Linear Fresnel photo-thermal power generation system technological process optimization process based on fused salt heat storage

Abstract

The invention discloses a linear Fresnel photo-thermal power generation system technological process optimization technology based on fused salt heat storage. The method comprises the following steps: collecting system operation historical data, preprocessing, mining features, and constructing a prediction model; the radiation intensity is predicted according to the model, the condensation angle and the tracking strategy of the heat collector are dynamically adjusted, fused salt flow distribution is optimized, and steam and steam turbine generator unit operation parameters are adjusted. And establishing a multi-objective optimization function for global optimization, and dynamically correcting the strategy through a feedback adjustment mechanism. In data processing, a sliding window method is adopted to correct radiation intensity data, and a piecewise linear interpolation method is adopted to process fused salt temperature data; in model construction, an LSTM network is used to predict radiation intensity and a compensation factor is introduced. The method can effectively improve the solar energy utilization efficiency of the power generation system, enhance the system stability, improve the power generation efficiency, reduce the operation cost and realize the self-adaptive optimization operation of the system.

Description

Technical Field

[0001] The present invention relates to the field related to solar thermal power generation systems, and specifically to a process flow optimization technology for a linear Fresnel solar thermal power generation system based on molten salt heat storage. Background Art

[0002] Solar thermal power generation technology, as a key area of renewable energy utilization, has garnered widespread attention. Linear Fresnel CSP systems, with their unique technical features, have captured a significant market share in the field. However, current linear Fresnel CSP systems based on molten salt heat storage still face numerous process challenges. Devices such as the linear Nefert heliostats mentioned in Prior Art Publication CN208578678U, an improved Brayton CSP system based on linear Nefert heliostats, can still fail to fully absorb solar radiation when tracking the sun's angle due to insufficient accuracy or interference from environmental factors. Furthermore, the molten salt in the system's heat collection circuit leads to inadequate heat absorption, resulting in energy waste. The uneven temperature distribution of the molten salt within the heat storage tank can lead to increased heat loss and reduced heat storage efficiency. Summary of the Invention

[0003] The object of the present invention is to provide a process flow optimization process for a linear Fresnel solar thermal power generation system based on molten salt heat storage to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a process flow optimization process for a linear Fresnel solar thermal power generation system based on molten salt heat storage, comprising the following steps:

[0005] S1: Collect historical operating data of the linear Fresnel CSP system, including multi-dimensional data on solar radiation intensity, molten salt temperature, steam parameters, and power generation, and establish a data repository;

[0006] S2: Use time series analysis and machine learning algorithms to mine the features of historical data in the data repository, identify potential correlations between solar radiation intensity, molten salt temperature, and power generation, and build a data-driven prediction model.

[0007] S3: Based on the prediction model, the solar radiation intensity under different meteorological conditions is predicted in real time, and the collector's focusing angle and tracking strategy are dynamically adjusted according to the predicted value to ensure that the linear Fresnel collector absorbs solar energy to the greatest extent;

[0008] S4: Based on the real-time solar radiation intensity prediction results and the heat absorption of the collector, optimize the flow distribution of molten salt between the heat collection circuit and the heat storage circuit, optimize the temperature distribution of molten salt in the heat storage tank, and maintain the thermal stability and maximum heat storage efficiency of the molten salt storage system;

[0009] S5: According to the optimized molten salt flow distribution strategy and molten salt temperature conditions, the steam generation parameters in the steam generation system are accurately adjusted to improve steam quality;

[0010] S6: Using high-quality steam as the working medium, the operating parameters of the steam turbine generator set, such as inlet steam pressure, temperature and flow, are optimized to improve power generation efficiency;

[0011] S7: Based on the operating status of each link of the system, real-time meteorological conditions, and market power demand, a multi-objective optimization function is established to globally optimize the process flow of the entire linear Fresnel CSP system based on molten salt heat storage;

[0012] S8: Continuously monitor the actual operating parameters of the system and compare and analyze them with the optimization targets. Dynamically modify the process optimization strategy through the feedback adjustment mechanism to achieve adaptive optimization operation of the system.

