Method and system for normal phase direct radiation trend prediction and cooperative operation of heat storage system

Abstract

The invention discloses a normal phase direct radiation trend prediction and heat storage system cooperative operation method and system, and belongs to the technical field of photo-thermal power generation. The method comprises the following steps: performing data fusion and migration on historical meteorological data to form a fusion data set, performing preprocessing and coarse-grained symbolization representation processing on data in the fusion data set, performing model training on the data as a training data set, performing clustering analysis and fitting after a prediction model is obtained, and forming a plurality of sub-models for different meteorological conditions; carrying out model matching and resource prediction according to a prediction demand and the sub-model; and running rule migration of the heat storage system is carried out according to the resource prediction result. According to the method, trend prediction and operation linkage of the heat storage system are achieved, the prediction method can be combined with different working conditions, the prediction result is output as the radiation trend, the requirement of the heat storage system is better met, the requirement for starting and stopping operation guidance of all parts of the heat storage system can be met, the prediction accuracy is higher, and the method is more practical.

Description

Technical Field

[0001] This invention belongs to the field of concentrated solar power generation technology, and specifically relates to a method and system for predicting the trend of normal phase direct radiation and coordinating the operation of a thermal storage system. Background Technology

[0002] Concentrated solar power (CSP) is a novel form of solar power generation. However, current industry practices for CSP radiation forecasting largely employ methods similar to those used for photovoltaic (PV) total radiation forecasting. Despite these differences, there are significant distinctions between the two. CSP and PV differ in energy harvesting methods, system processes, operational requirements, power generation capacity, and historical resource abundance. Furthermore, because CSP incorporates molten salt thermal storage systems, its requirements for the accuracy, timeliness, and key performance indicators in resource forecasting differ.

[0003] Solar normal phase direct radiation is strongly correlated with the power generation capacity of concentrated solar power (CSP) systems, while total solar radiation is strongly correlated with photovoltaic (PV) power generation capacity. However, due to the larger proportion and earlier development of PV power plants, current resource prediction methods primarily focus on total solar radiation, lacking a one-to-one mapping with direct radiation. Compared to total solar radiation, solar normal phase radiation exhibits stronger instantaneous fluctuations, greater diurnal variations, and more influencing factors. Current prediction methods fail to optimize for these characteristics and lack personalized solutions. Furthermore, because direct radiation data is subject to numerous perturbations, current mainstream techniques primarily operate directly on raw numerical data, making prediction results more susceptible to fluctuations and leading to problems such as learning sample bias or overfitting.

[0004] The main application scenario for solar normal phase radiation prediction is solar thermal power generation projects. For solar thermal power generation projects, especially tower molten salt projects, the thermal storage stage has already filtered out the impact of small resource fluctuations. Only large changes in resources caused by factors such as large-scale cloud cover may cause operations such as salt loading and unloading, thereby affecting the operation of the system.

[0005] Therefore, predicting radiation trends is more meaningful for concentrated solar power (CSP) projects with thermal storage systems. However, current methods for predicting direct solar radiation are mainly numerical, with a lack of trend prediction. Furthermore, the commercial operation of CSP is still in its early stages, and corresponding resource assessment methods are not yet mature, as is the application of resource utilization, especially trend-based resource prediction results. Moreover, current solar radiation prediction work is mostly used for photovoltaic power prediction and generally fails to integrate with the operation of energy storage systems. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes a radiation trend prediction method and system operation optimization method based on the characteristics of concentrated solar power (CSP) resource prediction requirements. The specific technical solution is as follows:

[0007] On the one hand, this invention proposes a method for predicting the trend of normal phase direct radiation and coordinating the operation of a thermal storage system, including the following steps:

[0008] Historical meteorological data are fused and migrated to form a fused dataset;

[0009] The data in the fused dataset are preprocessed and coarse-grained symbolic representation is performed. The processed data is then used as the training dataset to train the model and obtain the prediction model.

