Cold region multi-energy flow scheduling method based on photovoltaic power generation power prediction

Abstract

The invention relates to a cold region multi-energy flow scheduling method based on photovoltaic power generation power prediction. The invention relates to the technical field of multi-energy flow system management. The method comprises the following steps: collecting multi-source data of a target area and preprocessing the collected data; based on the preprocessed data, a photovoltaic power generation electric power prediction model is constructed by adopting a deep learning method and is used for predicting photovoltaic power generation electric power in different time periods in the future; according to the photovoltaic power generation electric power prediction model, real-time prediction is carried out on future photovoltaic power generation electric power based on meteorological data collected in real time, and a short-term power prediction result is generated; and according to a real-time prediction result, generating an optimized multi-energy flow scheduling strategy, and issuing the strategy to each generator set and a load management unit. By integrating an advanced data collecting and processing technology, a deep learning algorithm and a real-time scheduling strategy generation mechanism, the prediction accuracy of the multi-energy-flow scheduling system is remarkably improved, and the 15-minute-level response speed can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-energy flow system management, and is a cold-region multi-energy flow scheduling method based on photovoltaic power prediction. Background Art

[0002] With the transformation of the global energy structure, more and more countries and regions have begun to pay attention to the use of renewable energy. With the rapid development of industrialization and urbanization, the growth of electricity demand has also accelerated. As an important green power generation method, the application of photovoltaic power generation has become increasingly strong. However, in cold regions, due to the special climate conditions, the output of photovoltaic power generation is significantly affected by weather and environmental factors, and the dynamic change of its power generation power poses challenges to energy management, requiring more accurate power prediction and real-time scheduling capabilities.

[0003] Based on the actual situation in cold regions, using photovoltaic power generation to produce hydrogen to meet the various needs of heating, hydrogen supply, power supply, etc. in cold regions puts forward higher requirements for multi-energy flow scheduling strategies. Therefore, developing an efficient photovoltaic power prediction method to improve the utilization rate of photovoltaic power generation and the overall efficiency of the system has become an urgent problem to be solved.

[0004] Therefore, there is an urgent need for a new strategy for cold-region multi-energy flow systems that can dynamically optimize the scheduling between photovoltaic power generation, electrolyzer hydrogen production units, and battery units, so as to achieve the efficient coordinated utilization of photovoltaic power generation and other energy sources, and ultimately improve the operation efficiency and response ability of multi-energy flow systems. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a cold-region multi-energy flow scheduling strategy based on photovoltaic power prediction. The cold-region multi-energy flow system includes photovoltaics, batteries, and electrolyzers, and schedules various energies by clarifying the coupling laws of electric-thermal-hydrogen multi-energy flow in the processes of conversion, transportation, and storage. This multi-energy flow scheduling strategy will help improve the energy production and utilization methods in cold regions, promote the consumption of photovoltaic power generation, and maximize hydrogen production, thus solving the problems mentioned in the background art. The present invention provides a cold-region multi-energy flow scheduling method based on photovoltaic power prediction.

[0006] The present invention provides the following technical solutions:

[0007] A cold-region multi-energy flow scheduling method based on photovoltaic power prediction, the method comprising the following steps:

[0008] Step 1: Collect multi-source data of the target area, including historical photovoltaic power generation data, meteorological data, and electrical load data, and preprocess the collected data, including data cleaning, normalization, and noise reduction.

[0009] Step 2: Based on the preprocessed data, construct a photovoltaic power prediction model using deep learning methods to predict the photovoltaic power for different future time periods.

[0010] Step 3: According to the photovoltaic power prediction model and based on the real-time collected meteorological data, perform real-time prediction on the future photovoltaic power to generate short-term power prediction results.

[0011] Step 4: Generate an optimized multi-energy flow scheduling strategy according to the real-time prediction results and issue it to each generating unit and load management unit.

[0012] Step 5: Implement a self-learning and updating mechanism to dynamically adjust and optimize the photovoltaic power prediction model and the scheduling strategy generation model.

[0013] Preferably, step 1 further includes feature extraction of the collected meteorological data:

[0014] Calculate the Pearson correlation coefficient r between the meteorological data and the photovoltaic power:

[0015] r = cov(X,Y) / (σ X σ Y )

[0016] where cov(X,Y) is the covariance and σ is the standard deviation.