[0013] Furthermore, when optimizing the molten salt flow path in the heat collection stage in S4, the relationship between the molten salt flow rate and the heat collection efficiency is followed:

[0014] Q=m×C P ×ΔT

[0015] Among them, Q is the heat absorbed by the collector (KJ), m is the mass flow rate of molten salt (kg / s), C P is the constant-pressure specific heat capacity of the molten salt (kJ / (kg·K)), and ΔT is the temperature difference between the inlet and outlet of the molten salt (k). By calculating the heat Q absorbed by the molten salt under different flow paths, the molten salt flow path that maximizes Q is selected as the optimized flow path to improve the efficiency of the heat collection stage.

[0016] Furthermore, in said S4, when optimizing the temperature distribution of the molten salt in the heat storage tank, the heat conduction equation is considered:

[0017]

[0018] Where T is the temperature of the molten salt (k), t is the time (s), is the thermal diffusivity of the molten salt (m 2 / s), k is the thermal conductivity of the molten salt (W / (m·K)), ρ is the density of the molten salt (kg / m 3 ), C P is the constant-pressure specific heat capacity of the molten salt (kJ / (kg·K)). By simulating this heat conduction equation, the inlet and outlet locations and flow rates of the molten salt are adjusted to make the temperature distribution of the molten salt in the heat storage tank as uniform as possible, thereby reducing heat loss.

[0019] Furthermore, in S5, the heat exchange process between the molten salt and the water in the steam generator is optimized, and the calculation is performed based on the logarithmic mean temperature difference method:

[0020]

[0021] Where ΔT1 and ΔT2 are the temperature differences (K) between the hot and cold fluids at the steam generator inlet and outlet, respectively. By adjusting the flow rates of molten salt and water, as well as the structural parameters of the steam generator, the logarithmic mean temperature difference (LMTD) can be increased, thereby improving heat exchange efficiency and, consequently, power generation efficiency.

[0022] Furthermore, the pre-processing of the data repository in S1 includes:

[0023] The sliding window method is used to correct outliers in the solar radiation intensity data. The correction formula is:

[0024]

[0025] Among them, I corr (t) is the radiation intensity at time t after correction, N is the window length, and λ is the attenuation coefficient, which is used to weaken the lag effect of historical data;

[0026] The missing values of molten salt temperature data were compensated by piecewise linear interpolation, and the temperature change rate in the interpolation interval was calculated by the ratio of the temperature difference to the time difference between adjacent time stamps.

[0027] Furthermore, the prediction model construction in S2 includes:

[0028] The LSTM network is used to predict solar radiation intensity. The input layer is the time series I(t), T s (t), P g (t)), the output layer is the radiation intensity of the future Δt period

[0029] Introducing the molten salt temperature hysteresis effect compensation factor β=1-e -τ / γ , where τ is the thermal inertia time constant of the molten salt circuit and γ is the system response coefficient, which is used to correct the deviation between the predicted value and the actual working condition.

[0030] Furthermore, the strategy for dynamically adjusting the collector focusing angle in S3 is: based on the real-time solar altitude angle θ and azimuth angle Calculate the ideal focusing angle α:

[0031]

[0032] Considering the influence of wind speed v on tracking accuracy, the correction coefficient η=1-0.02v is introduced 2 , and finally the focusing angle is controlled to be α0=α·η, ensuring that the mirror field maintains efficient heat collection under wind load interference.

[0033] Furthermore, the molten salt flow rate distribution optimization in S4 includes:

[0034] Establish the flow ratio R between the heat collection circuit and the heat storage circuit f =Q c / Q s Dynamic adjustment model of:

[0035]

[0036] Among them, K p , K i is the proportional and integral coefficient, ΔT s is the molten salt temperature difference, ΔT t The maximum temperature difference threshold allowed by the system;

[0037] When the molten salt temperature at the collector field outlet is ≥550℃, the heat storage loop flow priority strategy is triggered to limit the heat collection loop flow to prevent overheating.

