Generating capacity prediction method for wind-solar hybrid power generation system

Abstract

The invention is suitable for the technical field of wind-solar power generation, and particularly relates to a wind-solar complementary power generation system generating capacity prediction method, which comprises the following steps: carrying out data acquisition to obtain historical data; preprocessing the historical data to obtain preprocessed data, and discretizing the preprocessed data to obtain discrete data; constructing an HMMC model, and constructing a historical data set through historical data to train the HMMC model; and collecting real-time data, calculating a thermal index based on the real-time data, and performing power prediction by the HMMC model according to the thermal index. By considering the interdependence between photovoltaic power generation and wind power generation and taking the thermal index as a meteorological variable into the model, the prediction accuracy of PV and wind power generation is remarkably improved, model training is performed by using historical data, and the model is dynamically updated to adapt to a new data sample, so that the prediction accuracy of the photovoltaic power generation and the wind power generation is improved. And the prediction accuracy and timeliness are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind-solar power generation, and in particular relates to a method for predicting power generation of a wind-solar complementary power generation system. Background Art

[0002] The wind-solar hybrid power generation system is a comprehensive power generation system that combines two renewable energy technologies, wind power generation and solar photovoltaic power generation. The system captures wind energy and converts it into electricity through wind turbines, and uses photovoltaic panels to convert solar energy into electricity, thereby maximizing energy acquisition. Due to the changes in wind and light conditions throughout the day and between different seasons, this complementary approach can balance the fluctuations in output power to a certain extent, improve the stability and reliability of the system, and is suitable for a variety of scenarios such as power supply in remote areas, public facilities, and residential electricity use, which helps promote the application and development of clean energy.

[0003] The main drawback of the existing technology (HMC) is that it only captures the temporal dynamics of photovoltaic (PV) / wind power generation, while ignoring other interdependencies. HMC is achieved by constructing two separate high-order Markov chains to predict PV output and wind power generation respectively, which means that this method only considers the time series dynamics of PV / wind power generation, but does not consider the interrelationship between PV output power and wind output power, or the dependence with other related variables (such as thermal index). The limitation of this method is that it may not fully capture all the key factors that affect the accuracy of PV and wind power generation forecasts, thereby limiting the overall performance and accuracy of the forecasting model. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting power generation of a wind-solar hybrid power generation system, aiming to solve the problem that all key factors affecting the accuracy of PV and wind power generation prediction cannot be fully captured, thereby limiting the overall performance and accuracy of the prediction model.

[0005] The present invention is implemented as follows: a method for predicting power generation of a wind-solar hybrid power generation system, the method comprising:

[0006] Performing data collection to obtain historical data, wherein the historical data at least includes photovoltaic power generation data, wind power generation data and meteorological data;

[0007] Preprocess the historical data to obtain preprocessed data, and discretize the preprocessed data to obtain discrete data;

[0008] Build the HMMC model and train the HMMC model by building a historical data set based on historical data;

[0009] Collect real-time data, calculate the thermal index based on the real-time data, and use the HMMC model to predict power based on the thermal index.

[0010] Preferably, in the step of preprocessing the historical data to obtain preprocessed data, the preprocessing process includes data clarification, missing value processing and data normalization processing on the historical data.

[0011] Preferably, in the step of discretizing the preprocessed data to obtain discrete data, the continuous historical data is converted into discrete data for HMMC model processing, and the process is: discretizing the historical data into several state intervals, and making the residence time of each state equal by solving the optimal time interval.

[0012] Preferably, the formula of the HMMC model is:

[0013]

[0014] In the formula, is the state probability distribution of the rrr-th data sequence at time t+1, k is the time lag length, indicating the number of historical time steps considered by the model, is the lag coefficient, which indicates the influence weight of data sequence j on sequence r at time lag i. is the state transition probability matrix, which represents the probability of transition from the state of the jth sequence at time ti to the state of the rth sequence at time t+1t+1t+1. is the state probability distribution of the rth sequence at time t+1-i.

