Intelligent scheduling method and system for intelligent mining of water-light complementary characteristics

Abstract

The invention discloses an intelligent scheduling method and system for intelligent mining of water-light complementation characteristics, and belongs to the technical field of renewable energy scheduling, and the method specifically comprises the steps: obtaining the operation data and meteorological data of a distributed photovoltaic power station and a cascade small hydropower station, and carrying out the preprocessing; analyzing a dynamic complementary relationship between the distributed photovoltaic and the cascade small hydropower station, and constructing a dynamic water-light complementary characteristic model; based on the dynamic water-light complementary characteristic model, in combination with photovoltaic output fluctuation, cascade small hydropower station regulation capability and reservoir scheduling constraint conditions, a distributed photovoltaic and cascade small hydropower station combined power generation model is constructed, and an optimal scheduling scheme is solved through an intelligent optimization scheduling algorithm; establishing a combined operation control system, dynamically adjusting an output plan according to the optimal scheduling scheme, and optimizing an intelligent optimization scheduling algorithm and a combined power generation model through a feedback mechanism; intelligent mining and optimal scheduling of water-light complementary characteristics are realized, and the energy utilization efficiency and the system stability are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of renewable energy scheduling, and specifically to an intelligent scheduling method and system for intelligently mining the complementary characteristics of water and light. Background Art

[0002] With the rapid development of distributed photovoltaic and cascaded small hydropower, the combined operation of hydropower stations and photovoltaic power plants has become an important approach to addressing energy volatility and intermittency. Distributed photovoltaic power is significantly affected by weather, resulting in fluctuating and intermittent output. However, cascaded small hydropower boasts strong regulation capabilities and high energy storage efficiency. Therefore, achieving combined operation control and intelligent scheduling of these two systems has become a key issue in the renewable energy sector. Existing technologies often employ fixed rules or simple optimization models for hydropower-photovoltaic hybrid scheduling, failing to fully exploit the dynamic complementary characteristics of distributed photovoltaic and cascaded small hydropower, resulting in suboptimal scheduling and low energy utilization.

[0003] For example, Chinese patent publication number CN117543721B discloses a method, apparatus, equipment, and medium for optimizing the scheduling of a cascaded hydropower-wind-photovoltaic system. These methods include establishing scheduling constraints for the cascaded hydropower-wind-photovoltaic system, including constraints on the power characteristics, hydraulic characteristics, and hydropower coupling characteristics of the cascaded hydropower stations, and constraints on the hydropower-wind-photovoltaic coupling characteristics and output characteristics of the cascaded hydropower-wind-photovoltaic system. Based on these constraints, a scheduling model for the cascaded hydropower-wind-photovoltaic system is constructed, taking the sum of the medium- and long-term scheduling utility and the short-term scheduling utility of the cascaded hydropower-wind-photovoltaic system as the objective function. The scheduling model is then solved using a linearization method to obtain an optimized scheduling result. This technical solution can provide a power generation plan for a cascaded hydropower-wind-photovoltaic system that satisfies these constraints, improving the overall system utility, reducing wind and solar curtailment, achieving resource integration of the new energy power generation system, and ensuring the safe and stable operation of the power grid.

[0004] The above existing technologies have the following problems: they rely on static constraints and objective functions and lack adaptability to dynamic changes in the system; they adopt traditional optimization methods, such as linearization methods, and do not fully utilize data-driven methods; they do not introduce real-time feedback mechanisms and online optimization technologies, and cannot dynamically adjust the scheduling strategy according to actual operating data; although the constraints of the water-wind-solar complementary system are taken into account, the dynamic complementary characteristics between water, wind and light are not deeply explored; they mainly focus on medium- and long-term scheduling utility and short-term scheduling utility, and do not fully consider multiple time scales, such as ultra-short-term, short-term, medium-term and long-term scheduling needs. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention proposes an intelligent scheduling method and system for intelligent mining of the complementary characteristics of water and photovoltaic power stations, which obtains the operating data and meteorological data of distributed photovoltaic power stations and cascade small hydropower stations and performs preprocessing; analyzes the dynamic complementary relationship between distributed photovoltaic and cascade small hydropower stations, and constructs a dynamic water-photovoltaic complementary characteristic model; based on the dynamic water-photovoltaic complementary characteristic model, combined with photovoltaic output fluctuations, cascade small hydropower regulation capabilities and reservoir scheduling constraints, constructs a joint power generation model of distributed photovoltaic and cascade small hydropower stations, and solves the optimal scheduling plan through an intelligent optimization scheduling algorithm; establishes a joint operation control system, dynamically adjusts the output plan according to the optimal scheduling plan, and optimizes the intelligent optimization scheduling algorithm and the joint power generation model through a feedback mechanism; realizes the intelligent mining and optimal scheduling of the complementary characteristics of water and photovoltaic power stations, and improves energy utilization efficiency and system stability.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The intelligent scheduling method for intelligent mining of the complementary characteristics of water and solar power includes:

[0008] Step S1: Obtaining the operating data of distributed photovoltaic power stations and cascade small hydropower stations, and at the same time, obtaining meteorological data, and preprocessing them to obtain operating characteristic data;

[0009] Step S2: Based on the operational characteristic data, a machine learning algorithm is used to analyze the output characteristics of distributed photovoltaics and cascaded small hydropower, and the dynamic complementary relationship between distributed photovoltaics and cascaded small hydropower is explored. Based on this dynamic complementary relationship, a deep reinforcement learning algorithm is combined to construct a dynamic hydro-photovoltaic complementary characteristic model to predict photovoltaic output and cascaded small hydropower regulation capacity at different time scales.