[0038] Furthermore, the steam turbine unit operating parameter optimization in S6 includes:

[0039] Establish the relationship between turbine efficiency η and inlet steam pressure P in , temperature T in The relational model:

[0040] η=0.85-0.02·(P in -P opt ) 2 -0.015·(T in -T opt ) 2

[0041] Among them, P opt =12.5MPa, T opt =540℃ is the optimal design parameter;

[0042] When the grid demand load changes, the steam inlet pressure is adjusted first to maintain the efficiency η≥0.78.

[0043] Furthermore, the multi-objective optimization function in S7 is:

[0044]

[0045] Among them, w1, w2, w3 are weight coefficients, P g is the real-time power generation, E s is the energy of the heat storage tank, C l For the system heat loss, the optimal weight combination is solved by particle swarm optimization.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention uses time series analysis and machine learning algorithms to perform feature mining on historical data, constructing a data-driven prediction model that can predict the solar radiation intensity under different meteorological conditions in real time. By dynamically adjusting the collector's focusing angle and tracking strategy based on the prediction results, the present invention can ensure that the linear Fresnel collector tracks the sun's movement more accurately, absorbs solar radiation to the greatest extent, reduces solar energy waste caused by inaccurate tracking, and improves the efficiency of photothermal conversion. By calculating the amount of heat absorbed by the molten salt under different flow paths and selecting the optimal flow path, the problem of insufficient heat absorption of the molten salt in the heat collection circuit in the prior art is effectively solved, thereby improving the efficiency of the heat collection stage. At the same time, in the molten salt heat storage stage, considering the heat conduction equation, by simulating and adjusting the inlet and outlet positions and flow rate of the molten salt, the temperature distribution of the molten salt in the heat storage tank is made more uniform, reducing heat loss, improving heat storage efficiency and stability, and reducing energy waste. By optimizing the flow distribution of the molten salt between the heat collection circuit and the heat storage circuit, the steam generation parameters in the steam generation system are accurately adjusted, which can improve the steam quality. Using high-quality steam as the working medium, optimizing the operating parameters of the steam turbine generator set, such as steam inlet pressure, temperature, and flow, establishing a relationship model between steam turbine efficiency and steam inlet parameters, and prioritizing the adjustment of steam inlet pressure according to changes in grid demand load can effectively improve power generation efficiency and solve the problem of limited power generation efficiency in existing technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Flowchart of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] See also Figure 1 The present invention provides a process flow optimization process for a linear Fresnel solar thermal power generation system based on molten salt heat storage, comprising the following steps:

[0051] S1. Collect historical operating data of the linear Fresnel solar thermal power generation system, including multi-dimensional data such as solar radiation intensity, molten salt temperature, steam parameters, and power generation.

[0052] Establish a data repository to store the above historical data.

[0053] Data preprocessing:

[0054] The sliding window method is used to correct outliers in the solar radiation intensity data. The correction formula is:

[0055]

[0056] Among them, I corr (t) is the radiation intensity at time t after correction, N is the window length, λ is the attenuation coefficient,

[0057] Used to reduce the lag effect of historical data;

[0058] The missing values of molten salt temperature data were compensated by piecewise linear interpolation, and the temperature change rate in the interpolation interval was calculated by the ratio of the temperature difference to the time difference between adjacent time stamps.

[0059] Specifically, high-precision sensors and automated monitoring equipment are used to comprehensively collect operational data from the power generation system. Multiple solar radiation sensors are installed in the collector area to obtain solar radiation intensity data at different locations. Multiple temperature sensors are deployed in the molten salt circuit to monitor real-time temperature changes in the molten salt during heat collection, storage, and exchange. Steam parameters are collected using specialized monitoring instruments in the steam generation system, and power generation data is obtained directly from the turbine generator unit's power monitoring device. This data is recorded at regular 10-minute intervals to ensure data continuity and integrity. A dedicated data repository is established, utilizing an advanced database management system to categorize and store the collected data. This data is organized by data type and time series, creating a clear data structure to facilitate subsequent data query, analysis, and processing. Furthermore, a data backup mechanism is implemented to regularly back up data to off-site storage devices to prevent data loss. A sliding window method is used to correct outliers in the solar radiation intensity data. The window length, N, is set to 15 based on data fluctuations and practical experience, and the attenuation coefficient, λ, is determined to be 0.2 through multiple experiments and data analysis. A formula is used to correct the radiation intensity data at each time point.

