Method for predicting output thermal power of tower type photo-thermal mirror field heat absorber
By using a generalized linear mixed quantile regression model and Bayesian theory, combined with a meteorological station and DCS control system for a tower-type solar thermal receiver, the accuracy and interpretability issues of thermal power prediction in existing technologies have been resolved, achieving high-precision thermal power prediction and grid dispatch support.
Patent Information
- Application Number
- CN202511438081.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-10
- Publication Date
- 2026-01-30
AI Technical Summary
Existing linear regression and deep neural network methods for predicting the thermal power of tower solar thermal receivers suffer from limitations in distribution assumptions, complex data requirements, and insufficient interpretability. Furthermore, traditional prediction methods struggle to provide upper and lower limits for receiver thermal power prediction, failing to meet the prediction requirements of power grid dispatch.
A generalized linear mixed quantile regression model was adopted, combined with Bayesian theory and Markov chain Monte Carlo method, to establish a heat receiver thermal power prediction model. By considering multiple random effect pairs such as wind speed and temperature, the generalized linear mixed quantile regression model prediction method was used, and historical data were obtained from the meteorological station and DCS control system of the solar thermal mirror field to make accurate predictions.
It achieves high-precision prediction of the thermal power of tower-type solar thermal absorbers, can handle outliers and anomalies, provides more flexible prediction results, and meets the needs of power grid dispatch.
Smart Images
Figure QLYQS_1
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of tower type light and heat mirror field, and particularly relates to a tower type light and heat mirror field heat absorber output heat power prediction method. BACKGROUND
[0002] Tower type light and heat power generation not only has the advantages of high power generation efficiency, stable output power and high performance heat storage, but also is one of new energy storage modes, can realize continuous 24-hour power generation, and is widely applied in thick oil exploitation, green hydrogen production and cross-season heating.
[0003] The principle of tower type light and heat power generation is that a large amount of low-density solar energy is reflected to the heat absorber to become high-density solar energy through tens of thousands of heliostats arranged around the heat absorber in different geometric shapes, and then the high-density solar energy is converted into working medium heat. The core links of the system include light energy collection, light and heat conversion, heat storage and heat and electricity conversion, each link is affected by multiple factors, and has high time variability and nonlinearity. Among them, accurate prediction of the heat absorber output heat power can reasonably plan the charging and discharging strategy of the heat storage system, maximize the use of the heat storage capacity, and is one of the core technologies to ensure efficient and stable operation of the power station and friendly access to the power grid.
[0004] The core idea of the heat absorber output heat power prediction is to comprehensively analyze the wind speed, DNI, visibility, position of the sun at a certain time, movement state of the heliostat at a certain time, state and historical data of the light and heat collection system, establish a mathematical model from solar radiation to the final output heat energy, and predict the heat function of the power station in advance, so as to provide key decision basis for operation strategy formulation, energy storage scheduling and power grid coordination. The factors affecting the heat power of the heat absorber not only include the rich solar resource condition, the optical efficiency and the operation efficiency of the mirror field, but also the heat loss of the heat absorber itself. Usually, the heat loss of the heat absorber includes reflection loss, convection loss and radiation loss, and the convection loss and the radiation loss are greatly related to the wind speed and the temperature. At the same time, the solar resource is greatly affected by time, and usually has periodicity, for example, the light and heat conversion efficiency is higher on the summer solstice day, and the light and heat conversion efficiency is lower on the winter solstice day, so the heat power prediction model also needs to consider the influence of the wind speed, the temperature and the time period on the heat power of the heat absorber.
[0005] The current commonly used prediction methods include linear regression, deep neural network learning and the like. The deep neural network has great advantages for text or picture data, but has little advantage for general panel data, and the modeling process is relatively complex. Generally, the linear model needs to meet the requirements that the dependent variable conforms to the normal distribution, is independent and has equal variance. However, for the heat absorber of the tower type light and heat type, affected by the unstable variables such as normal solar radiation and wind speed, the assumption conditions of normality and equal variance are often not met.