[0010] Step S3: Based on the hydropower-photovoltaic complementary characteristic model, a joint power generation model of distributed photovoltaic and cascade small hydropower is constructed, and an intelligent optimization scheduling algorithm is used to solve the optimal scheduling plan of the joint power generation model, and the optimal scheduling plan is output to the joint operation control system;

[0011] Step S4: According to the optimal scheduling plan, the joint operation control system dynamically adjusts the output plan of distributed photovoltaic and cascade small hydropower, monitors the operation effect in real time, collects actual operation data, and compares the actual operation data with the predicted results through the feedback mechanism. Based on the comparison results, the intelligent optimization scheduling algorithm and the joint power generation model are dynamically optimized using online learning technology.

[0012] Specifically, the specific steps of step S2 include:

[0013] S2.1: Obtaining operational characteristic data , input the running feature data into the pre-loaded random forest for training to obtain a trained decision tree model, where, represents the i-th operating feature, and , N represents the number of running features;

[0014] S2.2: According to each characteristic The split nodes and corresponding split gains in the trained decision tree model are combined with the interaction between features and time decay to calculate the importance score of each feature. , at the same time, Sort and select the first M feature data to obtain the filtered running feature data ,and ;

[0015] S2.3: Load the pre-built random forest model and input the filtered operating characteristic data into the pre-built random forest model for training to obtain the prediction results of photovoltaic output and cascade small hydropower regulation capacity.

[0016] Specifically, the specific steps of step S2 also include:

[0017] S2.4: Use a clustering algorithm to perform cluster analysis on the prediction results of PV output and cascade small hydropower regulation capacity to identify the complementary pattern between PV output and cascade small hydropower regulation capacity;

[0018] S2.5: Based on the identification results, extract the complementary relationship between photovoltaic and small hydropower under different meteorological conditions, and use the reinforcement learning algorithm to simulate the changes in the complementary relationship under different meteorological conditions and scheduling strategies;

[0019] S2.6: Combine the prediction results of the random forest model with the reinforcement learning simulation results to construct a dynamic hydro-photovoltaic complementarity model. Use this model to predict PV output and cascaded small hydropower regulation capacity at different time scales.

[0020] S2.7: Collect real-time operation data in real time, input the real-time operation data into the dynamic water-photovoltaic complementary characteristic model, use the online learning algorithm to update the dynamic water-photovoltaic complementary characteristic model parameters, and obtain an updated dynamic water-photovoltaic complementary characteristic model.

[0021] Specifically, the specific steps of S2.4 include:

[0022] S2.41: Align the forecast results of PV output and cascade small hydropower regulation capacity by time point, organize them into a forecast result dataset, and standardize the data in the forecast result dataset. Each row in the forecast result dataset represents the forecast value at a time point.

[0023] S2.42: Set the number of clusters and use the clustering algorithm to cluster the data in the standardized prediction result dataset, outputting the cluster label for each data point;

[0024] S2.43: For each cluster, the mean of the data points in each cluster is taken as the cluster center, and the photovoltaic output and cascade small hydropower regulation capacity values ​​of the cluster center are extracted;

[0025] If the center point of the cluster shows high PV output and low SHP regulation capacity, it means that in the current situation, PV is used as the power generation source and SHP is used as the supplementary power source;

[0026] If the center point of the cluster shows low PV output and high SHP regulation capacity, it means that under the current situation conditions, SHP serves as the power generation source and PV serves as the supplementary power source;

[0027] S2.44: Based on the judgment results of S2.43, define the types of complementary modes, including high photovoltaic + low small hydropower and low photovoltaic + high small hydropower;

[0028] S2.45: Based on the clustering results and complementary patterns, draw a scatter plot of PV output and cascade small hydropower regulation capacity, and use different colors to represent different clusters. At the same time, mark the center point of each cluster in the scatter plot.

[0029] Specifically, the specific steps of S2.5 include:

[0030] S2.51: Associate the center point of each cluster with the corresponding meteorological conditions to extract the complementary relationship between photovoltaic and small hydropower under different meteorological conditions;

[0031] S2.52: Based on the complementary relationship between photovoltaics and small hydropower under different meteorological conditions, a reinforcement learning environment is constructed, including the definition of a state space, an action space, and a reward function. The state space includes current meteorological conditions, photovoltaic output, small hydropower regulation capacity, and reservoir water level; the action space includes reservoir water release strategies and photovoltaic output adjustment strategies.

[0032] S2.53: Load a pre-trained reinforcement learning model and input different meteorological conditions and scheduling strategies into the pre-trained reinforcement learning model to simulate changes in the complementary relationship; the scheduling strategies include reservoir water release strategies and photovoltaic output adjustment strategies;

[0033] S2.54: Based on the simulation results, extract the changing patterns of the complementary relationship between photovoltaic and small hydropower under different meteorological conditions and scheduling strategies.

[0034] Specifically, the specific steps of S2.6 include:

[0035] S2.61: Align the prediction results of the random forest model with the reinforcement learning simulation results by timestamp and perform feature fusion to obtain a high-dimensional feature dataset;

[0036] S2.62: Load a pre-built machine learning model and train it using a high-dimensional feature dataset to obtain a dynamic water-light complementary characteristic model.

[0037] S2.63: Obtain meteorological conditions and reservoir water level data at different time scales, and input them into the dynamic water-photovoltaic complementary characteristic model to obtain photovoltaic output and cascade small hydropower regulation capacity at different time scales.