[0060] For molten salt temperature data, if there are missing values, compensation is performed to ensure data continuity and provide accurate data support for subsequent analysis and decision-making.

[0061] S2. Use time series analysis and machine learning algorithms to perform feature mining on historical data in the data repository to identify potential correlation patterns between solar radiation intensity, molten salt temperature, and power generation.

[0062] Prediction model construction:

[0063] The LSTM network is used to predict solar radiation intensity. The input layer is the time series I(t), T s (t), P g (t)), the output layer is the radiation intensity of the future Δt period

[0064] Introducing the molten salt temperature hysteresis effect compensation factor β=1-e -τ / γ , where τ is the thermal inertia time constant of the molten salt circuit and γ is the system response coefficient, which is used to correct the deviation between the predicted value and the actual working condition.

[0065] S3. Based on the above prediction model, the solar radiation intensity under different meteorological conditions is predicted in real time. The collector's focusing angle and tracking strategy are dynamically adjusted according to the predicted value to ensure that the linear Fresnel collector absorbs solar energy to the greatest extent possible.

[0066] Calculation of spotlight angle:

[0067] Based on real-time solar altitude angle θ and azimuth Calculate the ideal focusing angle α:

[0068]

[0069] Considering the influence of wind speed v on tracking accuracy, the correction coefficient η=1-0.02v is introduced 2 , and finally the focusing angle is controlled to be α0=α·η, ensuring that the mirror field maintains efficient heat collection under wind load interference.

[0070] S4. Based on the real-time solar radiation intensity prediction results and the heat absorption of the collector, optimize the flow distribution of molten salt between the heat collection circuit and the heat storage circuit to maintain the thermal stability and maximum heat storage efficiency of the molten salt storage system;

[0071] Dynamic adjustment of flow ratio:

[0072] Establish the flow ratio R between the heat collection circuit and the heat storage circuit f =Q c / Q s Dynamic adjustment model of:

[0073]

[0074] Among them, K p , K i is the proportional and integral coefficient, ΔT s is the molten salt temperature difference, ΔT t The maximum temperature difference threshold allowed by the system;

[0075] When the molten salt temperature at the collector field outlet is ≥550℃, the heat storage loop flow priority strategy is triggered to limit the heat collection loop flow to prevent overheating.

[0076] When optimizing the temperature distribution of molten salt in the thermal storage tank, the heat conduction equation is considered:

[0077]

[0078] Where T is the temperature of the molten salt (k), t is the time (s), is the thermal diffusivity of the molten salt (m 2 / s), k is the thermal conductivity of the molten salt (W / (m·K)), ρ is the density of the molten salt (kg / m 3 ), C P is the constant-pressure specific heat capacity of the molten salt (kJ / (kg·K)). By simulating this heat conduction equation, the inlet and outlet locations and flow rates of the molten salt are adjusted to make the temperature distribution of the molten salt in the heat storage tank as uniform as possible, thereby reducing heat loss.

[0079] S5. According to the optimized molten salt flow distribution strategy and molten salt temperature conditions, accurately adjust the steam generation parameters in the steam generation system to improve steam quality.

[0080] Optimize the heat exchange process between molten salt and water in the steam generator, and calculate based on the logarithmic mean temperature difference method:

[0081]

[0082] Where ΔT1 and ΔT2 are the temperature differences (K) between the hot and cold fluids at the steam generator inlet and outlet, respectively. By adjusting the flow rates of molten salt and water, as well as the structural parameters of the steam generator, the logarithmic mean temperature difference (LMTD) can be increased, thereby improving heat exchange efficiency and, consequently, power generation efficiency.