[0010] Cluster analysis and fitting are performed on the prediction model to form multiple sub-models for different meteorological conditions;

[0011] Based on the predicted demand and the sub-model, model matching and resource prediction are performed;

[0012] The operation rules of the thermal storage system are migrated based on the resource prediction results, and the fused dataset and sub-models are updated and saved based on the operation rule migration results.

[0013] Furthermore, the process of fusing and migrating historical meteorological data to form a fused dataset includes the following steps:

[0014] The historical meteorological data were analyzed for interannual variation of total radiation, interannual variation of sunshine percentage, and rationality detection analysis of long-sequence data.

[0015] A matching analysis was performed on the long-series and short-series meteorological data in the historical meteorological data.

[0016] The total radiation data and normal phase direct radiation data from professional photometric stations are fitted together, and the fitting relationship is used to separate the normal phase direct radiation data from long-term meteorological station and satellite data to obtain supplementary data.

[0017] The supplementary data and historical meteorological data are integrated according to time series to obtain a fused dataset.

[0018] Furthermore, when integrating the supplementary data with historical meteorological data according to time series, if multiple data exist at the same time, the dataset closest to the average value is selected as the fused dataset.

[0019] Furthermore, the preprocessing and coarse-grained symbolic representation processing for each data point in the fused dataset includes the following steps:

[0020] The data in the fused dataset is normalized to obtain normalized data;

[0021] Based on the different historical meteorological data, the normalized data at each moment is represented by different dimensions n and m related moments.

[0022] The associated data is represented by coarse-grained symbols, and the continuous data is discretized.

[0023] The model is trained by using each t*m dimensional data in the symbolized data as input data and the normal direct radiation data at time t as output data to obtain the mapping relationship and prediction model.

[0024] When the prediction model is used to predict the data for the first m time steps, and n-dimensional data is used as input data, the output is the predicted value.

[0025] Furthermore, cluster analysis and fitting are performed on the prediction model to form multiple sub-models for different meteorological conditions, including the following steps:

[0026] Cluster analysis is used to calculate the relationships between prediction models, and the similarity between two features is measured based on these relationships.

[0027] Based on the clustering analysis results, the fused dataset is divided into several data subsets;

[0028] The data in each subset is divided into a training set and a validation set;

[0029] The training and validation sets are used to fit the model, resulting in multiple sub-models.

[0030] Furthermore, the step of performing model matching and resource prediction based on the predicted demand and the sub-model includes the following steps:

[0031] Organize the input data of the prediction data according to the input data format used when the prediction model was built;

[0032] The collected input data is then used to match the model using clustering methods to determine the corresponding sub-models;

[0033] The rise and fall trend of normal phase direct radiation is predicted based on the corresponding normal phase direct radiation data output by the sub-model.

[0034] Furthermore, the migration of the operating rules of the thermal storage system based on resource prediction results includes the following steps:

[0035] Based on the prediction results, combined with the current operating conditions and subsequent predictions, the operating conditions are divided into multiple modes;

[0036] Based on the aforementioned multiple modes, collect the initial thermal storage system and operation mode coordination logic rules of existing operating projects;

[0037] For each mode, select the top n operation action types and their corresponding corresponding operations as the recommended rules for output.

[0038] Furthermore, the initial logic rules for the coordination of the thermal storage system and operation mode of existing operating projects are collected, including the collection of operational actions and evaluation indicators;

[0039] The types of operation actions include: whether the heat collection system is turned on, whether the molten salt pipeline starts preheating / insulation, whether the heat storage system enters the cold salt tank / hot salt tank, whether the power generation system is turned on, and the power generation system power.

[0040] The evaluation indicators include: photoelectric conversion efficiency and photothermal conversion efficiency after removing plant power.

[0041] Secondly, this invention proposes a system for predicting the direct radiation trend of the normal phase and coordinating the operation of a thermal storage system, comprising:

[0042] The data fusion unit is used to fuse and migrate historical meteorological data to form a fused dataset;

[0043] The model training unit is used to preprocess and coarse-grained symbolic representation of the data in the fused dataset, and use the processed data as the training dataset to train the model and obtain the prediction model.