[0038] Specifically, the specific steps of step S3 include:

[0039] S3.1: Collect data on photovoltaic output fluctuations and cascade small hydropower regulation capabilities;

[0040] S3.2: Set reservoir operation constraints ,in, Indicates the reservoir water level, and Represent the lower and upper limits of the reservoir water level, represents the reservoir flow, and represent the lower and upper limits of the reservoir flow, respectively;

[0041] S3.3: Setting up multi-objective optimization functions ,in, represents power generation efficiency, Cost represents power generation cost, represents carbon emissions, 、 、 represents the weight coefficient;

[0042] S3.4: Based on the hydropower-photovoltaic complementary characteristics model, combined with PV output fluctuation data, cascade small hydropower regulation capacity data, and reservoir scheduling constraints, a joint power generation model is constructed;

[0043] S3.5: Use the intelligent optimization scheduling algorithm to solve the joint power generation model and obtain the optimal scheduling solution;

[0044] S3.6: Transmit the optimal scheduling plan to the joint operation control system.

[0045] Specifically, the specific steps of S3.5 include:

[0046] S3.51: Set the differential evolution algorithm parameters, including population size, scaling factor, crossover probability, and maximum number of iterations;

[0047] S3.52: Randomly generate an initial population, each individual in the population Represents a scheduling plan;

[0048] S3.53: Calculate the objective function value of each individual based on the multi-objective optimization function in S3.3 ;

[0049] S3.54: Pass Generate mutant individuals , and perform crossover operations on the variant individuals and the original individuals to generate the test individuals ,in, 、 、 represents three different individuals selected randomly, represents the scaling factor;

[0050] S3.55: Comparison of test subjects and the original individual The objective function value of

[0051] like , then select the experimental individual Entering the next generation;

[0052] like , then select the original individual Entering the next generation;

[0053] S3.56: If the current number of iterations reaches the maximum number of iterations, the differential evolution algorithm is terminated and the individual with the highest objective function value in the population is selected as the optimal scheduling solution.

[0054] Specifically, the joint operation control system includes grid-connected joint control, fast smoothing control, off-grid coordinated control and advanced predictive control functions.

[0055] An intelligent scheduling system for intelligent mining of the complementary characteristics of hydropower and photovoltaic power generation, including: data acquisition module, complementary analysis module, joint power generation model construction module, and feedback optimization module;

[0056] The data acquisition module is used to obtain multi-source operating data from distributed photovoltaic power stations and cascade small hydropower systems, perform intelligent preprocessing, and store the preprocessed data in a distributed database;

[0057] The complementary analysis module uses machine learning algorithms and reinforcement learning technology to analyze the output characteristics of photovoltaic and small hydropower, explore dynamic complementary relationships, and build a dynamic hydro-photovoltaic complementary characteristic model;

[0058] The joint power generation model building module builds a joint power generation model based on the water-photovoltaic complementary characteristic model, and designs a multi-objective optimization function to solve the optimal scheduling solution;

[0059] The feedback optimization module dynamically adjusts the output plan according to the optimal scheduling plan, monitors the operation effect in real time, and optimizes the model and algorithm through the feedback mechanism.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] 1. The present invention proposes an intelligent scheduling system that intelligently mines the complementary characteristics of water and light, and optimizes and improves the architecture, operating steps and processes. The system has the advantages of simple processes, low investment and operating costs, and low production costs.

[0062] 2. The present invention proposes an intelligent scheduling method for intelligent mining of the complementary characteristics of water and photovoltaic power. This intelligent scheduling method integrates distributed photovoltaic power stations, cascade small hydropower and meteorological data, and uses machine learning and deep reinforcement learning algorithms to mine the complementary characteristics of water and photovoltaic power. It achieves accurate prediction of photovoltaic output and hydropower regulation capacity at different time scales, thereby improving the prediction accuracy and flexibility of the energy system.

[0063] 3. The present invention proposes an intelligent scheduling method for intelligent mining of the complementary characteristics of water and photovoltaic power generation. By constructing a joint power generation model and applying an intelligent optimization scheduling algorithm, it can formulate an optimal hydro-photovoltaic complementary power generation scheduling plan and adjust the operation plan in real time to adapt to actual changes. At the same time, it uses online learning technology to continuously optimize the model and algorithm, ensuring the efficient and stable operation of the energy system and improving energy utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 Schematic diagram of the intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics of the present invention;

[0065] Figure 2 This is a flow chart showing the principle of the intelligent scheduling method for intelligently mining the water-photovoltaic complementary characteristics of the present invention;

[0066] Figure 3 This is the architecture diagram of the intelligent scheduling system for intelligent mining of the water-photovoltaic complementary characteristics of the present invention. DETAILED DESCRIPTION

[0067] Example 1

[0068] See also Figure 1 and Figure 2 The present invention provides an embodiment of an intelligent scheduling method for intelligently mining the complementary characteristics of water and light, comprising the following steps:

[0069] Step S1: Obtaining the operating data of distributed photovoltaic power stations and cascade small hydropower stations, and at the same time, obtaining meteorological data, and preprocessing them to obtain operating characteristic data;

[0070] Furthermore, the specific steps of step S1 include:

[0071] (1) Collecting real-time operating data from photovoltaic power stations and small hydropower systems, and obtaining meteorological data from weather stations or weather forecast systems, and storing the collected operating data and meteorological data in a distributed database; the meteorological data includes light intensity, temperature, rainfall, etc.;

[0072] (2) Clean the collected real-time operation data and meteorological data, process missing values ​​and outliers, and ensure data quality;

[0073] (3) The meteorological data, geographic information system data and real-time operation data are integrated to generate a high-dimensional feature data set, and the high-dimensional feature data set is normalized to obtain the operation feature data.