[0015] Preferably, in the step of performing power prediction based on the thermal index using the HMMC model, the prediction formula is:

[0016]

[0017] In the formula, is the value of the rth data sequence predicted at time t+1, r represents photovoltaic power generation, wind power generation or thermal index, k is the time window length of historical observations, and the number of past moments used for prediction, is the lag coefficient, which indicates the influence weight of time lag i and data sequence j on the rth data sequence. This coefficient is used to adjust the contribution of each time point and sequence to the prediction result. This is the state transition probability matrix The c (r) Column, where c (r) is the current state of the rth data sequence, describes the transition probability from the state of the jth sequence at time ti to the state of the rth sequence at time t+1, Represents the vector of all possible states of the rth data sequence, N (r) is the number of discrete states of the rth data sequence. This vector represents the representative value of the sequence in different states, c (r) It is the current state number of the r-th data sequence, indicating the discrete state of the r-th sequence at time t.

[0018] Preferably, the meteorological data includes temperature, relative humidity and heat index.

[0019] Preferably, the calculation formula of the thermal index is:

[0020]

[0021] Where HI is the heat index, which represents the combined effect of ambient temperature and relative humidity on human perception and is a meteorological variable used to enhance prediction accuracy. ij It is a coefficient determined experimentally to describe the effect of different combinations of temperature T and relative humidity RH on the heat index. i is the ambient temperature in degrees Celsius, an input to the model

[0022] The present invention provides a method for predicting power generation of a wind-solar complementary power generation system, which can improve the prediction accuracy: by considering the interdependence between photovoltaic and wind power generation, and incorporating the thermal index as a meteorological variable into the model, the present invention significantly improves the prediction accuracy of PV and wind power generation, uses historical data for model training, and dynamically updates the model to adapt to new data samples, ensuring the accuracy and timeliness of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of a method for predicting power generation of a wind-solar hybrid power generation system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0025] like Figure 1 FIG. 1 is a flow chart of a method for predicting power generation of a wind-solar hybrid power generation system provided by an embodiment of the present invention, the method comprising:

[0026] Data collection is performed to obtain historical data, wherein the historical data at least includes photovoltaic power generation data, wind power generation data and meteorological data.

[0027] In this step, data collection is performed on the server side, including collecting historical power data, ambient temperature and relative humidity from photovoltaic power plants and wind power plants; Data source: Use historical photovoltaic and wind power generation data. In the case study, the data comes from the actual site, including: wind power generation data (unit: kW), photovoltaic power generation data (unit: kW), meteorological variables such as temperature (T), relative humidity (RH) and heat index (HI), which are collected at intervals of 15 minutes for short-term power forecasting; Meteorological data: The key meteorological variable is the heat index (HI), which is calculated from temperature and relative humidity and used as an additional variable to enhance the accuracy of power forecasting. The calculation formula of the heat index is:

[0028]

[0029] Where HI is the heat index, which represents the combined effect of ambient temperature and relative humidity on human perception and is a meteorological variable used to enhance prediction accuracy. ij It is a coefficient determined experimentally to describe the effect of different combinations of temperature T and relative humidity RH on the heat index. i is the ambient temperature in degrees Celsius, an input variable of the model, RH j is the relative humidity in percentage.

[0030] The historical data is preprocessed to obtain preprocessed data, and the preprocessed data is discretized to obtain discrete data.

[0031] In this step, data preprocessing is performed on the server side, involving operations such as cleaning data, processing missing values, and data normalization. Data discretization is performed on the server side, converting continuous data into discrete data suitable for high-order multivariate Markov chain (HMMC) processing. The present invention adopts a 15-minute prediction interval, and the continuous data sequence must be discretized into a finite state model. The purpose of discretization is to ensure that each state can evenly represent the changes in data. To this end, the data is discretized using the following steps: discretizing wind power generation, photovoltaic power generation, and thermal index data into several state intervals. By solving the optimal time interval δ(r)\delta^{(r)}δ(r), the residence time of each state is equal.

[0032] Build an HMMC model and train the HMMC model by building a historical data set using historical data.

