Multi-energy complementary power generation intelligent optimization scheduling method, system, equipment, medium and product

Abstract

The invention discloses a multi-energy complementary power generation intelligent optimization scheduling method, system and device, a medium and a product, and relates to the field of novel power systems, and the method comprises the steps: obtaining historical sunlight parameters, historical photoelectric output data and historical wind power output data from meteorological data; based on historical sunlight parameters, time sequence data changes corresponding to historical photoelectric output data and historical wind power output data are determined according to the 2D modeling time sequence network model, and wind-light-electricity output data in a short period are predicted; the wind-light-electricity output data comprises photoelectric output power and wind-electricity output power; determining an optimal scheduling scheme based on a deep reinforcement learning model; the deep reinforcement learning model is obtained by carrying out interactive training on the wind-light-electricity output data and a reinforcement learning environment based on a deep convolutional neural network model, the wind-light-electricity output power in a short period can be accurately predicted, and wind power, photoelectricity and hydropower high-precision time dimension intelligent multi-energy complementary optimization scheduling is realized.

Description

Technical Field

[0001] The present application relates to the field of new power systems, and in particular to a method, system, equipment, medium and product for intelligent optimization and scheduling of multi-energy complementary power generation. Background Art

[0002] The new power system is mainly reflected in the planning and construction of the power grid, equipment management, dispatching and operation, and marketing, mainly including auxiliary analysis of the power grid dispatching and operation. At present, the development of new quality productivity is listed as the top priority. The construction of the new power system is inseparable from the new quality productivity. The new quality productivity also needs the new power system as an important carrier. The construction of the new power system is very important for energy transformation. At the same time, the construction of the new power system also faces many challenges. The correlation between the new power system and the weather is an important part of many challenges.

[0003] The current climate is changeable, and the regularity of historical climate data is decreasing. The combination of historical meteorological data and short-term meteorological data is particularly important. In existing methods, historical wind and solar power output and sensor data are mainly used to predict the output power of wind and solar points in the future. Due to the strong correlation between the two data, the data dimension is small, and the cumulative error is large, it is impossible to adaptively predict future power in the face of climate mutations, and high-precision short-term predictions are inaccurate; and the use of traditional methods to predict wind and solar power output or deep learning methods only uses historical data, and cannot use recently collected data to make decisions, resulting in low prediction accuracy.

[0004] It can be seen that the existing methods cannot make high-precision predictions and adaptations to sudden weather conditions due to the use of historical data, resulting in the inability to accurately predict short-term photovoltaic output power and wind power output power, and the inability to achieve high-precision multi-energy complementarity of wind power, photovoltaic power and hydropower. Summary of the invention

[0005] The purpose of this application is to provide a method, system, equipment, medium and product for intelligent optimization scheduling of multi-energy complementary power generation, so as to solve the problem that short-term photovoltaic output power and wind power output power cannot be accurately predicted, and high-precision multi-energy complementarity of wind power, photovoltaic power and hydropower cannot be achieved.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a multi-energy complementary power generation intelligent optimization scheduling method, comprising:

[0008] Obtaining historical solar parameters, historical photovoltaic output data, and historical wind power output data from meteorological data; the meteorological data is calculated by a mesoscale meteorological model;

[0009] Based on the historical sunlight parameters, the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data are determined according to the 2D modeling time series network model, and the wind photovoltaic output data within a short period is predicted; the wind photovoltaic output data includes photovoltaic output power and wind power output power;

[0010] Based on a deep reinforcement learning model, an optimal scheduling plan is determined; the deep reinforcement learning model is based on a deep convolutional neural network model, and is obtained by interactively training the wind and solar power output data with a reinforcement learning environment; the reinforcement learning environment includes the wind and solar power output data, real-time electricity demand, hydropower generation and hydropower constraints; the optimal scheduling plan is an optimal hydropower regulation instruction.