[0044] The model fitting unit is used to perform cluster analysis and fitting on the prediction model to form multiple sub-models for different meteorological conditions.

[0045] The resource prediction unit is used to perform model matching and resource prediction based on the prediction requirements and the sub-model.

[0046] The update unit is used to migrate the operation rules of the thermal storage system based on the resource prediction results, and to update and save the fused dataset and sub-model based on the operation rule migration results.

[0047] Thirdly, the present invention proposes an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0048] Memory, which stores computer programs;

[0049] The processor, when executing the program stored in the memory, implements the method for predicting the normal phase direct radiation trend and coordinating the operation of the thermal storage system.

[0050] Fourthly, the present invention proposes a computer-readable storage medium storing a computer program, which, when run, executes the method for predicting the normal phase direct radiation trend and coordinating the operation of the thermal storage system.

[0051] The beneficial effects of this invention are:

[0052] This invention directly symbolizes radiation data during the processing of historical meteorological data, removes data features based on the characteristics of direct radiation, realizes trend prediction according to application requirements, and links trend prediction with the operation of thermal storage system, proposing a collaborative method for prediction and operation strategy.

[0053] This invention also involves data migration, specifically the fusion of multiple datasets for direct radiation data, which solves the problems of timing issues and insufficient regional coverage in this type of novel data.

[0054] This invention proposes key operational action types in historical rules, corresponding rule recommendation and selection methods, and corresponding incremental update methods. The prediction method can combine different operating conditions, and the prediction results output as radiation trends, which is more in line with the needs of thermal storage systems. It can meet the requirements for start-up and shutdown operation guidance for various parts of the thermal storage system, and has higher prediction accuracy and is more practical.

[0055] This invention combines the characteristics of direct radiation variation patterns and strong correlation with weather, uses clustering methods to decompose datasets and sub-models, and achieves multi-model prediction through corresponding model partitioning and new data matching methods, thereby improving prediction accuracy.

[0056] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 A flowchart of the method for predicting the direct radiation trend of the normal phase and coordinating the operation of the thermal storage system in an embodiment of the present invention is shown;

[0059] Figure 2 A schematic diagram of coarse-graining in an embodiment of the present invention is shown;

[0060] Figure 3 This invention illustrates a schematic diagram of the coordination between prediction results and operational logic in an embodiment of the invention.

[0061] Figure 4 A schematic diagram of an electronic device proposed in an embodiment of the present invention is shown. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] This invention addresses the characteristics of direct radiation in the normal phase by symbolizing the raw data and building multiple models based on different scenarios. It combines transfer learning with multiple project scenarios and historical data, and integrates resource prediction results with the operation of the thermal storage system. Ultimately, it achieves the prediction and optimization linkage of the entire system of direct radiation resources, energy storage, and power generation. Figure 1 As shown, the specific technical solution includes the following steps:

[0064] S1: Collect historical meteorological data of the area to be predicted, and then perform data fusion and migration on the historical meteorological data to form a fused dataset;

[0065] The historical meteorological data includes direct normal radiation, total radiation, wind speed, and temperature.

[0066] Data fusion and migration can be performed through methods including long-sequence association analysis, long-short-sequence data matching analysis, direct-to-dispersive separation of total radiation data, and database integration. Exemplary processes include the following:

[0067] S1.1: Perform association analysis on long sequences, the specific process is as follows:

[0068] For the longest time series data (long-series meteorological station data), we conducted interannual variation analysis of total radiation, age variation analysis of sunshine percentage, and rationality detection analysis.

[0069] S1.2: Perform matching analysis on long and short sequence data. The specific process is as follows:

[0070] Calculate the deviation rate of data from multiple sources within the same period. When the maximum deviation rate is less than a first set value and the average deviation rate is less than a second set value, the data difference is considered small and can be used directly. For example, the first set value can be 10%, and the second set value can be 5%.