[0033] In this step, model training is performed on the server side, and the HMMC model is trained using historical data sets, including setting state transition probabilities, calculating λ and Q parameters, etc. The heat index (HI) calculation is performed on the server side, and the HI value is calculated based on the collected temperature and humidity data. The present invention uses a high-order multivariable Markov chain (HMMC) to predict the power of wind and photovoltaic power generation. The model takes into account the intrinsic relationship between photovoltaic, wind power generation and meteorological variables to improve the prediction accuracy. The model assumes that the state probability distribution of the current time series depends on the state distribution at the previous moment. The formula is as follows:

[0034]

[0035] In the formula, is the state probability distribution of the rrr-th data sequence at time t+1, k is the time lag length, indicating the number of historical time steps considered by the model, is the lag coefficient, which indicates the influence weight of data sequence j on sequence r at time lag i. is the state transition probability matrix, which represents the probability of transition from the state of the jth sequence at time ti to the state of the rth sequence at time t+1t+1t+1. is the state probability distribution of the rth sequence at time t+1-i. Optimization parameters: The key parameters λ and Q in the model are estimated through linear programming optimization. The goal of this optimization problem is to minimize the difference between the model's predicted distribution and the long-term pattern of the actual data.

[0036] Collect real-time data, calculate the thermal index based on the real-time data, and use the HMMC model to predict power based on the thermal index.

[0037] In this step, the model performs a 15-minute short-term power forecast, and the forecast formula is as follows:

[0038]

[0039] In the formula, is the value of the rth data sequence predicted at time t+1, r represents photovoltaic power generation, wind power generation or thermal index, k is the time window length of historical observations, and the number of past moments used for prediction, is the lag coefficient, which indicates the influence weight of time lag i and data sequence j on the rth data sequence. This coefficient is used to adjust the contribution of each time point and sequence to the prediction result. This is the state transition probability matrix The c (r) Column, where c (r) is the current state of the rth data sequence, describes the transition probability from the state of the jth sequence at time ti to the state of the rth sequence at time t+1, Represents the vector of all possible states of the rth data sequence, N (r) is the number of discrete states of the rth data sequence. This vector represents the representative value of the sequence in different states, c (r) It is the current state number of the rth data sequence, indicating the discrete state of the rth sequence at time t. Dynamic update: The present invention proposes a mechanism for dynamically updating model parameters, that is, updating once every 4 hours to adapt to new measurement data, thereby improving the accuracy of prediction.

[0040] The present invention has the following characteristics:

[0041] 1. Improve forecast accuracy: By considering the interdependence between photovoltaic (PV) and wind power generation, and incorporating the heat index (HI) as a meteorological variable into the model, the present invention significantly improves the forecast accuracy of PV and wind power generation. The model is trained using historical data and dynamically updated to adapt to new data samples to ensure the accuracy and timeliness of the forecast.

[0042] 2. Flexible input vector expansion: This technical solution allows to increase the input vector dimension of the prediction algorithm in a flexible way, thereby improving the performance of the prediction algorithm. This is very important in practical applications because it allows to dynamically adjust the model according to the available data type and quantity to achieve the best performance.

[0043] 3. Adaptability and dynamic updating of the model: Compared with traditional prediction methods, the method proposed in this invention has better adaptability. Even when historical data is incomplete, it can capture the changes in PV and wind power generation by dynamically updating model parameters, thereby reducing the prediction error to a certain extent.

[0044] 4. Applicable to incomplete historical data sets: The present invention also demonstrates the effectiveness of its technical solution in processing incomplete historical data sets. Even when some data samples are missing, the proposed method can effectively predict the changes in PV and wind power generation, which is extremely important in practical applications.

[0045] In summary, the technical solution of the present invention not only improves the accuracy of photovoltaic and wind power generation predictions, but also provides flexible expansion of model input vectors, dynamic updating of model parameters, and effectiveness in processing incomplete data sets, thereby enabling it to provide more reliable and accurate prediction results in practical applications.

[0046] The present invention comprises a server and a client:

[0047] The present invention also relates to result feedback: client display, which displays the prediction results to the end user through the user interface.

[0048] Server and client interaction:

[0049] Data collection and preprocessing, model training and prediction are mainly completed on the server side.

[0050] The prediction results are displayed on the client side, and users can view them through the web interface or mobile application.