[0017] The value range of the Pearson correlation coefficient is between -1 and 1: where r = 1 indicates a perfect positive correlation, r = -1 indicates a perfect negative correlation, and r = 0 indicates no correlation.

[0018] Preferably, the construction of the photovoltaic power prediction model in step 2 includes a convolutional neural network and a long short-term memory network to capture the temporal features of the photovoltaic power data. Use CNN to extract local features in the photovoltaic power time series, and then input the features extracted by CNN into LSTM to capture the long-term dependencies of the time series. Finally, convert the output of LSTM into a power prediction result through a fully connected layer, and the output layer of the model uses a linear activation function to ensure the continuity of the predicted value.

[0019] Preferably, the real-time power prediction in step 3 uses a sliding window method to dynamically update the photovoltaic power sequence data, considering the order of the time series data, ensuring that the model uses past data to predict the future during training; predicting multiple future time steps, not limited to the next time point; in the time series data, continuously using more historical data to improve the learning effect of the model.

[0020] Preferably, the optimized multi-energy flow scheduling strategy in step 4 is based on operation constraint conditions and predicted photovoltaic power generation, including the start-stop arrangement of electrolyzer units and battery power regulation instructions. Through the above method, the system simulates the dynamic interaction between electrolyzers and batteries in a photovoltaic power system and monitors and analyzes the process of photovoltaic power generation in this area for hydrogen production in real time.

[0021] Preferably, the self-learning and updating mechanism in step 5 uses feedback data to adjust model parameters to improve the accuracy and effectiveness of the photovoltaic power generation prediction model and scheduling strategy.

[0022] Preferably, the generation of the multi-energy flow scheduling strategy also includes considering grid security constraints to ensure the security and economy of the scheduling strategy.

[0023] A cold-region multi-energy flow scheduling system based on photovoltaic power generation prediction, the system includes:

[0024] Data collection and processing module: used to collect the above-mentioned multi-source data and perform data preprocessing.

[0025] Photovoltaic power generation prediction model module: used to receive the preprocessed data and establish a prediction model.

[0026] Real-time power prediction module: combines real-time meteorological data to predict future photovoltaic power generation.

[0027] Scheduling strategy generation module: generates an optimized scheduling strategy according to the real-time power prediction results and other constraint conditions.

[0028] Scheduling execution and feedback module: executes the scheduling strategy and monitors and feedbacks the execution situation.

[0029] Self-learning and model iteration module: optimizes model parameters according to feedback data to continuously improve the accuracy and effectiveness of the model.

[0030] A computer-readable storage medium, on which a computer program is stored, and the program is executed by a processor to implement a cold-region multi-energy flow scheduling method based on photovoltaic power generation prediction.

[0031] A computer device, including a memory and a processor, the memory stores a computer program, and the processor implements a cold-region multi-energy flow scheduling method based on photovoltaic power generation prediction when executing the computer program.

[0032] The present invention has the following beneficial effects:

[0033] The present invention utilizes the multiple coupling mechanisms of the power flow, heat flow, and hydrogen flow in the green hydrogen production, storage, addition, and utilization integrated system in terms of time and space to establish a multi-energy flow system and its scheduling strategy;

[0034] By integrating advanced data collection and processing technologies, deep learning algorithms, and a real-time scheduling strategy generation mechanism, the present invention significantly improves the prediction accuracy of the multi-energy flow scheduling system, achieving a response speed at the 15-minute level. Using the real-time data acquisition and processing module and the photovoltaic power prediction model module, the system can accurately predict the change trend of photovoltaic power, and timely adjust the charge and discharge power of the battery and the start-stop state of the electrolyzer;

[0035] The present invention adopts a data-driven method to study the transport characteristics of power distribution under wide power fluctuations and formulate a scheduling strategy for a synergistic multi-energy flow system. By continuously collecting feedback information on scheduling execution and continuously updating the prediction model and scheduling strategy, the system can continuously learn and adapt to the complex and changing multi-energy flow demand environment. This dynamic learning and adjustment mechanism ensures that the multi-energy flow scheduling strategy always remains optimal, while enhancing the system's ability to respond to future changes in electrical, thermal, and hydrogen loads and emergencies, significantly improving the reliability and safety of the multi-energy flow system; BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0037] Figure 1 It shows a schematic diagram of the scheduling method for a cold-region multi-energy flow system based on photovoltaic power prediction of the present invention;