[0006] Traditional linear regression and deep neural networks both face significant challenges in predicting the thermal power of solar thermal receivers: the former is limited by distribution assumptions and linear constraints, while the latter is constrained by data requirements and interpretability limitations. In contrast, generalized linear mixed regression models not only have no requirements on independent and dependent variables, making them applicable to data of any hierarchical structure, but also consider both fixed and random effects, resulting in higher prediction accuracy. Furthermore, general power prediction typically forecasts the expected power (mean) at a given moment. However, for power grid dispatching, it is more desirable to know the upper and lower limits of power at a given moment, i.e., the quantiles. Power grid dispatching is based on the "worst-case" scenario: when solar energy and wind speed are stable, the confidence interval is small, and peak-shaving capacity is appropriately reduced; when solar energy and wind speed are unstable, the confidence interval is large, and peak-shaving capacity is appropriately increased. Therefore, a method for predicting the thermal power of tower-type solar thermal receivers based on a generalized linear mixed quantile regression model is needed. Summary of the Invention
[0007] To address the aforementioned technical problems, this invention provides a method for predicting the output thermal power of a tower-type photothermal mirror field receiver, achieving higher prediction accuracy and greater robustness against outliers or anomalies, resulting in a more accurate and flexible overall prediction effect for the receiver's thermal power.
[0008] To achieve the above objectives, the technical solution of the present invention is as follows:
[0009] A method for predicting the output thermal power of a tower-type photothermal mirror field receiver includes the following steps:
[0010] Step 1: Establish a generalized linear mixed quantile regression model for predicting the thermal power of a tower-type solar thermal absorber: ; in, The conditional quantile function, For the quantiles of the generalized linear mixed quantile regression model, ; In order to be in The heat output power of the absorber at any given moment; For fixed effects variables, ,in, for The normal direct irradiance value at that time. for The total reflective area of the mirror field at time t. for The optical efficiency of the mirror field at any given time; In order to be in Fixed effects coefficients at quantiles ,in, In Heat loss coefficient of the receiver at the quantile In Unknown fixed effect coefficients at the quantile, including heliostat surface cleanliness and reflectivity; For random effects variables, ; This indicates whether the current data collection time belongs to spring, summer, autumn, or winter. If yes, the value is 1; otherwise, the value is 0. This indicates whether the current wind speed falls within one of the following five corresponding wind speed ranges: 0 - 2.5 m / sec, 2.5–5 m / sec, 5–7.5 m / sec, 7.5–10 m / sec, and >10 m / sec. If yes, the value is 1; otherwise, the value is 0. This indicates whether the current data acquisition temperature is within one of the following three corresponding temperature ranges: <0℃, 0–10℃, and >10℃. If yes, the value is 1; otherwise, the value is 0. In order to be in Random effects coefficients at quantiles , They are independently and identically distributed; among them, These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different seasons on the heat output of the receiver; These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different wind speeds on the heat output of the absorber. These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different ambient temperatures on the heat output of the absorber. The error term is... Random variables that conform to a normal distribution under quantiles; Step 2: Solve for the unknown parameters in the model. , and ; (1) Obtain n sets of historical time data using the meteorological station and DCS control system of the solar thermal mirror field. sample; (2) Set quantiles The values are then used to obtain the final parameter estimates using the Markov chain Monte Carlo method based on Bayesian theory. , and ; Step 3: Use the obtained model to make predictions; When making predictions, the model obtains the value of m at a future time. Substituting into the model, we obtain the receiver thermal power at future time m. .
[0011] In the above scheme, in step two, the DCS control system based on the photothermal mirror field reads the data from the absorber. Output thermal power value at time t There are a total of n samples. Reading the solar thermal field meteorological station DNI value at time There are a total of n groups of samples Simultaneously, the DCS control system of the mirror field is read. Number of heliostats constantly participating in sun tracking The total reflective area at time t is calculated. , , Given the reflective area of a single heliostat, n sets of samples are obtained. The mirror field was calculated based on the Monte Carlo light tracing method. Optical efficiency at any time There are a total of n groups of samples Based on data from the meteorological station at the mirror field or surrounding meteorological bureaus, obtain seasonal data, temperature data, and wind speed data at the mirror field at time t, thereby obtaining n sets of samples. .
[0012] In the above scheme, in step three, the DNI value at time m in the future is obtained based on the photothermal mirror field meteorological station system. Simultaneously read the mirror field control system in the future Number of heliostats constantly participating in sun tracking The total reflective area at time m is obtained. The optical efficiency of the mirror field at time m in the future was calculated based on the Monte Carlo tracing method. Based on data from the meteorological bureau surrounding the mirror field, obtain the wind speed and temperature at time m in the future, and obtain the future... Moment .