[0011] In a second aspect, the present application provides a multi-energy complementary power generation intelligent optimization scheduling system, comprising:

[0012] A meteorological data acquisition module is used to obtain historical solar parameters, historical photovoltaic output data, and historical wind power output data from meteorological data; the meteorological data is calculated by a mesoscale meteorological model;

[0013] A wind and photovoltaic power output data determination module is used to determine the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data based on the historical sunlight parameters and the 2D modeling time series network model, and predict the wind and photovoltaic power output data in a short period; the wind and photovoltaic power output data includes photovoltaic output power and wind power output power;

[0014] An interface of a hydropower station dispatching system, used to read hydropower generator set data; the hydropower generator set data includes hydropower generation and hydropower constraint conditions;

[0015] The optimal scheduling scheme determination module is used to interact the wind and solar power output data with the reinforcement learning environment, train the deep reinforcement learning model, generate minute-level hydropower regulation instructions, and optimize the deep reinforcement learning model based on the hydropower regulation instructions to generate the optimal scheduling scheme.

[0016] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any one of the multi-energy complementary power generation intelligent optimization scheduling methods described above.

[0017] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned multi-energy complementary power generation intelligent optimization scheduling methods.

[0018] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the multi-energy complementary power generation intelligent optimization scheduling methods described above.

[0019] According to the specific embodiments provided by the present application, the present application discloses the following technical effects: the present application introduces meteorological data, and obtains historical solar parameters and historical photovoltaic output data from the meteorological data, and can also obtain historical solar parameters and historical photovoltaic output data when the climate changes suddenly, so as to accurately predict short-term wind power and photovoltaic output power, and determine the constraints between hydropower generation and hydropower constraints based on actual power demand. Based on the reinforcement learning environment, the deep reinforcement learning model is trained to determine the optimal scheduling strategy combining wind power, photovoltaic power and hydropower, so as to realize high-precision time dimension intelligent multi-energy complementary optimization scheduling of wind power, photovoltaic power and hydropower. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 This is a flow chart of a multi-energy complementary power generation intelligent optimization scheduling method in one embodiment of the present application;

[0022] Figure 2 A specific humidity contour map provided in one embodiment of the present application;

[0023] Figure 3 A wind speed contour map provided in an embodiment of the present application;

[0024] Figure 4 An isotherm diagram provided for an embodiment of the present application;

[0025] Figure 5 A photovoltaic power distribution diagram for a certain day provided in an embodiment of the present application;

[0026] Figure 6 Meteorological data and photovoltaic power distribution diagrams for different time periods provided in an embodiment of the present application;

[0027] Figure 7 Meteorological data and wind power distribution diagrams for different time periods provided in an embodiment of the present application;

[0028] Figure 8 This is an application flow chart of the multi-energy complementary power generation intelligent optimization scheduling system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0030] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0031] like Figure 1 As shown, the multi-energy complementary power generation intelligent optimization scheduling method provided in the embodiment of the present application includes:

[0032] Step 101: Obtain historical solar parameters, historical photovoltaic output data, and historical wind power output data from meteorological data; the meteorological data is calculated by a mesoscale meteorological model.

[0033] Step 102: Based on the historical sunlight parameters, determine the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data according to the 2D modeling time series network model, and predict the wind photovoltaic output data in a short period; the wind photovoltaic output data includes photovoltaic output power and wind power output power. The short period is usually within 7 days.

[0034] Step 103: Determine the optimal scheduling plan based on the deep reinforcement learning model; the deep reinforcement learning model is based on a deep convolutional neural network model, and is obtained by interactively training the wind and solar power output data with a reinforcement learning environment; the reinforcement learning environment includes the wind and solar power output data, real-time electricity demand, hydropower generation and hydropower constraints; the optimal scheduling plan is the optimal hydropower regulation instruction.

[0035] In an exemplary embodiment, the meteorological data is as follows: Figure 2-Figure 4 As shown in the figure, Longitude is longitude, Latitude is latitude, specific humidity is specific humidity, wind speed is wind speed, and tenperature is temperature. By obtaining the changes of the two time series data in the historical time period, the wind and solar power output data in the future short period is predicted. The relationship between meteorological data and wind and solar power output data is shown in the figure. Figure 5-Figure 7 shown.

[0036] Furthermore, meteorological data are calculated through a mesoscale meteorological model. Taking the Dadu River Basin as an example, the specific calculation process of meteorological data is explained.