[0071] When the data deviation is greater than or equal to the above requirements, other data should be corrected to align with the direction of a professional photometer station. The correction method is as follows:

[0072] Calculate the mean and variance for each dataset;

[0073] When the mean of a dataset / the mean of a professional photometer station is greater than the first value (e.g., 1.05) or less than the second value (e.g., 0.95), multiply all data in the dataset by the mean of the professional photometer station data / the mean of the dataset.

[0074] When, in the same sequence, the specific time value of a dataset / the specific time value of a professional photometer station is greater than the third value (e.g., it could be 1.1) or less than the fourth value (e.g., it could be 0.9), the specific value in the dataset will be calculated by multiplying the third value by the specific time value of the professional meteorological station, or by multiplying the fourth value by the specific time value of the professional meteorological station.

[0075] S1.3: Perform direct-to-dispersion separation on the total radiation data, specifically including the following steps:

[0076] For data from professional photometer stations, regression analysis was used to fit the total radiation data and normal phase direct radiation data to obtain the fitting relationship between direct radiation and total radiation at the project site.

[0077] Using this fitting relationship, direct radiation is separated from the data of nearby long-term meteorological stations (actual observation stations, which generally only have total radiation data) and satellite data, and supplementary data (i.e. normal phase direct radiation data) is obtained by supplementing the relevant data.

[0078] S1.4: Integrate the supplementary dataset and historical meteorological data according to time series. When multiple datasets exist at the same time, select the dataset that is closest to the average value as the typical dataset.

[0079] S2: Perform data preprocessing and coarse-grained symbolic representation processing on the fused dataset, and use the processed data as a training dataset to train the model to obtain a prediction model; exemplary, including the following steps:

[0080] S2.1: Statistically analyze the source of each type of data in the fused dataset and perform normalization processing;

[0081] S2.2: Based on different historical meteorological data, each moment is represented by a different dimension n and m related moments;

[0082] For example, for time t, collect m time points (t-1, t-2, ..., tm) and n related features such as solar altitude angle, solar azimuth angle, direct normal radiation, total radiation, wind speed, wind direction, cloud cover, and temperature. It should be noted that these features can be adjusted based on specific historical data.

[0083] S2.3: Perform coarse-grained symbolic representation on the correlated data, discretizing the continuous resource (trend) data into, for example... Figure 2 The types shown include general categories such as: large increase, small increase, unchanged, small decrease, and large decrease.

[0084] For example, a large increase is defined as: the value at time t divided by the value at time t-1 is greater than 150%;

[0085] The value at time t divided by the value at time t-1 is between 120% and 150%.

[0086] "Equal" means that the value at time t divided by the value at time t-1 is between 80% and 120%.

[0087] The smaller decrease is defined as: the value at time t divided by the value at time t-1 is between 50% and 80%;

[0088] A significant decrease is defined as: the value at time t divided by the value at time t-1 is less than 50%.

[0089] The corresponding range can be represented using an encoding format. When there are two types, a single binary code is used; when there are 3 to 4 types, a two-bit binary code is used; and when there are 5 to 8 types, a three-bit binary code is used.

[0090] S2.4: Take each piece of symbolically represented resource data containing t*m dimensional data as input data and the normal direct radiation data at time t as output data for model training to obtain the mapping relationship and prediction model;

[0091] S2.5: When using the prediction model to predict the data at the first m time steps, with n-dimensional data as input data, the output is the predicted value.

[0092] S3: Perform cluster analysis and fitting on the prediction model to form multiple sub-models for different meteorological conditions; exemplary, this includes the following steps:

[0093] S3.1: Calculate the interrelationships between prediction models and measure the similarity between two features based on the relationships;

[0094] A sample is generated from n*m historical meteorological data points (n different dimensions and m related times) at a specific moment, with the direct radiation rise and fall trend corresponding to that moment as the input and the direct radiation rise and fall trend as the output.