[0038] Figure 2 It shows a schematic diagram of the multi-energy flow scheduling strategy for photovoltaic power prediction of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The following will clearly and completely describe the technical solutions of the present invention with reference to the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0040] The following describes the present invention in detail with specific embodiments. Specific Embodiment 1:

[0042] According to Figures 1 to 2As shown in the figure, the specific optimized technical solution adopted by the present invention to solve the above technical problems is: The present invention relates to a multi-energy flow scheduling method for cold regions based on photovoltaic power generation prediction.

[0043] The present invention provides a multi-energy flow scheduling method for cold regions based on photovoltaic power generation prediction, and the method includes the following steps:

[0044] Step 1: Collect multi-source data of the target area, including historical photovoltaic power generation data, meteorological data, and electrical load data, and preprocess the collected data, including data cleaning, normalization, and noise reduction processing;

[0045] Step 2: Based on the preprocessed data, use deep learning methods to construct a photovoltaic power generation prediction model for predicting the photovoltaic power generation of different future time periods;

[0046] Step 3: According to the photovoltaic power generation prediction model, based on the real-time collected meteorological data, predict the future photovoltaic power generation in real time to generate a short-term power prediction result;

[0047] Step 4: According to the real-time prediction result, generate an optimized multi-energy flow scheduling strategy and issue it to each generator set and load management unit;

[0048] Step 5: Implement a self-learning and update mechanism to dynamically adjust and optimize the photovoltaic power generation prediction model and the scheduling strategy generation model. Specific Embodiment 2:

[0050] The difference between Embodiment 2 and Embodiment 1 of the present application is only that:

[0051] Step 1 also includes feature extraction of the collected meteorological data:

[0052] Calculate the Pearson correlation coefficient r between the meteorological data and the photovoltaic power generation:

[0053] r = cov(X,Y) / (σ X σ Y )

[0054] where cov(X,Y) is the covariance and σ is the standard deviation;

[0055] The value range of the Pearson correlation coefficient is between -1 and 1: where r = 1 indicates a perfect positive correlation, r = -1 indicates a perfect negative correlation, and r = 0 indicates no correlation. Specific Embodiment 3:

[0057] The difference between Embodiment 3 and Embodiment 2 of the present application is only that:

[0058] In step 2, the construction of the photovoltaic power prediction model includes a convolutional neural network and a long short-term memory network to capture the temporal characteristics of photovoltaic power data. The CNN is used to extract local features in the photovoltaic power time series, and then the features extracted by the CNN are input into the LSTM to capture the long-term dependencies of the time series. Finally, the output of the LSTM is converted into a power prediction result through a fully connected layer, and a linear activation function is used in the output layer of the model to ensure the continuity of the predicted value. Specific Embodiment Four:

[0060] The difference between Embodiment Four and Embodiment Three of this application is only that:

[0061] In step 3, the real-time power prediction uses a sliding window method to dynamically update the photovoltaic power sequence data, taking into account the order of the time series data, ensuring that the model uses past data to predict the future during training; predicting multiple future time steps, not limited to the next time point; in the time series data, continuously using more historical data to improve the learning effect of the model. Specific Embodiment Five:

[0063] The difference between Embodiment Five and Embodiment Four of this invention is only that:

[0064] In step 4, the optimized multi-energy flow scheduling strategy is based on operation constraint conditions and predicted photovoltaic power, including the start-stop arrangement of electrolyzer units and battery power adjustment instructions. Through the above method, the system simulates the dynamic interaction between electrolyzers and batteries in a photovoltaic power system, and monitors and analyzes the process of using photovoltaic power generation in this area for hydrogen production in real time. Specific Embodiment Six:

[0066] The difference between Embodiment Six and Embodiment Five of this invention is only that:

[0067] In step 5, the self-learning and updating mechanism uses feedback data to adjust the model parameters to improve the accuracy and effectiveness of the photovoltaic power prediction model and the scheduling strategy. Specific Embodiment Seven:

[0069] The difference between Embodiment Seven and Embodiment Six of this invention is only that:

[0070] The generation of the multi-energy flow scheduling strategy also includes considering grid security constraints to ensure the security and economy of the scheduling strategy. Specific Embodiment Eight:

[0072] The difference between Embodiment Eight and Embodiment Seven of this invention is only that:

[0073] This invention provides a cold-region multi-energy flow scheduling system based on photovoltaic power prediction, and the system includes:

[0074] Data collection and processing module: used to collect the above-mentioned multi-source data and perform data preprocessing.