[0037] The topography and weather and climate conditions in the Dadu River Basin are very complex. The total vertical drop from the source to the estuary is about 3,600 meters, and the overall terrain is high in the northwest and low in the southeast. From the perspective of terrain elevation, the upper reaches of Shuangjiangkou are hilly plateau landforms, Shuangjiangkou to Tongjiezi is a high mountain canyon area, and the lower reaches of Tongjiezi are hilly areas in the Sichuan Basin. From the perspective of precipitation spatial distribution characteristics, the upper reaches of Luding belong to the plateau climate zone. Due to the high terrain, it is difficult to transport water vapor up, there are few rainstorms, and it is cold and dry; the lower reaches of Luding belong to the subtropical humid climate zone, with abundant water vapor and frequent rainstorms. From the perspective of precipitation time distribution, precipitation in the Dadu River Basin mainly occurs from May to October, and rainstorms generally occur from June to September, mostly in July, and there are many night rains in the river valley. Rainstorms are usually caused by the appropriate weather system configuration conditions, superimposed with abundant water vapor conditions and strong upward movement, while night rains are local convective precipitation caused by the temperature difference between day and night between mountain valleys.

[0038] The weather systems that affect the Dadu River Basin include the South Branch Trough, the Westerly Jet Stream, the Southwest Low Vortex, the Plateau Shear Line, the Plateau Vortex, the Western Pacific Subtropical High Pressure, the Easterly Jet Stream, the Ural Blocking High Pressure and the Cut-off Low Pressure. Every year when the Indian monsoon and the East Asian monsoon break out, the warm and humid air from the Indian Ocean and the South China Sea will move northward with the air flow. When the weak cold air from the small ridge in the westerly belt or the strong cold air to the east of the Ural Blocking High Pressure meet these warm and humid air in the Dadu River Basin, it will bring precipitation of varying degrees.

[0039] Based on the topography and weather and climate characteristics above the Dadu River Basin, this application intends to use three layers of bidirectional nested grids, with grid spacings of 13.5 km, 4.5 km and 1.5 km respectively, with the outermost grid resolution of 215*191, the second inner grid resolution of 351*321, and the innermost grid resolution of 501*501. The innermost grid range is 750*750 km, which completely covers the Dadu River Basin, and the outermost range can basically be used to capture the weather system that affects the Dadu River Basin.

[0040] Since the development of mesoscale meteorological modeling, different models have different dynamic frameworks and physical process processing methods. Each model has different characteristics but also uncertainties. The mainstream technology of today's weather forecast is ensemble forecasting, which is to overcome the uncertainty of model forecasting by combining different forecast results. This application intends to combine the three more mature mesoscale meteorological models WRF, MM5 and ARPS in the world today for multi-model ensemble forecasting. The initial and boundary conditions are the global reanalysis data of the National Center for Environmental Prediction (NCEP) of the United States. It aims to overcome the uncertainty of a single model in weather forecasting in the Dadu River Basin, which is conducive to improving the forecast accuracy of weather forecasting in the Dadu River Basin.

[0041] In terms of the selection of the model's microphysical scheme, long- and short-wave radiation scheme, land surface process scheme, planetary boundary layer scheme, and cumulus convection parameterization scheme, this application will select a suitable scheme based on the spatiotemporal distribution characteristics of precipitation in the Dadu River Basin.

[0042] Weather forecasting under complex terrain and different weather and climate backgrounds is a major challenge facing the world today. This application will be based on the existing observation data in the Dadu River Basin, combined with the terrain altitude, slope, slope aspect and other information displayed by the Geographic Information System (GIS), and intends to use the empirical mathematical model of mountain precipitation to further revise the precipitation forecast results of the multi-model ensemble, so as to further improve the accuracy of precipitation forecast in the Dadu River Basin.

[0043] Based on the mesoscale model numerical weather forecast, this application will use advanced physical and statistical methods to establish wind farm and solar farm power prediction models based on wind farm and solar farm operation data, meteorological station monitoring data and high-precision numerical weather forecast data, and predict the power of wind farms and solar farms, thereby providing reference and basis for the Dadu River Basin Hydropower Development Co., Ltd. to formulate power generation plans.

[0044] The Dadu River Basin numerical weather forecast system plans to use APACHE+PHP technology to connect the National Energy Dadu River Basin Hydropower Development Co., Ltd. (WINDOWS) and the developer (LINUX). The system includes the following 5 subsystems:

[0045] 1. Software operation support system.