[0095] For data points in the existing database, calculate their similarity pairwise;

[0096] After the similarity calculation is completed, the corresponding numbers and similarity values ​​of the two data points should be recorded.

[0097] Record the similarity values ​​between all data, establish corresponding network relationships, use clustering methods to detect communities, and finally form multiple more closely connected communities.

[0098] S3.2: Based on the cluster analysis results, the fused dataset is divided into several data subsets; each closely related community is a data subset. According to the above analysis results, the data in each subset has stronger correlation and can be more effectively fitted to the model.

[0099] S3.3: In each subset, a machine learning algorithm is used to fit the model. The specific process is as follows:

[0100] The data in each subset is divided into a training set and a validation set, and the grouping method is random sampling or cross-validation.

[0101] Model fitting is performed using non-mechanistic modeling machine learning methods; these methods include, but are not limited to, neural networks (BP neural networks, radial basis function neural networks, deep neural networks), support vector machines, and logistic regression.

[0102] To avoid overfitting and underfitting, the parameters are optimized based on the performance of the prediction model on the training and validation sets, ultimately forming multiple sub-models for different meteorological conditions.

[0103] S4: Perform model matching based on the predicted demand and the sub-model, and then use the matched sub-model to perform resource prediction; specifically including the following steps:

[0104] S4.1: When it is necessary to make actual data predictions, the input of the prediction data should be organized in accordance with the input data format used when the model was built (e.g., the input data format described in step S2.4);

[0105] S4.2: Match the model based on the processed input data. Clustering methods (k-means, hierarchical clustering, k-nearest neighbors, etc.) can be used to determine the corresponding sub-model.

[0106] S4.3: Output the corresponding normal phase direct radiation data based on the sub-model. If longer sequence prediction data is needed, this process can be slid to continue making predictions at times t+1, t+2, and even t+k.

[0107] It should be noted that the resource forecast results represent the future situation of the aforementioned relevant meteorological data.

[0108] S5: Based on the resource prediction results, migrate the operation rules of the thermal storage system, and update and save the fused dataset and sub-models based on the migration results. The specific process is as follows: Figure 3 As shown, the main steps include resource prediction, collection of operational logic rules, generation of recommendation rules, and updating of operational data. For example, the following method can be used:

[0109] S5.1: Based on the resource and operational status predictions obtained from the above sub-models, and combined with the current operational status and subsequent predictions, the operational status is divided into multiple modes;

[0110] For example, it can be divided into the following modes:

[0111] a) Power-on status.

[0112] b) The facility is operating normally and has good future resources, with a demand for electricity.

[0113] c) The system is operating normally and has good future resources, but has no electricity demand.

[0114] d) Normal operation, future resources are average, there is a demand for electricity.

[0115] e) Normal operation, future resources are average, but there is no demand for electricity.

[0116] Based on the above models, collect the initial thermal storage system and operation mode coordination logic rules of existing operating projects; specifically, collect the types of operation actions and evaluation indicators (which can be used as operation sets). Among them, the types of operation actions include: whether the thermal collection system is turned on, whether the molten salt pipeline has started preheating / insulation, whether the thermal storage system enters the cold salt tank / hot salt tank, whether the power generation system is turned on and the power generation system power, etc.; the evaluation indicators include: photoelectric conversion efficiency and photothermal conversion efficiency after removing plant power.

[0117] For the five ae modes, the corresponding operation actions under the four operation action types with the top two photoelectric conversion efficiency and photothermal conversion efficiency are selected as recommended rules for output.

[0118] S5.2: After the operator adopts the recommended rule, the new real-time operational data is sent to the fused dataset and operation set to achieve real-time updates.

[0119] The updating and saving of the model in this invention includes the following aspects:

[0120] As time series data accumulates, the training and validation sets can be further revised and updated.