[0075] Photovoltaic power prediction model module: used to receive the preprocessed data and establish a prediction model.

[0076] Real-time power prediction module: combines real-time meteorological data to predict future photovoltaic power generation.

[0077] Scheduling strategy generation module: generates an optimized scheduling strategy according to the real-time power prediction results and other constraints.

[0078] Scheduling execution and feedback module: executes the scheduling strategy and monitors and feedbacks the execution situation.

[0079] Self-learning and model iteration module: optimizes the model parameters according to the feedback data, and continuously improves the accuracy and effectiveness of the model. Specific Embodiment Nine:

[0081] The only difference between Embodiment Nine and Embodiment Eight of the present invention is that:

[0082] The present invention provides a computer-readable storage medium, on which a computer program is stored, and the program is executed by a processor to implement a cold-region multi-energy flow scheduling method based on photovoltaic power prediction. Specific Embodiment Ten:

[0084] The only difference between Embodiment Ten and Embodiment Nine of the present invention is that:

[0085] The present invention provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements a cold-region multi-energy flow scheduling method based on photovoltaic power prediction. Specific Embodiment Eleven:

[0087] The only difference between Embodiment Eleven and Embodiment Ten of the present invention is that:

[0088] The present invention provides a cold-region multi-energy flow scheduling strategy for the multi-energy flow co-production system, which can gain a first-mover advantage in the future field of renewable energy hydrogen production and comprehensive utilization, seek important profit growth points for enterprises, and also provide important technical support and demonstration templates for promoting the large-scale utilization of renewable energy and the development of the entire hydrogen energy industry chain in China.

[0089] The following Figures 1 - 2 Combined with specific examples, the present invention is further described:

[0090] Figure 1 is a schematic diagram of the cold-region multi-energy flow system scheduling method based on photovoltaic power prediction according to the embodiments of the present invention;

[0091] Figure 2 Schematic diagram of a multi - energy flow scheduling strategy based on photovoltaic power generation prediction for an embodiment of the present invention

[0092] As Figure 1 shown, the schematic diagram of the scheduling method for a multi - energy flow system in cold regions based on photovoltaic power generation prediction includes the following steps: Step 1: Collect multi - source data of the target area, including historical photovoltaic power generation data, meteorological data, and electrical load data, and pre - process the collected data, including data cleaning, normalization, and noise reduction; Step 2: Based on the pre - processed data, use a deep learning algorithm to construct a photovoltaic power generation prediction model for predicting the photovoltaic power generation for different future time periods; Step 3: Use the constructed photovoltaic power generation prediction model, combined with the real - time collected meteorological data, to perform real - time prediction of the future photovoltaic power generation and generate short - term power prediction results; Step 4: According to the real - time prediction results, generate an optimized multi - energy flow scheduling strategy and issue it to each generator set and load management unit; Step 5: Implement a self - learning and update mechanism to dynamically adjust and optimize the photovoltaic power generation prediction model and the scheduling strategy generation model to improve the accuracy, response speed, and overall system efficiency and reliability of the multi - energy flow scheduling.

[0093] Step 1 also includes extracting the collected meteorological data to identify meteorological characteristics under different weather conditions. Specifically, first organize the collected historical meteorological data and historical photovoltaic power data, and perform preliminary cleaning to remove noise and abnormal data. Then, use the obtained data to analyze the relationship between different meteorological data and photovoltaic power, and quantify the strength and direction of this relationship through calculation to improve the accuracy of photovoltaic power generation prediction.