[0046] The support system integrates various compilers, drawing software and supporting graphics mathematical calculation libraries required by various numerical forecasting models into a unified system, mainly including the following software:

[0047] (1) Operating system (SL6.3 / FD17 / CentOS6.3 / RH6.3).

[0048] (2) Compiler PGI-11.7 / GNU / INTEL2011.

[0049] (3)NETCDF4.1.3 network data format library.

[0050] (4) MPICH2-1.4.1p1 / MVAPICH parallel software.

[0051] (5) NCL6.0 / GRADS drawing software / Vis5D animation demonstration.

[0052] (6) HDF-4.2.6 / jasper / libpng / zlib library (decoding GRIB2).

[0053] (7) BLAS / BUFRLIB / lapack libraries (matrix-vector computation).

[0054] (8) APACHE+PHP+MYSQL web page support software.

[0055] (9)GIS analysis software.

[0056] 2. Real-time data collection system.

[0057] (1) Global-scale numerical model forecast data reception. The initial and boundary conditions of the model are the NCEP global reanalysis data, which are automatically downloaded every day.

[0058] (2) Operation data of wind farms and solar farms of Dadu River Basin Hydropower Development Co., Ltd. and observation data of its own meteorological observation stations.

[0059] (3) Observation data from domestic meteorological departments (data interface reserved).

[0060] 3. Data processing system.

[0061] (1) Data quality control.

[0062] (2) Convert various messages into the format required by the model.

[0063] (3) Use the tools provided by the model to prepare the initial field and boundary conditions.

[0064] 4. Multi-model ensemble numerical forecasting system.

[0065] This system includes three mesoscale numerical prediction models. After the model integration, the system combines them to form the prediction results. The three models are:

[0066] (1)WRF model.

[0067] The WRF model was developed by the National Centers for Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR) in 2000. It is a fully compressible non-hydrostatic model that uses the Arakawa C grid in the horizontal direction and terrain-following mass coordinates in the vertical direction. It is currently the most widely used meteorological research and forecasting model at home and abroad.

[0068] (2)MM5 model.

[0069] The MM5 model is a mesoscale numerical prediction model jointly developed by the National Center for Atmospheric Research (NCAR) and Pennsylvania State University (PSU). It is a mesoscale non-hydrostatic model developed based on the hydrostatic model MM4. It uses the Arakawa B grid in the horizontal direction and a reference air pressure to define the vertical coordinate in the vertical direction. It has been widely used in meteorology, environment, ecology, hydrology and other fields.

[0070] (3)ARPS model.

[0071] The ARPS model was developed by Oklahoma State University in the United States. It is the world's first numerical forecast model specifically for convective-scale weather and is widely used in meteorological research and operational forecasting in many countries around the world.

[0072] 5. Inspection and analysis as well as power prediction systems for wind farms and solar farms.

[0073] (1) The systematic errors of the model element forecasts were calculated using the forecast results of the system model and the actual observations of the stations in the Dadu River Basin. The systematic errors of the forecasts of temperature, wind direction and speed, cloud distribution and other elements were then corrected.

[0074] (2) Combined with the terrain altitude, slope, and aspect information displayed by GIS, the mathematical empirical model of mountain precipitation is used to correct the precipitation forecast by the multi-model ensemble.

[0075] (3) Real-time weather forecast and rating.

[0076] (4) Use the wind farm and solar power plant operation data of the State Energy Dadu River Basin Hydropower Development Co., Ltd., meteorological station monitoring data, and numerical weather forecast data to predict the power of the wind farm and solar power plant of the State Energy Dadu River Basin Hydropower Development Co., Ltd.

[0077] Finally, APACHE+PHP technology is used to combine and unify the above five subsystems to form a complete numerical weather forecast system for the Dadu River Basin and a power forecast system for wind farms and solar power plants, so that the National Energy Dadu River Basin Hydropower Development Co., Ltd. can output forecast results in real time according to actual needs, and can also view the forecast results on any networked computer through IE.

[0078] The scheme for collecting meteorological data has the following characteristics.

[0079] 1. Contains multiple mesoscale numerical weather forecast models (i.e. mesoscale meteorological models).