[0121] The model can be adjusted according to different time scales, such as hourly or minute-level, but the format and scale of the training data, validation data, and prediction data should be kept consistent during the adjustment process.

[0122] When the prediction model is basically stable, the corresponding multi-model system can be saved by saving the parameters and model architecture of each sub-model. When the prediction results show a large deviation or abnormal weather conditions occur in a certain place, the prediction model should be reset and the model fitting and prediction should be carried out again according to the process of this invention.

[0123] Based on the same inventive concept, one embodiment of the present invention proposes a system for predicting the direct radiation trend of the normal phase and coordinating the operation of a thermal storage system, comprising:

[0124] The data fusion unit is used to fuse and migrate historical meteorological data to form a fused dataset;

[0125] The model training unit is used to preprocess and coarse-grained symbolic representation of the data in the fused dataset, and use the processed data as the training dataset to train the model and obtain the prediction model.

[0126] The model fitting unit is used to perform cluster analysis and fitting on the prediction model to form multiple sub-models for different meteorological conditions.

[0127] The resource prediction unit is used to perform model matching and resource prediction based on the prediction requirements and the sub-model.

[0128] The update unit is used to migrate the operation rules of the thermal storage system based on the resource prediction results, and to update and save the fused dataset and sub-model based on the operation rule migration results.

[0129] The execution steps of each unit in the data fusion unit, model training unit, model fitting unit, resource prediction unit, and update unit are similar to the methods described above, and will not be explained in detail here.

[0130] Another exemplary embodiment of the present invention provides an electronic device. For example... Figure 4 As shown, the electronic device includes at least one processor 401, at least one communication interface 402, at least one memory 403 and at least one communication bus 404; wherein the processor 401, communication interface 402 and memory 403 communicate with each other through the communication bus 404.

[0131] Memory 403 stores computer programs;

[0132] The processor 401, when executing the program stored in the memory 403, implements the method for predicting the normal phase direct radiation trend and coordinating the operation of the thermal storage system.

[0133] Optionally, the communication interface can be an interface of a communication module, such as the interface of a GSM module; the processor may be a CPU, an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The memory may include high-speed RAM and may also include non-volatile memory, such as at least one disk storage device. The memory stores a program, and the processor calls the program stored in the memory to execute some or all of the above-described method embodiments.

[0134] Based on the same inventive concept, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed, implements some or all of the above-described method embodiments. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0135] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for operating a phase-direct radiation trend prediction and thermal storage system, characterized in that, The method comprises the following steps: Data fusion and migration of historical meteorological data to form a fusion data set; Preprocessing and coarse-grained symbolic representation of data in the fusion data set, using the processed data as a training data set for model training to obtain a prediction model; Cluster analysis and fitting of the prediction model to form multiple sub-models for different meteorological conditions; Model matching and resource prediction according to the prediction requirements and the sub-models; Operation rule migration of the thermal storage system according to the resource prediction results, and updating and saving of the fusion data set and the sub-models according to the operation rule migration results.

2. The method of claim 1, wherein the phase direct radiation trend prediction and thermal storage system coordination operation method is characterized by, The data fusion and migration of historical meteorological data to form a fusion data set comprises the following steps: Annual variation analysis of total radiation, annual variation analysis of sunshine percentage, and rationality detection analysis of long sequence data in the historical meteorological data; Matching analysis of long sequence meteorological data and short sequence meteorological data in the historical meteorological data; Fitting of total radiation data and direct normal radiation data in the professional photometric station data, and separation of direct normal radiation data of long time sequence meteorological stations and satellite data using the fitting relationship to obtain supplementary data; Integration of the supplementary data and the historical meteorological data according to the time sequence to obtain a fusion data set.

3. The method of claim 2, wherein the phase direct radiation trend prediction and thermal storage system coordination operation method is characterized by, If there are multiple data at the same time during the integration of the supplementary data and the historical meteorological data according to the time sequence, the data set closest to the average value is selected as the fusion data set.