[0074] Step S2: Based on the operational characteristic data, a machine learning algorithm is used to analyze the output characteristics of distributed photovoltaics and cascaded small hydropower, and the dynamic complementary relationship between distributed photovoltaics and cascaded small hydropower is explored. Based on this dynamic complementary relationship, a deep reinforcement learning algorithm is combined to construct a dynamic hydro-photovoltaic complementary characteristic model to predict photovoltaic output and cascaded small hydropower regulation capacity at different time scales.

[0075] Step S3: Based on the hydropower-photovoltaic complementary characteristic model, combined with the photovoltaic output fluctuation, the regulation capacity of cascade small hydropower and reservoir scheduling constraints, a joint power generation model of distributed photovoltaic and cascade small hydropower is constructed, and an intelligent optimization scheduling algorithm is used to solve the optimal scheduling plan of the joint power generation model, and the optimal scheduling plan is output to the joint operation control system;

[0076] The joint operation control system includes grid-connected joint control, fast smoothing control, off-grid coordinated control and advanced predictive control functions;

[0077] Step S4: According to the optimal scheduling plan, the joint operation control system dynamically adjusts the output plan of distributed photovoltaic and cascade small hydropower, monitors the operation effect in real time, collects actual operation data, and compares the actual operation data with the predicted results through the feedback mechanism. Based on the comparison results, the intelligent optimization scheduling algorithm and the joint power generation model are dynamically optimized using online learning technology.

[0078] Furthermore, the specific steps of step S4 include:

[0079] (1) Analyze the optimal dispatching plan, obtain the output plan of distributed photovoltaic and cascade small hydropower, dynamically adjust the output plan of distributed photovoltaic and cascade small hydropower according to the current meteorological conditions, load demand and equipment status, and output the adjusted output plan to the distributed photovoltaic power station and cascade small hydropower system to execute the dispatching;

[0080] (2) Collect actual output data, meteorological data, and reservoir water level data of distributed photovoltaic and cascade small hydropower, and store the collected actual operating data in a database or data warehouse;

[0081] (3) Compare the actual operation data with the predicted results, and calculate the error by averaging the differences between all the actual operation data and the predicted results;

[0082] (4) Using the stochastic gradient descent algorithm to update the parameters of the intelligent optimization scheduling algorithm and the joint power generation model, and optimizing the intelligent optimization scheduling algorithm and the joint power generation model according to the updated parameters to achieve rolling optimization, wherein the stochastic gradient descent algorithm is the existing technical content in this field and is not the inventive solution of this application, and will not be described in detail here.

[0083] The specific steps of step S2 include:

[0084] S2.1: Obtaining operational characteristic data , input the running feature data into the pre-loaded random forest for training to obtain a trained decision tree model, where, represents the i-th operating feature, and , N represents the number of running features;

[0085] S2.2: According to each characteristic The split nodes and corresponding split gains in the trained decision tree model are combined with the interaction between features and time decay to calculate the importance score of each feature. , at the same time, Sort and select the first M feature data to obtain the filtered running feature data ,and ,in, represents the importance score of feature i, n represents the number of decision trees, represents the splitting node of the j-th decision tree, represents the splitting gain of feature i at node t, represents the weight coefficient of feature i at node t, represents the interaction coefficient between feature i and feature k, represents a time attenuation factor; the operating characteristic data include photovoltaic power generation, light intensity, reservoir water level, and flow;

[0086] Further, Represents the weight coefficient of feature i at node t, and satisfies ,in, represents the depth of node t, Indicates the number of samples of node t.

[0087] Further, Represents the interaction coefficient between feature i and feature k, and satisfies ,in, represents the interaction strength coefficient, Represents the correlation coefficient between feature i and feature k.

[0088] Further, represents the time decay factor and satisfies ,in, represents the decay rate, Indicates the current time, Represents the timestamp of node t.

[0089] It should be noted that the split gain measures the splitting effect of feature i at the current node t. The feature importance is dynamically adjusted by the split depth and the number of samples to avoid the interference of too deep nodes or too few samples on the feature importance evaluation. The introduction of synergistic effects between features enhances the ability to evaluate feature interactions. The time decay factor can dynamically adjust the feature importance to make the model more adaptable to the changing trend of the data.

[0090] S2.3: Load the pre-built random forest model and input the filtered operational characteristic data into the pre-built random forest model for training to obtain the prediction results of PV output and cascade small hydropower regulation capacity;

[0091] Furthermore, the specific steps of S2.3 include:

[0092] (1) Load the pre-built random forest model framework from the model storage path and use the machine learning library Scikit-learn to load the model;

[0093] (2) The filtered operational characteristic data are divided into a characteristic matrix and target variables. The characteristic matrix includes light intensity, temperature, reservoir water level, and flow rate, and the target variables include photovoltaic output and cascade small hydropower regulation capacity.

[0094] (3) Use the train_test_split function of Scikit-learn to divide the filtered running feature data into a ratio of 8:2 to obtain the training set and test set;

[0095] (4) Input the training set data into the random forest model for training to obtain a trained random forest model;

[0096] (5) Using the test set data for prediction, the prediction results of photovoltaic output and cascade small hydropower regulation capacity are obtained, and the prediction results are stored. Among them, for the random forest model, the prediction result is the average value of the prediction results of all decision trees. The random forest model is the existing technology content in this field and is not the inventive solution of this application. It will not be described in detail here.