[0080] This application brings together three relatively mature mesoscale numerical weather forecast models in the world today: WRF, MM5 and ARPS models, effectively overcoming the uncertainty of individual WRF, MM5 and ARPS models, integrating the advantages of multiple models for multi-model aggregate forecasting, which is conducive to improving the accuracy of weather forecasts in the Dadu River Basin.

[0081] 2. Combination of GIS, observational data and mathematical models of mountain precipitation.

[0082] Based on the existing observation data in the Dadu River Basin, combined with the terrain altitude, slope, slope aspect and other information displayed by GIS, the mathematical empirical model of mountain precipitation is used to make in-depth corrections to the precipitation forecast results of the multi-model ensemble.

[0083] 3. Autonomous and controllable.

[0084] With the advent of the big data era, data resources will become an increasingly important strategic resource. The meteorological system has increasingly stringent control over meteorological observation data. Now there is a trend that is becoming increasingly obvious, that is, in the future, the meteorological system will most likely no longer provide original observation data to the outside world, but only provide forecast products. Therefore, if the numerical weather forecast system can be run autonomously on the computers of Dadu River Basin Hydropower Development Co., Ltd. without control and produce the weather forecast products it needs, this will bring great flexibility to the power production and dispatching of Dadu River Basin Hydropower Development Co., Ltd., and will definitely produce huge economic and social benefits.

[0085] 4. Power prediction model for wind farms and solar farms.

[0086] Based on the independently controllable numerical weather forecast data of Guoneng Dadu River Basin Hydropower Development Co., Ltd., combined with the company's own wind farm and solar power plant operation data and meteorological station monitoring data, a prediction model for the power of the wind farm and solar power plant of Dadu River Basin Hydropower Development Co., Ltd. was established.

[0087] 5.Flexible system data interface.

[0088] The system currently uses the global atmospheric reanalysis data of the National Center for Marine and Atmospheric Environment (NCEP) of the United States, which is open to users, as initial and boundary conditions, but at the same time reserves interfaces for accessing observation data from self-built observation stations of my country's meteorological department and Dadu River Basin Hydropower Development Co., Ltd.

[0089] 6. Advanced WEB control technology.

[0090] In the browser, you can use images and tables to display various elements of the forecast and the actual situation. The wind can be displayed using the wind vane method commonly used in meteorology. Users do not need to be familiar with various LINUX shell commands to complete system browsing, parameter modification, forecast effect inspection and other functions.

[0091] In an exemplary embodiment, step 102 may be replaced by steps 201 to 203, which include the following steps.

[0092] Step 201: Determine the light intensity based on historical sunlight parameters obtained from meteorological data; the historical sunlight parameters include sunlight reflectivity and light radiation intensity in a local area.

[0093] Step 202: Based on the light intensity, a 2D modeling time series network model is used to determine the time series data changes corresponding to the historical photovoltaic output data, and the photovoltaic output power in a short period is predicted.

[0094] The photoelectric output power P s for:

[0095] P s =f(E(R,le)×I×A×α)

[0096] Among them, I is the photoelectric efficiency, A is the area of ​​the solar panel, α is the attenuation rate, E(R,le) is the light intensity, R is the sunlight reflectivity in the local area, le is the light radiation intensity in the local area, and f(.) is the 2D modeling time series network model.

[0097] Step 203: using a 2D modeling time series network model to determine the time series data changes corresponding to the historical wind power output data, and predicting the wind power output power in a short period.

[0098] The wind power output power P c for:

[0099] P c =f(NWP)

[0100] Where NWP = [t, h, P air ,v,v dir ,R rain ], NWP is meteorological data, t is temperature, h is humidity, P air is the air pressure, v is the wind speed at the fan hub height, v dir is the wind direction at the fan hub height, R rain is the rainfall, and f(.) is the 2D modeling time series network model.

[0101] In an exemplary embodiment, building a deep reinforcement learning model requires establishing a reinforcement learning environment, defining a state space and an action space, and setting a reward function.

[0102] Among them, the reinforcement learning environment includes predicted wind and solar power output data, real-time electricity demand, hydropower generation and hydropower constraints.