[0013] Through the above technical solution, the method for predicting the output heat power of a tower-type photothermal mirror field receiver provided by the present invention has the following beneficial effects:
[0014] Compared with other traditional methods (general linear models, neural network regression), this invention adopts a generalized linear mixed quantile model. This model not only has no requirements for independent and dependent variables and can be applied to data of any hierarchical structure, but also considers both fixed and random effects, resulting in high prediction accuracy. Furthermore, it is more robust to outliers or abnormal values and takes into account quantiles, making the overall prediction of receiver heat power more accurate and flexible.
[0015] This model considers a variety of random effects, including periodic factors, wind speed, and temperature, and is more accurate in measuring the heat loss of the receiver. Detailed Implementation
[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below.
[0017] This invention provides a method for predicting the output thermal power of a tower-type photothermal mirror field receiver, comprising the following steps:
[0018] Step 1: Establish a generalized linear mixed quantile regression model for predicting the thermal power of a tower-type solar thermal absorber: ; in, The conditional quantile function, For the quantiles of the generalized linear mixed quantile regression model, Commonly used values include 0.25 (25th percentile), 0.5 (median), and 0.75 (75th percentile). Different... The regression coefficients corresponding to the values are , The difference lies in the thermal power of tower-type solar thermal absorbers. Typically, a value of 0.5 is used, with a focus on studying the median distribution of the absorber's thermal power.
[0019] In order to be in The corresponding heat output power of the absorber at any given time.
[0020] For fixed effects variables, ,in, for The normal direct irradiance value at that time. for The total reflective area of the mirror field at time t. for The optical efficiency of the mirror field at any given time;
[0021] In order to be in Fixed effects coefficients at quantiles ,in, In Heat loss coefficient of the receiver at the quantile In Unknown fixed effect coefficients at the quantile, including heliostat surface cleanliness and reflectivity; For random effects variables, ; These values indicate whether the current data collection time belongs to spring, summer, autumn, or winter, respectively. If yes, the value is 1; otherwise, the value is 0. For example, if time t is currently in spring... , For example, at time t, when the region where the mirror is located is in winter, , ; This indicates whether the current wind speed falls within one of the following five corresponding wind speed ranges: 0 - 2.5 m / sec, 2.5–5 m / sec, 5–7.5 m / sec, 7.5–10 m / sec, and >10 m / sec. A value of 1 indicates whether the wind speed falls within these ranges; otherwise, a value of 0 indicates whether it falls within these ranges. For example, if the wind speed at time t is 8 m / s, then... , ; This indicates whether the current data acquisition temperature falls within one of the following three corresponding temperature ranges: <0℃, 0–10℃, and >10℃. If yes, the value is 1; otherwise, it is 0. For example, if the temperature of the area where the mirror is located at time t is 15℃, then... , ; In order to be in Random effects coefficients at quantiles , They are independently and identically distributed; among them, These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different seasons on the heat output of the receiver; These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different wind speeds on the heat output of the absorber. These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different ambient temperatures on the heat output of the absorber. For error terms, It is The quantiles conform to a normal distribution (mean 0, variance ). A random variable; Step 2: Solve for the unknown parameters in the model. , and ; (1) Obtain n sets of historical time data using the meteorological station and DCS control system of the solar thermal mirror field. sample.
[0022] Specifically, the DCS control system based on the photothermal mirror field reads the data from the receiver. Output thermal power value at time t There are a total of n samples. Reading the solar thermal field meteorological station DNI value at time There are a total of n groups of samples Simultaneously, the DCS control system of the mirror field is read. Number of heliostats constantly participating in sun tracking The total reflective area at time t is calculated. , , Given the reflective area of a single heliostat, n sets of samples are obtained. The mirror field was calculated based on the Monte Carlo light tracing method. Optical efficiency at any time There are a total of n groups of samples Based on data from the meteorological station at the mirror field or surrounding meteorological bureaus, obtain seasonal data, temperature data, and wind speed data at the mirror field at time t. The first four columns indicate whether the current data collection time belongs to a specific season (spring, summer, autumn, or winter). The first column indicates which interval the wind speed falls into, and the last three columns indicate which interval the temperature falls into. Z is composed of seasonal, wind speed, and temperature factors, and is obtained from data from the Jingyuan meteorological station or surrounding meteorological bureaus, thus obtaining n sets of samples. .
[0023] (2) Set quantiles The value is taken as follows. Alternatively, using fixed values such as 0.95 (0.5 represents the median, 0.95 represents the 95th percentile), the final parameter estimates can be obtained using the Markov chain Monte Carlo method based on Bayesian theory. , and ;
[0024] Step 3: Use the obtained model to make predictions.