[0097] S2.4: Use a clustering algorithm to perform cluster analysis on the prediction results of PV output and cascade small hydropower regulation capacity to identify the complementary pattern between PV output and cascade small hydropower regulation capacity;

[0098] S2.5: Based on the identification results, extract the complementary relationship between photovoltaic and small hydropower under different meteorological conditions, and use the reinforcement learning algorithm to simulate the changes in the complementary relationship under different meteorological conditions and scheduling strategies;

[0099] S2.6: Combine the prediction results of the random forest model with the reinforcement learning simulation results to construct a dynamic hydro-photovoltaic complementarity model. Use this model to predict PV output and cascaded small hydropower regulation capacity at different time scales.

[0100] S2.7: Collect real-time operation data in real time, input the real-time operation data into the dynamic water-photovoltaic complementary characteristic model, use the online learning algorithm to update the dynamic water-photovoltaic complementary characteristic model parameters, and obtain an updated dynamic water-photovoltaic complementary characteristic model.

[0101] Furthermore, the specific steps of S2.7 include:

[0102] (1) Real-time collection of distributed photovoltaic and cascade small hydropower operation data, and input of real-time operation data into the dynamic water-photovoltaic complementary characteristic model for model update and prediction;

[0103] (2) Using a stochastic gradient descent algorithm to update the parameters of the dynamic water-light complementary characteristic model, and using real-time operation data to evaluate the performance of the updated dynamic water-light complementary characteristic model, the mean square error and mean absolute error are calculated. The formulas for calculating the mean square error and mean absolute error are the existing technical content in this field and are not the inventive solution of this application, so they are not described in detail here;

[0104] (3) The updated dynamic water-photovoltaic complementary characteristic model is stored in the model library. At the same time, the updated model is deployed to the joint operation control system for real-time prediction and scheduling.

[0105] The specific steps of S2.4 include:

[0106] S2.41: Align the forecast results of PV output and cascade small hydropower regulation capacity by time point, organize them into a forecast result dataset, and standardize the data in the forecast result dataset. Each row in the forecast result dataset represents the forecast value at a time point.

[0107] S2.42: Set the number of clusters and use a clustering algorithm to cluster the data in the normalized prediction result dataset, outputting the cluster label for each data point. The clustering algorithm uses the k-means algorithm, which is prior art in this field and does not constitute an inventive solution of the present application, and is not described in detail here.

[0108] S2.43: For each cluster, the mean of the data points in each cluster is taken as the cluster center, and the photovoltaic output and cascade small hydropower regulation capacity values ​​of the cluster center are extracted;

[0109] If the center point of the cluster shows high PV output and low SHP regulation capacity, it means that in the current situation, PV is used as the power generation source and SHP is used as the supplementary power source;

[0110] If the center point of the cluster shows low PV output and high SHP regulation capacity, it means that under the current situation conditions, SHP serves as the power generation source and PV serves as the supplementary power source;

[0111] In the present invention, if the output power of a distributed photovoltaic power station is greater than or equal to 70% of its rated power, it is considered to be in a high photovoltaic output state. For example, if a photovoltaic power station with a rated power of 1,000 kilowatts has an actual output power of 700 kilowatts or above, it is considered to be in a high photovoltaic output state; if the adjustable range of the small hydropower generation power is less than 30% of its rated power and the adjustment response time exceeds 30 minutes, it is considered to be in a low small hydropower regulation capacity. For example, if a small hydropower station with a rated power of 500 kilowatts has an adjustable power range of less than 150 kilowatts and a response time of 1 hour, it is considered to be in a low small hydropower regulation capacity.

[0112] S2.44: Based on the judgment results of S2.43, define the types of complementary modes, including high photovoltaic + low small hydropower and low photovoltaic + high small hydropower;

[0113] S2.45: Based on the clustering results and complementary patterns, draw a scatter plot of PV output and cascade small hydropower regulation capacity, and use different colors to represent different clusters. At the same time, mark the center point of each cluster in the scatter plot.

[0114] The specific steps of S2.5 include:

[0115] S2.51: Associate the center point of each cluster with the corresponding meteorological conditions to extract the complementary relationship between photovoltaic and small hydropower under different meteorological conditions;

[0116] Furthermore, the specific steps of S2.51 include:

[0117] (1) Extract the center point data of each cluster from the clustering results, including photovoltaic output and cascade small hydropower regulation capacity;

[0118] (2) Obtain meteorological condition data corresponding to the timestamp of the cluster center, including light intensity, temperature, and rainfall;

[0119] (3) Align the cluster center point data with the meteorological condition data by timestamp to ensure data consistency;

[0120] (4) The cluster center point data and meteorological condition data are combined into one data set, where each data point includes the cluster center point features and the corresponding meteorological conditions;

[0121] (5) Use the correlation coefficient analysis method to analyze the relationship between the cluster center point characteristics and the meteorological conditions, and store the association analysis results. The correlation coefficient analysis method is the existing technical content in this field and is not the inventive solution of this application, so it will not be described here.

[0122] S2.52: Based on the complementary relationship between photovoltaics and small hydropower under different meteorological conditions, a reinforcement learning environment is constructed, including the definition of a state space, an action space, and a reward function. The state space includes current meteorological conditions, photovoltaic output, small hydropower regulation capacity, and reservoir water level; the action space includes reservoir water release strategies and photovoltaic output adjustment strategies.

[0123] S2.53: Load a pre-trained reinforcement learning model and input different meteorological conditions and scheduling strategies into the pre-trained reinforcement learning model to simulate changes in the complementary relationship. The scheduling strategies include reservoir water release strategies and photovoltaic output adjustment strategies. The reinforcement learning model is prior art in this field and does not constitute an inventive solution of this application, and is not described in detail here.