[0103] By obtaining the real-time power demand P total , according to the predicted wind and solar power output data P c and P s , establish hydropower generation capacity P w and the hydropower constraint ε.

[0104] The state space is a set of variables that describe the environmental state, and the set includes wind and solar power output data, short-term electricity demand, and hydropower output data.

[0105] The action space is a set of actions that can be taken by the deep reinforcement learning model, and the set includes parameters and duration for controlling the amount of water and electricity.

[0106] The reward function is mainly used to evaluate the quality of the actions taken by the deep reinforcement learning model at each time. The reward function is as follows:

[0107] R(s,a)=ω1·initial reward+ω2·dispatching hydropower reward

[0108] Among them, s represents the current state, a represents the action taken by the deep reinforcement learning model, and ω1 and ω2 represent the weights of various factors.

[0109] Rewards are set at the beginning of the task to reward those who take immediate action at the beginning of the task. Reward values ​​are set in relation to the start time of the task. Quick adjustment of the parameters for controlling the water volume is rewarded, while delayed adjustment of the water volume is penalized. Rewards for scheduling hydropower are set based on whether the water volume parameters are controlled quickly and accurately. This is used to reward the deep reinforcement learning model for controlling the balance between the water volume parameters and the photovoltaic output. If the hydropower control parameters are given quickly and accurately, the reward is positive, otherwise it is negative.

[0110] In an exemplary embodiment, a deep convolutional neural network model is used as the network model body of the deep reinforcement learning model, and wind and solar power output data are used to interact with the reinforcement environment for training, and the scheduling plan is continuously optimized through a reward function. Step 103 can be replaced by steps 301-302, including the following steps.

[0111] Step 301: define the state space and action space of the deep convolutional neural network model, and set a reward function; the state space is a set of variables that describe the environmental state, and the variable set includes wind and solar power output data, electricity demand in a short period, and hydropower output data; the action space is a set of actions taken by the deep convolutional neural network model, and the action set includes parameters and duration for controlling the amount of hydropower.

[0112] Step 302: Based on the state space and the action space, the wind and photovoltaic power output data are interactively trained with the reinforcement learning environment, and the wind power, photovoltaic and hydropower scheduling plans output by the deep reinforcement learning model are continuously optimized through the reward function to determine the optimal scheduling plan.

[0113] In an exemplary embodiment, building a deep convolutional neural network model also requires setting network parameters, which include the number of network layers, the number of neurons, and the activation function. The structure of the deep convolutional neural network model is Q, and the expression is:

[0114] Q(s,a,θ)

[0115] Among them, Q represents the output of the deep convolutional neural network model, and θ represents the parameters of the deep convolutional neural network model.

[0116] The state space expression is:

[0117] s i =(s1,s2,…,s n )

[0118] s n is the nth state space.

[0119] The action space expression is:

[0120] a i =(a1,a2,…,a n )

[0121] s i represents the i-th state variable in the state space, s n represents the nth state variable in the state space, a i represents the i-th action variable in the action space, a n Represents the nth action variable in the action space, i=1,2,3,...,n.

[0122] The deep reinforcement learning model obtains the optimal actions under different states, and the minimum error value enables the deep reinforcement learning model to make the optimal decision at each stage and obtain the optimal scheduling strategy.

[0123] This application uses actual data to evaluate the deep reinforcement learning model generated by test training, applies the reinforcement learning scheduling model to actual data, and evaluates the model effect. It can be clearly concluded that the multi-energy complementary power generation intelligent optimization scheduling method of this application can achieve high-speed and intelligent multi-energy complementarity of water, photovoltaic and electricity, creating value for practical applications.

[0124] Based on the same inventive concept, the embodiment of the present application also provides a multi-energy complementary power generation intelligent optimization scheduling system for implementing the multi-energy complementary power generation intelligent optimization scheduling method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more multi-energy complementary power generation intelligent optimization scheduling system embodiments provided below can refer to the limitations of the multi-energy complementary power generation intelligent optimization scheduling method above, and will not be repeated here.

[0125] In an exemplary embodiment, a multi-energy complementary power generation intelligent optimization scheduling system is provided, comprising:

[0126] The meteorological data acquisition module is used to obtain historical solar parameters, historical photovoltaic output data and historical wind power output data from meteorological data; the meteorological data is calculated by a mesoscale meteorological model.