[0083] S6. Using high-quality steam as the working medium, optimize the operating parameters of the steam turbine generator set, including steam inlet pressure, temperature, and flow, to improve power generation efficiency;

[0084] Operation parameter optimization:

[0085] Establish the relationship between turbine efficiency η and inlet steam pressure P in , temperature T in The relational model:

[0086] η=0.85-0.02·(P in -P opt ) 2 -0.015·(T in -T opt ) 2

[0087] Among them, P opt =12.5MPa, T opt =540℃ is the optimal design parameter;

[0088] When the grid demand load changes, the steam inlet pressure is adjusted first to maintain the efficiency η≥0.78.

[0089] S7. Comprehensively consider the operating status of each link in the system, real-time meteorological conditions, and market power demand, establish a multi-objective optimization function, and perform global optimization of the entire process flow of the linear Fresnel CSP system based on molten salt heat storage;

[0090] The multi-objective optimization function is:

[0091]

[0092] Among them, w1, w2, w3 are weight coefficients, P g is the real-time power generation, E s is the energy of the heat storage tank, C l For the system heat loss, the optimal weight combination is solved by particle swarm optimization.

[0093] S8. Considering the operating status of each link of the system, real-time meteorological conditions and market electricity demand, a multi-objective optimization function is established to globally optimize the process flow of the entire linear Fresnel solar thermal power generation system based on molten salt heat storage.

[0094] Feedback regulation mechanisms include:

[0095] Real-time calculation of system comprehensive efficiency deviation Δη:

[0096]

[0097] When Δη≤-5%, the molten salt flow redistribution and concentration angle coordinated correction strategy is triggered. By adjusting the flow distribution of molten salt in the thermal collection circuit and the heat storage circuit, and adjusting the concentration angle of the collector at the same time, the system operating efficiency is restored until Δη returns to within the ±2% threshold.

Claims

1. A process flow optimization process for a linear Fresnel solar thermal power generation system based on molten salt heat storage, characterized in that: The following steps are involved: S1: Collect historical operating data of the linear Fresnel CSP system, including multi-dimensional data on solar radiation intensity, molten salt temperature, steam parameters, and power generation, and establish a data repository; S2: Use time series analysis and machine learning algorithms to mine the features of historical data in the data repository, identify potential correlations between solar radiation intensity, molten salt temperature, and power generation, and build a data-driven prediction model. S3: Based on the prediction model, the solar radiation intensity under different meteorological conditions is predicted in real time, and the collector's focusing angle and tracking strategy are dynamically adjusted according to the predicted value to ensure that the linear Fresnel collector absorbs solar energy to the greatest extent; S4: Based on the real-time solar radiation intensity prediction results and the heat absorption of the collector, optimize the flow distribution of molten salt between the heat collection circuit and the heat storage circuit, optimize the temperature distribution of molten salt in the heat storage tank, and maintain the thermal stability and maximum heat storage efficiency of the molten salt storage system; S5: According to the optimized molten salt flow distribution strategy and molten salt temperature conditions, the steam generation parameters in the steam generation system are accurately adjusted to improve steam quality; S6: Using high-quality steam as the working medium, the operating parameters of the steam turbine generator set, such as inlet steam pressure, temperature and flow, are optimized to improve power generation efficiency; S7: Based on the operating status of each link of the system, real-time meteorological conditions, and market power demand, a multi-objective optimization function is established to globally optimize the process flow of the entire linear Fresnel CSP system based on molten salt heat storage; S8: Continuously monitor the actual operating parameters of the system and compare and analyze them with the optimization targets. Dynamically modify the process optimization strategy through the feedback adjustment mechanism to achieve adaptive optimization operation of the system.

2. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: When optimizing the molten salt flow path in the heat collection stage in S4, the relationship between the molten salt flow rate and the heat collection efficiency is followed: Q=m×C P ×ΔT Among them, Q is the heat absorbed by the collector (KJ), m is the mass flow rate of molten salt (kg / s), C P is the constant-pressure specific heat capacity of the molten salt (kJ / (kg·K)), and ΔT is the temperature difference between the inlet and outlet of the molten salt (k). By calculating the heat Q absorbed by the molten salt under different flow paths, the molten salt flow path that maximizes Q is selected as the optimized flow path to improve the efficiency of the heat collection stage.

3. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: In S4, when optimizing the temperature distribution of the molten salt in the heat storage tank, the heat conduction equation is considered: Where T is the temperature of the molten salt (k), t is the time (s), is the thermal diffusivity of the molten salt (m 2 / s), k is the thermal conductivity of the molten salt (W / (m·K)), ρ is the density of the molten salt (kg / m 3 ), C P is the constant-pressure specific heat capacity of the molten salt (kJ / (kg·K)). By simulating the heat conduction equation, the inlet and outlet positions and flow rate of the molten salt are adjusted to make the temperature distribution of the molten salt in the heat storage tank as uniform as possible and reduce heat loss.

4. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: In S5, the heat exchange process between the molten salt and the water in the steam generator is optimized, and the calculation is performed based on the logarithmic mean temperature difference method: Where ΔT1 and ΔT2 are the temperature differences (K) between the hot and cold fluids at the steam generator inlet and outlet, respectively. By adjusting the flow rates of molten salt and water, as well as the structural parameters of the steam generator, the logarithmic mean temperature difference (LMTD) can be increased, thereby improving heat exchange efficiency and, consequently, power generation efficiency.

5. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: The preprocessing of the data repository in S1 includes: The sliding window method is used to correct outliers in the solar radiation intensity data. The correction formula is: Among them, I corr (t) is the radiation intensity at time t after correction, N is the window length, and λ is the attenuation coefficient, which is used to weaken the lag effect of historical data; The missing values of molten salt temperature data were compensated by piecewise linear interpolation, and the temperature change rate in the interpolation interval was calculated by the ratio of the temperature difference to the time difference between adjacent time stamps.

6. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: The prediction model construction in S2 includes: The LSTM network is used to predict solar radiation intensity. The input layer is the time series I(t), T s (t), P g (t)), the output layer is the radiation intensity of the future Δt period Introducing the molten salt temperature hysteresis effect compensation factor β=1-e -τ / γ , where τ is the thermal inertia time constant of the molten salt circuit and γ is the system response coefficient, which is used to correct the deviation between the predicted value and the actual working condition.

7. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: The strategy for dynamically adjusting the collector focusing angle in S3 is: based on the real-time solar altitude angle θ and azimuth angle Calculate the ideal focusing angle α: Considering the influence of wind speed v on tracking accuracy, the correction coefficient η=1-0.02v is introduced 2 , and finally the focusing angle is controlled to be α0=α·η, ensuring that the mirror field maintains efficient heat collection under wind load interference.

8. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: The optimization of molten salt flow distribution in S4 includes: Establish the flow ratio R between the heat collection circuit and the heat storage circuit f =Q c / Q s Dynamic adjustment model of: Among them, K p , K i is the proportional and integral coefficient, ΔT s is the temperature difference of molten salt, ΔT t The maximum temperature difference threshold allowed by the system; When the molten salt temperature at the collector field outlet is ≥550℃, the heat storage loop flow priority strategy is triggered to limit the heat collection loop flow to prevent overheating.

9. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: The steam turbine unit operating parameter optimization in S6 includes: Establish the relationship between turbine efficiency η and inlet steam pressure P in , temperature T in The relational model: η=0.85-0.02·(P in -P opt ) 2 -0.015·(T in -T opt ) 2 Among them, P opt =12.5MPa, T opt =540℃ is the optimal design parameter; When the grid demand load changes, the steam inlet pressure is adjusted first to maintain the efficiency η≥0.

78.

10. The process flow optimization process of the linear Fresnel solar thermal power generation system based on molten salt heat storage according to claim 1 is characterized in that: The multi-objective optimization function in S7 is: Among them, w1, w2, w3 are weight coefficients, P g is the real-time power generation, E s is the energy of the heat storage tank, C l For the system heat loss, the optimal weight combination is solved by particle swarm optimization.

Images