[0124] Among them, reservoir water release strategies and photovoltaic output adjustment strategies are key strategies for optimizing scheduling and resource allocation in cascade hydropower, wind-solar hybrid systems. The reservoir water release strategy regulates the amount of water released from reservoirs to control the power generation capacity of cascade hydropower stations to meet power demand, optimize water resource utilization, and maximize the overall benefits of the system. The photovoltaic output adjustment strategy adjusts the output of photovoltaic power stations to complement the output of cascade hydropower stations and optimize the overall output characteristics of the system. Because photovoltaic output is significantly affected by weather conditions and is volatile and intermittent, while the output of hydropower stations can be flexibly adjusted through reservoir water release strategies, the synergistic effect of reservoir water release strategies and photovoltaic output adjustment strategies can achieve hydropower complementarity and optimize the overall output characteristics of the system. For example, when photovoltaic output is high, the output of the hydropower station can be reduced to save water resources; when photovoltaic output is low, the output of the hydropower station can be increased to compensate for the shortfall in photovoltaic power generation.

[0125] S2.54: Based on the simulation results, extract the changing patterns of the complementary relationship between photovoltaic and small hydropower under different meteorological conditions and scheduling strategies.

[0126] The specific steps of S2.6 include:

[0127] S2.61: Align the prediction results of the random forest model with the reinforcement learning simulation results by timestamp and perform feature fusion to obtain a high-dimensional feature dataset. For example, features such as photovoltaic output, cascade small hydropower regulation capacity, meteorological conditions, and reservoir water levels can be combined into a single dataset.

[0128] S2.62: Load a pre-built machine learning model and train it using the high-dimensional feature dataset to obtain a dynamic water-photovoltaic complementary characteristic model. The machine learning model is prior art in this field and does not constitute an inventive solution of the present application, and is not described in detail here.

[0129] S2.63: Obtain meteorological conditions and reservoir water level data at different time scales, and input them into the dynamic water-photovoltaic complementary characteristic model to obtain photovoltaic output and cascade small hydropower regulation capacity at different time scales.

[0130] The specific steps of step S3 include:

[0131] S3.1: Collect data on photovoltaic output fluctuations and cascade small hydropower regulation capabilities;

[0132] S3.2: Set reservoir operation constraints ,in, Indicates the reservoir water level, and Represent the lower and upper limits of the reservoir water level, represents the reservoir flow, and represent the lower and upper limits of the reservoir flow, respectively;

[0133] S3.3: Setting up multi-objective optimization functions ,in, represents power generation efficiency, Cost represents power generation cost, represents carbon emissions, 、 、 represents the weight coefficient;

[0134] S3.4: Based on the hydropower-photovoltaic complementary characteristics model, combined with PV output fluctuation data, cascade small hydropower regulation capacity data, and reservoir scheduling constraints, a joint power generation model is constructed;

[0135] The process of building a joint power generation model includes:

[0136] (1) Select linear programming as the joint power generation model;

[0137] (2) Set model parameters, including objective function, constraints, and decision variables;

[0138] (3) Obtain the joint power generation model.

[0139] S3.5: Use the intelligent optimization scheduling algorithm to solve the joint power generation model and obtain the optimal scheduling solution;

[0140] S3.6: Transmit the optimal scheduling plan to the joint operation control system.

[0141] The specific steps of S3.5 include:

[0142] S3.51: Set the differential evolution algorithm parameters, including population size, scaling factor, crossover probability, and maximum number of iterations;

[0143] S3.52: Randomly generate an initial population, each individual in the population Represents a scheduling plan;

[0144] S3.53: Calculate the objective function value of each individual based on the multi-objective optimization function in S3.3 ;

[0145] S3.54: Pass Generate mutant individuals , and perform crossover operations on the variant individuals and the original individuals to generate the test individuals ,in, 、 、 represents three different individuals selected randomly, represents the scaling factor;

[0146] S3.55: Comparison of test subjects and the original individual The objective function value of

[0147] like , then select the experimental individual Entering the next generation;

[0148] like , then select the original individual Entering the next generation;

[0149] S3.56: If the current number of iterations reaches the maximum number of iterations, the differential evolution algorithm is terminated and the individual with the highest objective function value in the population is selected as the optimal scheduling solution.

[0150] Example 2

[0151] See also Figure 3 Another embodiment provided by the present invention is an intelligent scheduling system for intelligently mining the complementary characteristics of water and light, comprising:

[0152] Data acquisition module, complementary analysis module, joint power generation model construction module, feedback optimization module;

[0153] The data acquisition module is used to obtain multi-source operating data from distributed photovoltaic power stations and cascade small hydropower systems, perform intelligent pre-processing, and store the pre-processed data in a distributed database;

[0154] The complementary analysis module uses machine learning algorithms and reinforcement learning techniques to analyze the output characteristics of photovoltaic and small hydropower, explore dynamic complementary relationships, and build a dynamic hydro-photovoltaic complementary characteristic model;

[0155] The joint power generation model construction module builds a joint power generation model based on the hydropower-photovoltaic complementary characteristic model, and designs a multi-objective optimization function to solve the optimal scheduling plan;

[0156] The feedback optimization module dynamically adjusts the output plan according to the optimal scheduling plan, monitors the operating results in real time, and optimizes the model and algorithm through the feedback mechanism.