[0127] The wind and photovoltaic power output data determination module is used to determine the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data based on the historical sunlight parameters and the 2D modeling time series network model, and predict the wind and photovoltaic power output data in a short period; the wind and photovoltaic power output data includes photovoltaic output power and wind power output power.

[0128] The interface of the hydropower station dispatching system is used to read the data of the hydropower generator set; the hydropower generator set data includes the hydropower generation and the hydropower constraint conditions.

[0129] The optimal scheduling scheme determination module is used to interact the wind and solar power output data with the reinforcement learning environment, train the deep reinforcement learning model, generate minute-level hydropower regulation instructions, and optimize the deep reinforcement learning model based on the hydropower regulation instructions to generate the optimal scheduling scheme.

[0130] The deep reinforcement learning model that has achieved the effect is integrated into the wind, solar and hydropower station dispatching system, the meteorological data collection module and the wind and solar power output data determination module to predict the wind and solar power output data in a short period, and read the hydropower generator data through the interface of the hydropower station dispatching system to generate minute-level hydropower regulation instructions, intelligently adjust the power generation of the hydropower generator set, and achieve constant or smooth unified power output. According to the hydropower regulation instructions, adjust the parameters of the deep reinforcement learning model, and optimize the deep reinforcement learning model to provide better dispatching instructions. The application process of the multi-energy complementary power generation intelligent optimization dispatching system is as follows: Figure 8 shown.

[0131] In an exemplary embodiment, the trained deep reinforcement learning model can obtain the real-time output data of actual power generation from the external query interface of the wind power plant and photovoltaic power plant system. Among them, reinforcement learning is to reward and punish and adjust the deep reinforcement learning model itself by comparing the difference between the predicted output data and the actual output data.

[0132] The trained deep reinforcement learning model obtains the power output power target instructions issued by the superior department through the power dispatching interface.

[0133] The trained deep reinforcement learning model is used to calculate high-precision data on the required hydropower output power, and through the hydropower station power dispatching system interface, the hydropower generator adjustment instructions are output in real time to the hydropower station dispatching system for dispatching.

[0134] During the dispatching process, the deep reinforcement learning model is adjusted and optimized based on real-time meteorological data, power output target instructions issued by superior departments, and wind power, photovoltaic and hydropower output power data, so as to continuously and intelligently optimize the entire multi-energy complementary power generation intelligent optimization dispatching method and multi-energy complementary power generation intelligent optimization dispatching system.

[0135] Compared with the traditional multi-energy complementary scheduling method, the multi-energy complementary power generation intelligent optimization scheduling method based on reinforcement learning in this application introduces a large amount of meteorological data and predicts short-term wind and solar power output with higher accuracy.

[0136] Since this method refers to meteorological data, it can more accurately predict the wind and solar power output in the short term in the future. Compared with the common use of historical data to predict future output data, it has higher sensitivity and can predict short-term output data, thereby controlling hydropower in a short period of time.

[0137] This application integrates the characteristics of short time and high precision, and can more intelligently adjust the hydropower output automatically according to weather changes, solving the problem of abandoned photovoltaic and wind power.

[0138] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store multi-energy complementary power generation intelligent optimization scheduling data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a multi-energy complementary power generation intelligent optimization scheduling method is implemented.

[0139] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0140] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0141] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0142] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0143] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0144] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0145] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A multi-energy complementary power generation intelligent optimization scheduling method, characterized in that: The multi-energy complementary power generation intelligent optimization scheduling method comprises: Obtaining historical solar parameters, historical photovoltaic output data, and historical wind power output data from meteorological data; the meteorological data is calculated by a mesoscale meteorological model; Based on the historical sunlight parameters, the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data are determined according to the 2D modeling time series network model, and the wind photovoltaic output data within a short period is predicted; the wind photovoltaic output data includes photovoltaic output power and wind power output power; Based on a deep reinforcement learning model, an optimal scheduling plan is determined; the deep reinforcement learning model is based on a deep convolutional neural network model, and is obtained by interactively training the wind and solar power output data with a reinforcement learning environment; the reinforcement learning environment includes the wind and solar power output data, real-time electricity demand, hydropower generation and hydropower constraints; the optimal scheduling plan is an optimal hydropower regulation instruction.