[0094] The construction of the photovoltaic power generation prediction model in Step 2 includes convolutional neural network (CNN) and long short - term memory network (LSTM) technologies to capture the temporal characteristics of photovoltaic power data. Specifically, use CNN to extract local characteristics in the photovoltaic power time series, and then input the features extracted by CNN into LSTM to capture the long - term dependence relationship of the time series. Finally, convert the output of LSTM into a power prediction result through a fully - connected layer, and the output layer of the model uses a linear activation function to ensure the continuity of the predicted value.

[0095] In Step 3, real-time power prediction uses the sliding window technique to dynamically update the input data of the photovoltaic prediction model, improving the timeliness of the prediction model. This extraction method takes into account the order of time series data, ensuring that the model uses past data to predict the future during training; it predicts multiple future time steps, not limited to the next time point; in time series data, more historical data is continuously used to improve the learning effect of the model. Specifically, the time series data of photovoltaic power generation is divided at a fixed step size, and the data within each window is used as an input sample; each time the sliding window moves forward by one time step, the latest data of the input sample is updated to ensure that the input data of the prediction model always reflects the latest power change trend; finally, the power of the data within each window is predicted to generate a rolling prediction result.

[0096] In Step 4, the optimized multi-energy flow scheduling strategy is based on operation constraint conditions and predicted photovoltaic power generation, including the start-stop arrangement of electrolyzer units and the battery power regulation command. Through the above method, the system simulates the dynamic interaction between electrolyzers and batteries in a photovoltaic power system, and monitors and analyzes the process of using photovoltaic power generation in this area for hydrogen production in real time.

[0097] In Step 5, the self-learning and update mechanism uses feedback data to adjust the model parameters to improve the accuracy and effectiveness of the photovoltaic power generation prediction model and the scheduling strategy. Specifically, it includes:

[0098] Real-time detection of photovoltaic power changes: Obtain the actual power change data at 15-minute intervals through the detection system of the photovoltaic power generation group, compare it with the prediction results, and adjust the prediction model parameters;

[0099] Start-stop adjustment strategy: According to the volatility of photovoltaic power, adjust the start-stop arrangement of electrolyzer units, and schedule the start-up and shutdown of electrolyzers through the strategy to maximize the hydrogen production of the target cold-region photovoltaic power generation for hydrogen production;

[0100] Power changes during the adjustment process: Set thresholds through the multi-energy flow system and its strategy to adaptively assist in controlling the photovoltaic hydrogen production system: when the photovoltaic output power is greater than the threshold, the excess power is temporarily absorbed by the battery to avoid the loss of excess power; when the photovoltaic output power is equal to the threshold, direct coupling is used for hydrogen production at this time; when the photovoltaic output power is at a low valley and lower than the threshold, the energy stored in the battery is then fed back for hydrogen production, thus achieving the maximum utilization of energy.

[0101] Dynamic adjustment strategy: During the adjustment process, continuously detect the operating status of electrolyzer units and the fluctuations of photovoltaic power generation, and continuously adjust the start-stop arrangement and battery power output to ensure the stable operation of the power system;

[0102] Feedback data collection: Collect the operation data of the electrolyzer unit and the photovoltaic power generation data after continuous adjustment as the input data for the self-learning and parameter iteration of the subsequent photovoltaic power generation prediction model, and continuously optimize the prediction model and the multi-energy flow scheduling strategy.

[0103] Furthermore, considering the grid security constraints is also included in the generation of the multi-energy flow scheduling strategy to ensure the security and economy of the scheduling strategy. Specifically, collect the real-time price data of the hydrogen market as one of the inputs for generating the scheduling strategy; then, when generating the scheduling strategy, combine the photovoltaic power generation prediction results and market price information to calculate the economic costs of different scheduling schemes; finally, select the scheduling scheme with the lowest economic cost to minimize the operating cost on the premise of ensuring that the multi-energy flow system meets the electrolyzer hydrogen production electricity load demand and the electrolyzer operation safety constraints.

[0104] Furthermore, the generation of the scheduling strategy also includes real-time monitoring of the hydrogen demand information to optimize the scheduling strategy to meet the hydrogen demand and achieve the hydrogen production target.

[0105] As Figure 2 shown, the schematic diagram of the multi-energy flow scheduling strategy for photovoltaic power generation prediction includes the following modules:

[0106] Data collection and processing module: Used to collect multi-source data, including historical photovoltaic power generation data, historical meteorological data, historical electrolyzer hydrogen production data, and historical heat demand data in this area; and perform data preprocessing, including data screening, data normalization, and data denoising, to ensure the accuracy and integrity of the data.