4. The method of claim 2, wherein the phase direct radiation trend prediction and thermal storage system coordination operation method is characterized by, The preprocessing and coarse-grained symbolic representation of data in the fusion data set comprises the following steps: Normalization processing of the data in the fusion data set to obtain normalized data; According to the different historical meteorological data, the normalized data at each time is selected using different dimensions n, and m associated time points are represented; Coarse-grained symbolic representation of the associated data, and discretization processing of the continuous data; Using each piece of t*m-dimensional data in the symbolic representation data as input data, and the direct normal radiation data at time t as output data for model training to obtain a mapping relationship and a prediction model; Using the prediction model to predict data m times in advance, n-dimensional data as input data, and the output is the predicted value.

5. The method of claim 1, wherein the phase direct radiation trend prediction and thermal storage system coordination operation method is characterized by, The cluster analysis and fitting of the prediction model to form multiple sub-models for different meteorological conditions comprises the following steps: Using a cluster analysis method to calculate the mutual relationship between the prediction models, and measuring the similarity between two features according to the mutual relationship; According to the cluster analysis result, the fusion data set is divided into several data subsets; The data in each data subset is divided into a training set and a validation set; Model fitting using the training set and the validation set to obtain multiple sub-models.

6. The method of claim 1, wherein the phase direct radiation trend prediction and thermal storage system coordination operation method is characterized by, The model matching and resource prediction according to the prediction requirements and the sub-models comprise the following steps: Organizing the input data of the prediction data according to the input data format used when establishing the prediction model; Using the organized input data to match the model using a clustering method to determine the corresponding sub-model; According to the corresponding law phase direct radiation data output by the sub-model, a rising and falling trend of the law phase direct radiation is predicted.

7. The phase direct radiation trend prediction and thermal storage system coordinated operation method according to claim 1, characterized in that, The operation rule migration of the thermal storage system according to the resource prediction result comprises the following steps: According to the prediction result, the current operation condition and the subsequent prediction condition, the operation condition is divided into multiple modes; According to the multiple modes, the initial thermal storage system and operation mode coordination logic rules of existing operation projects are collected respectively; The specific operation actions corresponding to the operation action types ranked in the top several positions in each mode are selected as recommended rules.

8. The phase direct radiation trend prediction and heat storage system coordinated operation method according to claim 7, characterized in that, The collection of the initial thermal storage system and operation mode coordination logic rules of existing operation projects includes the collection of operation actions and evaluation indexes; The operation actions include whether the heat collection system is started, whether the molten salt pipeline is preheated / insulated, whether the thermal storage system enters the cold salt tank / hot salt tank, whether the power generation system is started and the power generation system power; The evaluation indexes include the photoelectric conversion efficiency and the photo-thermal conversion efficiency after removing the auxiliary power.

9. A phase direct radiation trend prediction and heat storage system coordinated operation system, characterized in that, It comprises: a data fusion unit for data fusion and migration of historical meteorological data to form a fusion data set; a model training unit for pre-processing and coarse-grained symbolic representation processing of data in the fusion data set, using the processed data as a training data set for model training to obtain a prediction model; a model fitting unit for clustering analysis and fitting of the prediction model to form multiple sub-models for different weather conditions; a resource prediction unit for model matching and resource prediction according to prediction requirements and the sub-models; an update unit for operation rule migration of the thermal storage system according to the resource prediction result, and updating and saving the fusion data set and the sub-models according to the operation rule migration result.

10. An electronic device, comprising: It comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete communication with each other through the communication bus; The memory stores a computer program; The processor is used to execute the program stored in the memory to realize the law phase direct radiation trend prediction and thermal storage system coordinated operation method of any one of claims 1-8.

11. A computer readable storage medium storing a computer program, wherein the computer program comprises program instructions configured to cause a processor to perform the method according to any one of claims 1 to 10. The computer program is executed to perform the law phase direct radiation trend prediction and thermal storage system coordinated operation method of any one of claims 1-8.

Images