0077] In the deviation detection unit, the grid frequency is an important indicator reflecting the power balance of the grid. Under normal circumstances, the grid frequency should be maintained within a certain range (e.g., 50Hz ± 0.2Hz). The deviation detection unit is connected to the grid frequency monitoring equipment to obtain grid frequency data in real time and compare it with the theoretical frequency corresponding to the planned output to calculate the frequency deviation value. The frequency deviation value is equal to the actual frequency minus the planned frequency. When the frequency deviation exceeds the set threshold, it indicates that the grid power is unbalanced and the output of the power generation unit needs to be adjusted.

[0078] In the state assessment unit, the Kalman filter is a commonly used state estimation method that can provide an optimal estimate of the system state even in the presence of noise. For energy storage devices, the State of Charge (SOC) is an important parameter that reflects the remaining available capacity of the energy storage device.

[0079] Assume the state equation of the energy storage device is as follows:

[0080] x k =A k x k-1 +B k u k +w k

[0081] The observation equation is,

[0082] y k =C k x k +v k

[0083] Where, x k Let A be the state at time k (i.e., SOC). k B k C k Let u be the system matrix. k To control the input (such as charging and discharging current), w k and v k These are process noise and observation noise, respectively.

[0084] By using the prediction and update steps of the Kalman filter, the State of Charge (SOC) of energy storage devices can be continuously estimated, providing an accurate basis for energy storage scheduling.

[0085] The path adjustment unit specifically includes:

[0086] State space construction sub-units define a joint space for grid state, energy storage state, and meteorological state.

[0087] The action strategy subunit generates force adjustment actions based on a deep deterministic policy gradient algorithm.

[0088] The reward function sub-unit uses a weighted sum of power balance, equipment loss rate, and energy utilization rate as the reward signal.

[0089] The specific workflow of the path adjustment unit is as follows: Figure 4 As shown.

[0090] In the state space construction sub-unit, the power grid state, in addition to frequency deviation and voltage deviation, can also include information such as the distribution of active and reactive power of the power grid and line power flow; the energy storage state not only considers the state of charge (SOC), but can also include parameters such as the charging and discharging power limits and charging and discharging efficiency of energy storage devices; the meteorological state, in addition to basic meteorological parameters, can also consider the changing trends of meteorological conditions, such as the rate of change of wind speed and the changing trend of light intensity. Combining the above information together forms a high-dimensional state space, which can more comprehensively describe the real-time state of the system.

[0091] In the action policy subunit, when using the deep deterministic policy gradient algorithm, two neural networks need to be constructed first: a policy network and a value network. The policy network generates actions, taking the state space as input and outputting a deterministic action. The value network evaluates the quality of the policy, taking the state and action as input and outputting a value function. During training, the parameters of the policy network and the value network are continuously updated so that the action generated by the policy network maximizes the value function; specifically, based on the current state s... t The strategy network outputs action a. t The environment returns to a new state based on the action. t+1 and reward r t Then, these data are used to update the parameters of the value network and policy network, continuously optimizing the action strategy.

[0092] In the reward function subunit, by setting reasonable weight coefficients for power balance, equipment loss rate and energy utilization rate, the relationship between different objectives is balanced according to actual needs, guiding the path adjustment unit to generate an optimized path that better meets actual needs.

[0093] It should be noted that the dynamic feedback module corrects and optimizes the path in real time based on grid frequency deviation, energy storage state of charge, and meteorological abrupt changes, ensuring that the output power of the power generation unit closely matches the grid load demand, effectively reducing grid frequency and voltage fluctuations. Taking sudden power changes in wind farms and photovoltaic power plants as examples, this system can quickly adjust the output of other power generation equipment or the charging and discharging state of energy storage, maintain stable grid operation, reduce the risk of grid failures caused by power generation fluctuations, and ensure the reliability of power supply.

[0094] Example 3, the third embodiment of the present invention, differs from the previous two embodiments in that: a regional-level multi-objective collaborative optimization prediction method for hydropower, wind power, and solar power generation includes:

[0095] Real-time data collection of power generation data from hydropower stations, wind farms, and photovoltaic power stations, as well as meteorological data and grid load data.

[0096] The collected data are subjected to spatiotemporal feature extraction to generate power fluctuation features, meteorological correlation features, and load cycle features.

[0097] The features are input into the pre-trained reservoir inflow prediction model, wind power prediction model, and photovoltaic power output model to generate initial power output prediction results.

[0098] Based on the initial prediction results and power grid constraints, several candidate optimization paths are generated.

[0099] Calculate the power balance, equipment loss rate, and energy utilization rate for each path.

[0100] The optimal collaborative optimization path is selected through a dynamic weight allocation algorithm.

[0101] The power output of the power generation unit is controlled according to the optimal collaborative optimization path, and feedback corrections are made in real time.