[0107] Photovoltaic power generation prediction model module: Used to receive the preprocessed data and establish a photovoltaic power generation prediction model using machine learning methods.

[0108] Real-time power prediction module: Use the photovoltaic power generation prediction model to construct a complete multi-energy flow system, and combine the real-time photovoltaic power generation collected from the data collection and processing module with real-time meteorological data to predict the short-term photovoltaic power generation.

[0109] Scheduling strategy generation module: Generate an optimized scheduling strategy according to the real-time power prediction results and other constraint conditions, including the start-stop arrangement of the electrolyzer unit, the battery power adjustment instruction, and the heat load plan.

[0110] Scheduling execution and feedback module: Execute the scheduling strategy and monitor and feedback the execution situation, including executing the scheduling strategy on the electrolyzer unit, the battery unit, and the heat load-related units, and continuously detecting the strategy execution situation, and feedback the adjustment information to the adjustment strategy generation module according to the actual execution situation and feedback information, and finally achieve a minute-level rapid response to dynamically adjust the multi-energy flow scheduling strategy.

[0111] Self-learning and model iteration module: Optimize model parameters according to feedback data to continuously improve the accuracy and effectiveness of the model. Specific Embodiment Twelve:

[0113] To achieve the above object, the present invention adopts the following technical solutions:

[0114] A multi-energy flow scheduling strategy based on photovoltaic power generation prediction includes the following steps:

[0115] Step 1: Collect multi-source data of the target area, including historical photovoltaic power generation data, meteorological data, and electrical load data, and preprocess the collected data, including data cleaning, normalization, and noise reduction.

[0116] Step 2: Based on the preprocessed data, use a deep learning algorithm to construct a photovoltaic power generation prediction model for predicting the photovoltaic power generation of different future time periods.

[0117] Step 3: Use the constructed photovoltaic power generation prediction model, combined with the real-time collected meteorological data, to perform real-time prediction of the future photovoltaic power generation and generate short-term power prediction results.

[0118] Step 4: Generate an optimized multi-energy flow scheduling strategy according to the real-time prediction results and issue it to each generator set and load management unit.

[0119] Step 5: Implement a self-learning and update mechanism to dynamically adjust and optimize the photovoltaic power generation prediction model and the scheduling strategy generation model to improve the accuracy, response speed, and overall efficiency and reliability of the multi-energy flow scheduling.

[0120] Further, Step 1 further includes extracting the collected meteorological data to identify meteorological characteristics under different weather conditions. Specifically, first organize the collected historical meteorological data and historical photovoltaic power data, and perform preliminary cleaning to remove noise and abnormal data. Then, use the obtained data to calculate the Pearson correlation coefficient r between the meteorological data and the photovoltaic power. The purpose of introducing this correlation coefficient is to analyze the relationship between different meteorological data and the photovoltaic power, and quantify the strength and direction of this relationship through calculation to facilitate improving the accuracy of photovoltaic power generation prediction. The specific calculation formula is: r = cov(X,Y) / (σ X σ Y ), where cov(X,Y) is the covariance and σ is the standard deviation. The value range of the Pearson correlation coefficient is between -1 and 1: where r = 1 indicates a perfect positive correlation, r = -1 indicates a perfect negative correlation, and r = 0 indicates no correlation.

[0121] Further, the construction of the photovoltaic power prediction model in Step 2 includes convolutional neural network (CNN) and long short-term memory network (LSTM) technologies to capture the temporal characteristics of photovoltaic power data. Specifically, CNN is used to extract local features in the photovoltaic power time series, and then the features extracted by CNN are input into LSTM to capture the long-term dependencies of the time series. Finally, the output of LSTM is converted into a power prediction result through a fully connected layer, and a linear activation function is used in the output layer of the model to ensure the continuity of the predicted value.