[0025] When making predictions, the model obtains the value of m at a future time. Substituting into the model, we obtain the receiver thermal power at future time m. .
[0026] Specifically, based on the solar thermal field meteorological station system, the DNI value at time m in the future is obtained. Simultaneously read the mirror field control system in the future Number of heliostats constantly participating in sun tracking The total reflective area at time m is obtained. The optical efficiency of the mirror field at time m in the future was calculated based on the Monte Carlo tracing method. Based on data from the meteorological bureau surrounding the mirror field, obtain the wind speed and temperature at time m in the future, and obtain the future... Moment .
[0027] Substituting the data obtained above into the model, we obtain the future time m. Conditional quantile function under quantiles: ; Thus, we can obtain Predicted power value of the receiver at the quantile level .
[0028] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting the thermal power output of a tower optical thermal mirror field absorber, characterized in that, Comprising the following steps: Step one, based on the establishment of tower type heat power prediction of photothermal heat absorption generalized linear mixed quantile regression model: ； wherein, is a conditional quantile function, is a quantile of a generalized linear mixed quantile regression model, is the output thermal power of the heat sink at the time instant ; is the fixed effect variable, wherein, is the normal direct irradiance value at time t, is the total reflective area of the mirror field at time t, is the optical efficiency of the mirror field at time t; to be in fixed effect coefficient at quantile q, wherein, denotes the heat loss coefficient of the heat receiver at quantile q, fixed effect coefficient at quantile q, denotes the unknown fixed effect coefficient at quantile q, containing the mirror surface cleanliness, reflectivity of the heliostat at quantile q; for the random effect variables, ; represent whether the current data collection time belongs to spring, summer, autumn, winter, respectively, if yes, the value is 1, if not, the value is 0; represent whether the current data collection wind speed belongs to the following five corresponding wind speed intervals, 0 - 2.5 m / sec, 2.5 - 5 m / sec, 5 - 7.5 m / sec, 7.5 - 10 m / sec, > 10 m / sec, respectively, if yes, the value is 1, if not, the value is 0; represent whether the current data collection temperature is in the following three corresponding temperature intervals, <0℃, 0 - 10℃, >10℃, respectively, if yes, the value is 1, if not, the value is 0; In order to be in Random effects coefficients at quantiles , They are independently and identically distributed; among them, These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different seasons on the heat output of the receiver; These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different wind speeds on the heat output of the absorber. These are the random effect variables Z(t) The coefficient represents the intensity of the influence of different ambient temperatures on the heat output of the absorber. is an error term, is a random variable that follows a normal distribution at the quantile a random variable that follows a normal distribution at the quantile Step two, solving for unknown parameters in the model , and ; (1) Obtain n groups of samples at historical time by using the meteorological station and the DCS control system of the photothermal mirror field sample; (2) Set the value of the quantile number , and obtain the final parameter estimation value by using the Markov chain Monte Carlo method based on Bayesian theory , and ; Step three, using the model obtained for prediction; When the model is predicted, the future time m is obtained , the model is substituted, and the heat absorption heat power of the future time m is obtained .
2. A method of predicting the thermal power output of a tower optical thermal mirror field absorber according to claim 1, characterized in that, In step two, the DCS control system based on the field of the CTP reads the output thermal power value of the heat absorber at time t , , n sets of samples ; the field of the CTP weather station reads the DNI value at time t , , n sets of samples ; at the same time, the DCS control system of the field of the CTP reads the number of heliostats participating in the sun tracking at time t , , and the total reflection area at time t is calculated , , , which is the reflection area of a single heliostat, n sets of samples ; the optical efficiency of the field of the CTP at time t is calculated by the light path tracking method based on Monte Carlo , , n sets of samples ; according to the data of the field of the CTP weather station or the surrounding weather bureau, the seasonal data at the field of the CTP at time t is obtained, the temperature data and the wind speed data, so as to obtain n sets of samples .
3. A method of predicting the thermal power output of a tower optical thermal mirror field absorber according to claim 1, characterized in that, In step three, according to the light-thermal mirror field meteorological station system, the DNI value at the future m time is obtained , the mirror field control system is read at the same time The number of heliostats participating in the sun tracking at the future m time , the total reflection area at the m time is obtained , the optical efficiency of the mirror field at the future m time is calculated based on the Monte Carlo light trace tracking method ; according to the meteorological data of the mirror field periphery, the wind speed and temperature at the future m time are obtained, and the future .