[0102] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation, characterized in that: include, The data acquisition module is used to acquire real-time power generation data, meteorological environment data, and grid load demand data of hydropower stations, wind farms, and photovoltaic power stations, and feed the data back to the optimization processor; An optimized processor is used to receive data from the data acquisition module, construct a multi-objective optimization model based on spatiotemporal feature extraction, and generate several candidate optimization paths based on historical and real-time data. The filtering module is used to calculate the power balance, equipment loss rate and energy utilization rate of each candidate path, and to filter the candidate paths through a dynamic weight allocation algorithm to select the optimal collaborative optimization path, and control the output power of the power generation unit according to the optimal collaborative optimization path. The dynamic feedback module is used to correct the optimal collaborative optimization path online based on grid frequency deviation, energy storage state of charge, and meteorological abrupt change signals during real-time optimization.

2. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 1, characterized in that: The optimized processor includes, The spatiotemporal feature extraction unit is used to perform multi-scale decomposition of the spatiotemporal distribution of the collected data and extract power fluctuation features, meteorological correlation features and load cycle features. The model fusion unit is used to input all features into the pre-trained reservoir inflow prediction model, wind power prediction model and photovoltaic output model to generate initial output prediction results. The path generation unit is used to generate the candidate optimized path based on the initial output prediction results and the power grid constraints.

3. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 2, characterized in that: The spatiotemporal feature extraction unit includes, The power generation data is decomposed into sub-units, and the empirical mode decomposition algorithm is used to perform time-frequency decomposition on the power generation data to obtain the intrinsic mode components. The correlation subunit constructs a spatiotemporal correlation matrix between meteorological data and power generation based on a graph convolutional network; The periodic identification sub-unit uses Fourier transform to extract the periodic characteristics of the load data and constructs the load fluctuation spectrum.

4. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 3, characterized in that: The path generation unit includes a multi-objective optimization calculation module and an extreme scenario contingency plan module; The multi-objective optimization calculation module is used to solve the Pareto front of the candidate optimization path using an improved non-dominated sorting genetic algorithm, and to select the optimal solution from the Pareto solution set based on the entropy weight method. The extreme scenario contingency plan module stores emergency optimization strategies for typhoons, droughts, and extreme temperatures.

5. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 4, characterized in that: The multi-objective optimization calculation module includes, The constraint processing subunit transforms power grid security constraints, equipment physical limits, and environmental limitations into multi-objective optimization boundary conditions. The population initialization sub-unit generates the initial population based on the distribution of historical best solutions. The crossover mutation subunit introduces an adaptive crossover probability and a Cauchy mutation operator for global search. Reference point generation sub-units are used to dynamically adjust the distribution of reference points based on hypervolume indices. The solution set is filtered into sub-units, and a clustering algorithm is used to reduce the dimensionality of the Pareto solution set while retaining representative solutions.

6. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 5, characterized in that: The extreme scenario contingency plan module includes: The scene matching unit triggers corresponding contingency plans based on satellite remote sensing data and meteorological warning signals; The strategy migration unit adapts historical extreme scenario handling solutions to the current scenario based on case reasoning.

7. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 6, characterized in that: The candidate optimization paths include power allocation strategies for hydro, wind and solar power generation equipment, energy storage scheduling schemes and grid interaction strategies. The dynamic weight allocation algorithm includes weight adjustment based on game theory, weight iteration based on Markov chains, and weight fusion based on fuzzy logic.

8. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 7, characterized in that: The dynamic feedback module includes, The deviation detection unit monitors the deviation between the power grid frequency and the planned output in real time. The condition assessment unit predicts the remaining available capacity of the energy storage device based on a Kalman filter. The path adjustment unit uses a reinforcement learning algorithm to generate a corrected optimized path.

9. The regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in claim 8, characterized in that: The path adjustment unit includes, State space construction sub-units define a joint space for grid state, energy storage state, and meteorological state; The action strategy subunit generates force adjustment actions based on a deep deterministic policy gradient algorithm; The reward function sub-unit uses a weighted sum of power balance, equipment loss rate, and energy utilization rate as the reward signal.

10. A regional-level multi-objective collaborative optimization prediction method for hydropower, wind power, and solar power generation, using the regional-level multi-objective collaborative optimization prediction system for hydropower, wind power, and solar power generation as described in any one of claims 1 to 9, characterized in that: include, Real-time data collection of power generation data, meteorological data, and grid load data from hydropower stations, wind farms, and photovoltaic power stations; Spatiotemporal features are extracted from the collected data to generate power fluctuation features, meteorological correlation features, and load cycle features; The features are input into the pre-trained reservoir inflow prediction model, wind power prediction model, and photovoltaic output model to generate initial output prediction results; Based on the initial prediction results and power grid constraints, several candidate optimization paths are generated. Calculate the power balance, equipment loss rate, and energy utilization rate for each path; The optimal collaborative optimization path is selected using a dynamic weight allocation algorithm; The power output of the power generation unit is controlled according to the optimal collaborative optimization path, and feedback corrections are made in real time.

Images