[0122] Further, in Step 3, real-time power prediction uses the sliding window technique to dynamically update the input data of the photovoltaic prediction model, improving the timeliness of the prediction model. This extraction method takes into account the order of time series data, ensuring that the model uses past data to predict the future during training; predicting multiple future time steps, not limited to the next time point; and continuously using more historical data in the time series data to improve the learning effect of the model. Specifically, the photovoltaic power time series data is divided according to a fixed step size, and the data within each window is used as an input sample; each time the sliding window moves forward by one time step, the latest data of the input sample is updated to ensure that the input data of the prediction model always reflects the latest power change trend; finally, power prediction is performed on the data within each window to generate rolling prediction results.

[0123] Further, the optimized multi-energy flow scheduling strategy in Step 4 is based on operation constraint conditions and predicted photovoltaic power, including the start-stop arrangement of electrolyzer units and battery power adjustment instructions. Through the above method, the system simulates the dynamic interaction between electrolyzers and batteries in a photovoltaic power system, and monitors and analyzes the process of using photovoltaic power generation in this area for hydrogen production in real time.

[0124] Further, the self-learning and updating mechanism in Step 5 uses feedback data to adjust model parameters to improve the accuracy and effectiveness of the photovoltaic power prediction model and the scheduling strategy. Specifically, it includes:

[0125] Real-time detection of photovoltaic power changes: Obtain the actual power change data at the 15-minute level through the detection system of the photovoltaic power generation group, and compare it with the prediction result to adjust the prediction model parameters;

[0126] Start-stop adjustment strategy: According to the volatility of photovoltaic power, adjust the start-stop arrangement of electrolyzer units, and dispatch the start-up and shutdown of electrolyzers through the strategy to maximize the hydrogen production of the target cold-region photovoltaic power generation for hydrogen production;

[0127] Power variation during the adjustment process: Set thresholds through the multi-energy flow system and its strategy to adaptively assist the photovoltaic hydrogen production system for control: When the photovoltaic output power is greater than the threshold, the excess power is temporarily absorbed by the battery to avoid the loss of excess power; When the photovoltaic output power is equal to the threshold, the direct coupling method is used for hydrogen production at this time; When the photovoltaic output power is at a low valley and lower than the threshold, the energy stored in the battery is then fed back for hydrogen production, so as to maximize the utilization of energy.

[0128] Dynamic adjustment strategy: During the adjustment process, the operating status of the electrolyzer unit and the fluctuation of the photovoltaic power generation are detected in real time, and the start-stop arrangement and the battery power output are continuously adjusted to ensure the stable operation of the power system;

[0129] Feedback data collection: Collect the operating data of the electrolyzer unit and the photovoltaic power generation data after continuous adjustment, as the input data for the self-learning and parameter iteration of the subsequent photovoltaic power generation prediction model, and continuously optimize the prediction model and the multi-energy flow scheduling strategy.

[0130] Furthermore, the generation of the multi-energy flow scheduling strategy also includes considering the grid security constraints to ensure the security and economy of the scheduling strategy. Specifically, collect the real-time price data of the hydrogen market as one of the inputs for generating the scheduling strategy; Then when generating the scheduling strategy, combine the photovoltaic power generation prediction results and the market price information to calculate the economic costs of different scheduling schemes; Finally, select the scheduling scheme with the lowest economic cost to minimize the operating cost on the premise of ensuring that the multi-energy flow system meets the electrolyzer hydrogen production electricity load demand and the electrolyzer operation security constraints.

[0131] Furthermore, the generation of the scheduling strategy also includes real-time monitoring of the hydrogen demand information to optimize the scheduling strategy to meet the hydrogen demand and achieve the hydrogen production target.

[0132] The present invention provides a multi-energy flow scheduling system for photovoltaic power generation prediction, including the following modules:

[0133] Data collection and processing module: Used to collect the above-mentioned multi-source data and perform data preprocessing.

[0134] Photovoltaic power generation prediction model module: Used to receive the preprocessed data and establish a prediction model.

[0135] Real-time power prediction module: Combine real-time meteorological data to predict the future photovoltaic power generation.

[0136] Scheduling strategy generation module: Generate an optimized scheduling strategy according to the real-time power prediction results and other constraint conditions.

[0137] Scheduling execution and feedback module: Execute the scheduling strategy and monitor and feedback the execution situation.

[0138] Self-learning and model iteration module: Optimize model parameters according to feedback data to continuously improve the accuracy and effectiveness of the model.