[0157] The complementary analysis modules include: feature extraction unit, machine learning analysis unit, reinforcement learning modeling unit, and prediction unit;

[0158] Feature extraction unit, used to extract key features from preprocessed data, such as the light dependence of photovoltaic output and the flow regulation characteristics of cascaded small hydropower;

[0159] Machine learning analysis unit, which uses machine learning algorithms to analyze the output characteristics of photovoltaic and small hydropower;

[0160] The reinforcement learning modeling unit uses reinforcement learning algorithms to simulate the changes in the complementary relationship under different meteorological conditions and scheduling strategies, and build a dynamic water-solar complementary characteristic model;

[0161] The prediction unit predicts the photovoltaic output and cascade small hydropower regulation capacity at different time scales based on the dynamic water-photovoltaic complementary characteristics model, where the different time scales include short-term, medium-term and long-term.

[0162] The joint power generation model construction module includes: model construction unit, optimization function design unit, intelligent optimization scheduling unit, and solution output unit;

[0163] The model building unit combines the photovoltaic output fluctuation, the regulation capacity of cascade small hydropower stations and the reservoir scheduling constraints to build a joint power generation model;

[0164] Optimization function design unit, which designs multi-objective optimization functions, where the objectives include maximizing power generation efficiency, minimizing power generation costs, and reducing carbon emissions;

[0165] Intelligent optimization scheduling unit, which uses intelligent optimization scheduling algorithm to solve the optimal scheduling plan;

[0166] The plan output unit is used to output the optimal scheduling plan to the joint operation control system.

[0167] The feedback optimization module includes: operation control unit, real-time monitoring unit, feedback comparison unit, and model optimization unit;

[0168] The operation control unit dynamically adjusts the output plans of distributed photovoltaic and cascade small hydropower according to the optimal scheduling plan;

[0169] Real-time monitoring unit, used to collect system operation data in real time, such as actual power generation, equipment status, and weather changes;

[0170] Feedback comparison unit, used to compare actual operating data with the predicted results, calculate the prediction error and analyze the source of the error;

[0171] The model optimization unit uses online learning technology to dynamically update the water-photovoltaic complementary characteristic model and the joint power generation model, and optimize the intelligent optimization scheduling algorithm.

[0172] In summary, through the design of data acquisition module, complementary analysis module, joint power generation model construction module, feedback optimization module and their corresponding units, the system realizes a complete closed loop from data acquisition, model construction, optimization scheduling to feedback optimization, which is logically reasonable and efficient.

[0173] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific embodiments. The above-mentioned specific embodiments are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention. These are all protected by the present invention.

Claims

1. An intelligent scheduling method for intelligently mining the complementary characteristics of water and light, characterized by: include: Step S1: Obtaining the operating data of distributed photovoltaic power stations and cascade small hydropower stations, and at the same time, obtaining meteorological data, and preprocessing them to obtain operating characteristic data; Step S2: Based on the operational characteristic data, a machine learning algorithm is used to analyze the output characteristics of distributed photovoltaics and cascaded small hydropower, and the dynamic complementary relationship between distributed photovoltaics and cascaded small hydropower is explored. Based on this dynamic complementary relationship, a deep reinforcement learning algorithm is combined to construct a dynamic hydro-photovoltaic complementary characteristic model to predict photovoltaic output and cascaded small hydropower regulation capacity at different time scales. Step S3: Based on the hydropower-photovoltaic complementary characteristic model, a joint power generation model of distributed photovoltaic and cascade small hydropower is constructed, and an intelligent optimization scheduling algorithm is used to solve the optimal scheduling plan of the joint power generation model, and the optimal scheduling plan is output to the joint operation control system; Step S4: According to the optimal scheduling plan, the joint operation control system dynamically adjusts the output plan of distributed photovoltaic and cascade small hydropower, monitors the operation effect in real time, collects actual operation data, and compares the actual operation data with the predicted results through the feedback mechanism. Based on the comparison results, the intelligent optimization scheduling algorithm and the joint power generation model are dynamically optimized using online learning technology.

2. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 1, characterized in that: The specific steps of step S2 include: S2.1: Obtaining operational characteristic data , input the running feature data into the pre-loaded random forest for training to obtain a trained decision tree model, where, represents the i-th operating feature, and , N represents the number of running features; S2.2: According to each characteristic The split nodes and corresponding split gains in the trained decision tree model are combined with the interaction between features and time decay to calculate the importance score of each feature. , at the same time, Sort and select the first M feature data to obtain the filtered running feature data ,and ; S2.3: Load the pre-built random forest model and input the filtered operating characteristic data into the pre-built random forest model for training to obtain the prediction results of photovoltaic output and cascade small hydropower regulation capacity.

3. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 2, characterized in that: The specific steps of step S2 also include: S2.4: Use a clustering algorithm to perform cluster analysis on the prediction results of PV output and cascade small hydropower regulation capacity to identify the complementary pattern between PV output and cascade small hydropower regulation capacity; S2.5: Based on the identification results, extract the complementary relationship between photovoltaic and small hydropower under different meteorological conditions, and use the reinforcement learning algorithm to simulate the changes in the complementary relationship under different meteorological conditions and scheduling strategies; S2.6: Combine the prediction results of the random forest model with the reinforcement learning simulation results to construct a dynamic hydro-photovoltaic complementarity model. Use this model to predict PV output and cascaded small hydropower regulation capacity at different time scales. S2.7: Collect real-time operation data in real time, input the real-time operation data into the dynamic water-photovoltaic complementary characteristic model, use the online learning algorithm to update the dynamic water-photovoltaic complementary characteristic model parameters, and obtain an updated dynamic water-photovoltaic complementary characteristic model.

4. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 3, characterized in that: The specific steps of S2.4 include: S2.41: Align the forecast results of PV output and cascade small hydropower regulation capacity by time point, organize them into a forecast result dataset, and standardize the data in the forecast result dataset. Each row in the forecast result dataset represents the forecast value at a time point. S2.42: Set the number of clusters and use the clustering algorithm to cluster the data in the standardized prediction result dataset, outputting the cluster label for each data point; S2.43: For each cluster, the mean of the data points in each cluster is taken as the cluster center, and the photovoltaic output and cascade small hydropower regulation capacity values ​​of the cluster center are extracted; If the center point of the cluster shows high PV output and low SHP regulation capacity, it means that in the current situation, PV is used as the power generation source and SHP is used as the supplementary power source; If the center point of the cluster shows low PV output and high SHP regulation capacity, it means that under the current situation conditions, SHP serves as the power generation source and PV serves as the supplementary power source; S2.44: Based on the judgment results of S2.43, define the types of complementary modes, including high photovoltaic + low small hydropower and low photovoltaic + high small hydropower; S2.45: Based on the clustering results and complementary patterns, draw a scatter plot of PV output and cascade small hydropower regulation capacity, and use different colors to represent different clusters. At the same time, mark the center point of each cluster in the scatter plot.

5. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 4, characterized in that: The specific steps of S2.5 include: S2.51: Associate the center point of each cluster with the corresponding meteorological conditions to extract the complementary relationship between photovoltaic and small hydropower under different meteorological conditions; S2.52: Based on the complementary relationship between photovoltaics and small hydropower under different meteorological conditions, a reinforcement learning environment is constructed, including the definition of a state space, an action space, and a reward function. The state space includes current meteorological conditions, photovoltaic output, small hydropower regulation capacity, and reservoir water level; the action space includes reservoir water release strategies and photovoltaic output adjustment strategies. S2.53: Load a pre-trained reinforcement learning model and input different meteorological conditions and scheduling strategies into the pre-trained reinforcement learning model to simulate changes in the complementary relationship; the scheduling strategies include reservoir water release strategies and photovoltaic output adjustment strategies; S2.54: Based on the simulation results, extract the changing patterns of the complementary relationship between photovoltaic and small hydropower under different meteorological conditions and scheduling strategies.

6. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 5, characterized in that: The specific steps of S2.6 include: S2.61: Align the prediction results of the random forest model with the reinforcement learning simulation results by timestamp and perform feature fusion to obtain a high-dimensional feature dataset; S2.62: Load a pre-built machine learning model and train it using a high-dimensional feature dataset to obtain a dynamic water-light complementary characteristic model. S2.63: Obtain meteorological conditions and reservoir water level data at different time scales, and input them into the dynamic water-photovoltaic complementary characteristic model to obtain photovoltaic output and cascade small hydropower regulation capacity at different time scales.

7. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 6, characterized in that: The specific steps of step S3 include: S3.1: Collect data on photovoltaic output fluctuations and cascade small hydropower regulation capabilities; S3.2: Set reservoir operation constraints ,in, Reservoir water level, and Represent the lower and upper limits of the reservoir water level, represents the reservoir flow, and represent the lower and upper limits of the reservoir flow, respectively; S3.3: Setting up multi-objective optimization functions ,in, represents power generation efficiency, Cost represents power generation cost, represents carbon emissions, 、 、 represents the weight coefficient; S3.4: Based on the hydropower-photovoltaic complementary characteristics model, combined with PV output fluctuation data, cascade small hydropower regulation capacity data, and reservoir scheduling constraints, a joint power generation model is constructed; S3.5: Use the intelligent optimization scheduling algorithm to solve the joint power generation model and obtain the optimal scheduling solution; S3.6: Transmit the optimal scheduling plan to the joint operation control system.

8. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 7, characterized in that: The specific steps of S3.5 include: S3.51: Set the differential evolution algorithm parameters, including population size, scaling factor, crossover probability, and maximum number of iterations; S3.52: Randomly generate an initial population, each individual in the population Represents a scheduling plan; S3.53: Calculate the objective function value of each individual based on the multi-objective optimization function in S3.3 ; S3.54: Pass Generate mutant individuals , and perform crossover operations on the variant individuals and the original individuals to generate the test individuals ,in, 、 、 represents three different individuals selected randomly, represents the scaling factor; S3.55: Comparison of test subjects and the original individual The objective function value of like , then select the experimental individual Entering the next generation; like , then select the original individual Entering the next generation; S3.56: If the current number of iterations reaches the maximum number of iterations, the differential evolution algorithm is terminated and the individual with the highest objective function value in the population is selected as the optimal scheduling solution.

9. The intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to claim 8, characterized in that: The joint operation control system includes grid-connected joint control, fast smoothing control, off-grid coordinated control and advanced predictive control functions.

10. An intelligent scheduling system for intelligent mining of water-photovoltaic complementary characteristics, which is used to implement the intelligent scheduling method for intelligent mining of water-photovoltaic complementary characteristics according to any one of claims 1 to 9, characterized in that: include: Data acquisition module, complementary analysis module, joint power generation model construction module, feedback optimization module; The data acquisition module is used to obtain multi-source operating data from distributed photovoltaic power stations and cascade small hydropower systems, perform intelligent preprocessing, and store the preprocessed data in a distributed database; The complementary analysis module uses machine learning algorithms and reinforcement learning technology to analyze the output characteristics of photovoltaic and small hydropower, explore dynamic complementary relationships, and build a dynamic hydro-photovoltaic complementary characteristic model; The joint power generation model building module builds a joint power generation model based on the water-photovoltaic complementary characteristic model, and designs a multi-objective optimization function to solve the optimal scheduling solution; The feedback optimization module dynamically adjusts the output plan according to the optimal scheduling plan, monitors the operation effect in real time, and optimizes the model and algorithm through the feedback mechanism.

Images