[0051] Modular design: The system is designed into data preprocessing module, data discretization module, HMMC model training module and prediction module, etc. Each module is responsible for different functions, and modules exchange and communicate data through defined interfaces.

[0052] Multi-layer server-side distribution:

[0053] The present invention may involve a multi-layer architecture including a data processing server, a model training server and an application server.

[0054] The data processing server is responsible for data collection and preprocessing.

[0055] The model training server is used to train the HMMC model.

[0056] The application server processes user requests, calls the prediction model and feeds the results back to the client.

[0057] Software and hardware collaborative workflow:

[0058] The hardware part involves data collection equipment (such as sensors in photovoltaic and wind power plants) used to monitor and record power generation, ambient temperature and humidity in real time.

[0059] The software part includes data processing, model training and prediction algorithms on the server side, and the user interface on the client side.

[0060] The data collected by the hardware is transmitted to the server via the network for processing and analysis. The server generates a prediction report based on the analysis results and then sends it to the client via the network for display.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting power generation of a wind-solar hybrid power generation system, characterized in that: The method comprises: Performing data collection to obtain historical data, wherein the historical data at least includes photovoltaic power generation data, wind power generation data and meteorological data; Preprocess the historical data to obtain preprocessed data, and discretize the preprocessed data to obtain discrete data; Build the HMMC model and train the HMMC model by building a historical data set based on historical data; Collect real-time data, calculate the thermal index based on the real-time data, and use the HMMC model to predict power based on the thermal index.

2. The method for predicting power generation of a wind-solar hybrid power generation system according to claim 1, characterized in that: In the step of preprocessing the historical data to obtain preprocessed data, the preprocessing process includes data clarification, missing value processing and data normalization processing on the historical data.

3. The method for predicting power generation of a wind-solar hybrid power generation system according to claim 1, characterized in that: In the step of discretizing the preprocessed data to obtain discrete data, the continuous historical data is converted into discrete data for HMMC model processing. The process is: discretizing the historical data into several state intervals, and making the residence time of each state equal by solving the optimal time interval.

4. The method for predicting power generation of a wind-solar hybrid power generation system according to claim 1, characterized in that: The formula of the HMMC model is: In the formula, is the state probability distribution of the rrr-th data sequence at time t+1, k is the time lag length, indicating the number of historical time steps considered by the model, is the lag coefficient, which indicates the influence weight of data sequence j on sequence r at time lag i. is the state transition probability matrix, which represents the probability of transition from the state of the jth sequence at time ti to the state of the rth sequence at time t+1t+1t+1. is the state probability distribution of the rth sequence at time t+1-i.

5. The method for predicting power generation of a wind-solar hybrid power generation system according to claim 1, characterized in that: In the step of performing power prediction based on the thermal index using the HMMC model, the prediction formula is: In the formula, is the value of the rth data sequence predicted at time t+1, r represents photovoltaic power generation, wind power generation or thermal index, k is the time window length of historical observations, and the number of past moments used for prediction, is the lag coefficient, which indicates the influence weight of time lag i and data sequence j on the rth data sequence. This coefficient is used to adjust the contribution of each time point and sequence to the prediction result. This is the state transition probability matrix The c (r) Column, where c (r) is the current state of the rth data sequence, describes the transition probability from the state of the jth sequence at time ti to the state of the rth sequence at time t+1, Represents the vector of all possible states of the rth data sequence, N (r) is the number of discrete states of the rth data sequence. This vector represents the representative value of the sequence in different states, c (r) It is the current state number of the r-th data sequence, indicating the discrete state of the r-th sequence at time t.

6. The method for predicting power generation of a wind-solar hybrid power generation system according to claim 1, characterized in that: The meteorological data includes temperature, relative humidity and heat index.

7. The method for predicting power generation of a wind-solar hybrid power generation system according to claim 6, characterized in that: The calculation formula of the thermal index is: Where HI is the heat index, which represents the combined effect of ambient temperature and relative humidity on human perception and is a meteorological variable used to enhance prediction accuracy. ij It is a coefficient determined experimentally to describe the effect of different combinations of temperature T and relative humidity RH on the heat index. i is the ambient temperature in degrees Celsius, an input variable of the model, RH j is the relative humidity in percentage.

Images