2. The multi-energy complementary power generation intelligent optimization scheduling method according to claim 1 is characterized in that: Based on the historical sunlight parameters, the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data are determined according to the 2D modeling time series network model, specifically including: Determine the light intensity based on historical sunlight parameters obtained from meteorological data; the historical sunlight parameters include sunlight reflectivity and light radiation intensity in a local area; Based on the light intensity, a 2D modeling time series network model is used to determine the time series data changes corresponding to the historical photovoltaic output data, and to predict the photovoltaic output power in a short period; The 2D modeling time series network model is used to determine the time series data changes corresponding to the historical wind power output data, and predict the wind power output power in a short period.

3. The multi-energy complementary power generation intelligent optimization scheduling method according to claim 2 is characterized in that: The photoelectric output power P s for: P s =f(E(R,le)×I×A×α) Among them, I is the photoelectric efficiency, A is the area of ​​the solar panel, α is the attenuation rate, E(R,le) is the light intensity, R is the sunlight reflectivity in the local area, le is the light radiation intensity in the local area, and f(.) is the 2D modeling time series network model.

4. The multi-energy complementary power generation intelligent optimization scheduling method according to claim 2 is characterized in that: The wind power output power P c for: P c =f(NWP) Where NWP = [t, h, P air ,v,v dir ,R rain ], NWP is meteorological data, t is temperature, h is humidity, P air is the air pressure, v is the wind speed at the fan hub height, v dir is the wind direction at the fan hub height, R rain is the rainfall, and f(.) is the 2D modeling time series network model.

5. The multi-energy complementary power generation intelligent optimization scheduling method according to claim 1 is characterized in that: Based on the deep reinforcement learning model, the optimal scheduling solution is determined, including: Define the state space and action space of the deep convolutional neural network model, and set a reward function; the state space is a set of variables describing the environmental state, and the variable set includes wind and solar power output data, electricity demand in a short period, and hydropower output data; the action space is a set of actions taken by the deep convolutional neural network model, and the action set includes parameters and duration for controlling the amount of hydropower; Based on the state space and the action space, the wind and photovoltaic power output data are interactively trained with the reinforcement learning environment, and the dispatching schemes of wind power, photovoltaic power and hydropower output by the deep reinforcement learning model are continuously optimized through the reward function to determine the optimal dispatching scheme.

6. The multi-energy complementary power generation intelligent optimization scheduling method according to claim 5 is characterized in that: The reward function is R(s,α): R(s,a)=ω1·initial reward+ω2·dispatching hydropower reward Among them, s is the current state, a is the action, ω1 and ω2 are the weights of each factor.

7. A multi-energy complementary power generation intelligent optimization dispatching system, characterized in that: The multi-energy complementary power generation intelligent optimization scheduling system is based on the multi-energy complementary power generation intelligent optimization scheduling method according to any one of claims 1 to 6, and the multi-energy complementary power generation intelligent optimization scheduling system comprises: A meteorological data acquisition module is used to obtain historical solar parameters, historical photovoltaic output data, and historical wind power output data from meteorological data; the meteorological data is calculated by a mesoscale meteorological model; A wind and photovoltaic power output data determination module is used to determine the time series data changes corresponding to the historical photovoltaic output data and the historical wind power output data based on the historical sunlight parameters and the 2D modeling time series network model, and predict the wind and photovoltaic power output data in a short period; the wind and photovoltaic power output data includes photovoltaic output power and wind power output power; An interface of a hydropower station dispatching system, used to read hydropower generator set data; the hydropower generator set data includes hydropower generation and hydropower constraint conditions; The optimal scheduling scheme determination module is used to interact the wind and solar power output data with the reinforcement learning environment, train the deep reinforcement learning model, generate minute-level hydropower regulation instructions, and optimize the deep reinforcement learning model based on the hydropower regulation instructions to generate the optimal scheduling scheme.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-energy complementary power generation intelligent optimization scheduling method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the multi-energy complementary power generation intelligent optimization scheduling method described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the multi-energy complementary power generation intelligent optimization scheduling method described in any one of claims 1 to 6 is implemented.

Images