[0139] The above is only a preferred embodiment of a cold-region multi-energy flow scheduling method based on photovoltaic power prediction. The protection scope of a cold-region multi-energy flow scheduling method based on photovoltaic power prediction is not limited to the above embodiments. Any technical solutions within this concept belong to the protection scope of the present invention. It should be noted that for those skilled in the art, several improvements and variations made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. A multi-energy flow scheduling method for cold regions based on photovoltaic power generation prediction, characterized in that: The method includes the following steps: Step 1: Collect multi-source data of the target area, including historical photovoltaic power generation data, meteorological data, and electrical load data, and preprocess the collected data, including data cleaning, normalization, and noise reduction; Step 2: Based on the preprocessed data, use a deep learning method to construct a photovoltaic power generation prediction model for predicting the photovoltaic power generation at different future time periods; Step 3: According to the photovoltaic power generation prediction model, based on the real-time collected meteorological data, perform real-time prediction of the future photovoltaic power generation to generate a short-term power prediction result; Step 4: According to the real-time prediction result, generate an optimized multi-energy flow scheduling strategy and issue it to each generator set and load management unit; Step 5: Implement a self-learning and update mechanism to dynamically adjust and optimize the photovoltaic power prediction model and the scheduling strategy generation model.

2. The method according to claim 1, wherein: Step 1 further includes feature extraction of the collected meteorological data: Calculate the Pearson correlation coefficient r between the meteorological data and the photovoltaic power: r = cov(X,Y) / (σ X σ Y ) where cov(X,Y) is the covariance and σ is the standard deviation; The value range of the Pearson correlation coefficient is between -1 and 1: where r = 1 indicates a perfect positive correlation, r = -1 indicates a perfect negative correlation, and r = 0 indicates no correlation.

3. The method according to claim 2, wherein: The construction of the photovoltaic power generation prediction model in Step 2 includes a convolutional neural network and a long short-term memory network to capture the temporal characteristics of the photovoltaic power data. Use CNN to extract local features in the photovoltaic power time series, and then input the features extracted by CNN into LSTM to capture the long-term dependencies of the time series. Finally, convert the output of LSTM into a power prediction result through a fully connected layer, and the output layer of the model uses a linear activation function to ensure the continuity of the predicted value.

4. The method according to claim 3, wherein: The real-time power prediction in Step 3 uses a sliding window method to dynamically update the photovoltaic power sequence data, considering the order of the time series data, ensuring that the model uses past data to predict the future during training; predicting multiple future time steps, not limited to the next time point; in the time series data, continuously using more historical data to improve the learning effect of the model.

5. The method according to claim 4, wherein: The optimized multi-energy flow scheduling strategy in Step 4 is based on operation constraint conditions and predicted photovoltaic power, including the start-stop arrangement of electrolyzer units and battery power adjustment instructions. Through the above method, the system simulates the dynamic interaction between electrolyzers and batteries in a photovoltaic power system and monitors and analyzes the process of using photovoltaic power generation in this area for hydrogen production in real time.

6. The method according to claim 5, wherein: The self-learning and update mechanism in Step 5 uses feedback data to adjust the model parameters to improve the accuracy and effectiveness of the photovoltaic power prediction model and the scheduling strategy.

7. The method according to claim 6, wherein: The generation of the multi-energy flow scheduling strategy also includes considering the grid security constraints to ensure the security and economy of the scheduling strategy.

8. A multi-energy flow scheduling system for cold regions based on photovoltaic power generation prediction, characterized in that: The system includes: Data collection and processing module: used to collect the above-mentioned multi-source data and perform data preprocessing; Photovoltaic power prediction model module: used to receive the preprocessed data and establish a prediction model; Real-time power prediction module: combines real-time meteorological data to predict the future photovoltaic power generation; Scheduling strategy generation module: generates an optimized scheduling strategy according to the real-time power prediction results and other constraint conditions; Scheduling execution and feedback module: executes the scheduling strategy and monitors and feedbacks the execution situation; Self-learning and model iteration module: optimizes the model parameters according to the feedback data to continuously improve the accuracy and effectiveness of the model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the method as claimed in claims 1-7.

10. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, the method as claimed in claims 1-7 is implemented.

Images