A molten salt energy storage system coordinated regulation method and system

Abstract

This invention relates to the fields of molten salt energy storage and concentrated solar power (CSP) technology, specifically to a collaborative control method and system for molten salt energy storage systems. The collaborative control method for molten salt energy storage systems includes the following steps: calculating the comprehensive operating condition coefficient at the current time t, characterizing the comprehensive operating state of the molten salt energy storage system at time t; using the comprehensive operating condition coefficient as the core input, predicting the sustainable output power of the molten salt energy storage system at future times based on an autoregressive exogenous model; using the sustainable output power at future times as a feedforward setpoint, combined with feedback correction from the comprehensive operating condition coefficient, calculating the optimal molten salt volumetric flow rate to be set at the current time, achieving adaptive molten salt flow rate control; after the charge-discharge cycle ends, calculating the actual cycle efficiency of the molten salt energy storage system, constructing a multi-objective optimization function; using the multi-objective optimization function as the objective, updating the parameters of the autoregressive exogenous model online, and using the updated autoregressive exogenous model parameters for prediction of the next cycle, forming closed-loop control.

Description

Technical Field

[0001] This invention relates to the fields of molten salt energy storage and solar thermal power generation, specifically to a method and system for coordinated control of molten salt energy storage systems. Background Technology

[0002] Molten salt energy storage technology, with its advantages of high energy density, low cost, and wide operating temperature range, has become one of the preferred solutions for large-scale, long-term energy storage, and is widely used in solar thermal power generation, grid peak shaving, and industrial waste heat recovery. The operation and control of molten salt energy storage systems are directly related to their safety, economy, and flexibility.

[0003] The mainstream control strategies for existing molten salt energy storage systems are mainly divided into fixed-parameter control based on fixed rules and feedback control based on a single objective (such as temperature or power). While fixed-parameter control (such as fixed flow rate and fixed temperature setpoint) is simple and reliable, it struggles to cope with dynamic changes in external loads and internal thermal-hydraulic states, easily leading to low system efficiency or safety risks such as molten salt solidification and excessive pipeline thermal stress under extreme conditions. Single-objective feedback control (such as adjusting flow rate based on outlet temperature), while capable of regulating a single variable, lacks comprehensive consideration of the coupled effects of multiple physical fields (hydraulic, thermal, and mechanical) within the system. For example, controlling only temperature may ignore changes in pipeline pressure drop and heat transfer intensity, leading to a decrease in overall system energy efficiency; tracking only power demand may fail to respond promptly to the accumulation of internal risks.

[0004] Furthermore, molten salt systems exhibit significant thermal inertia and nonlinear characteristics, leading to response lag issues when relying solely on feedback control, making it difficult to quickly track the grid's rapid power commands. With the increasing penetration of renewable energy, the grid's requirements for the regulation speed and accuracy of energy storage systems are becoming increasingly stringent, highlighting the limitations of existing control methods. Summary of the Invention

[0005] The purpose of this invention is to provide a collaborative control method for molten salt energy storage systems to solve the problem of response lag in the mainstream control methods of molten salt energy storage systems in the prior art.

[0006] To address the aforementioned problems, this invention proposes a collaborative control method for molten salt energy storage systems. The technical solution adopted is as follows: A method for coordinated control of a molten salt energy storage system includes the following steps: The method executes the following steps sequentially within a single control cycle to achieve logical progression and control closed loop: Step S1: Couple the hydraulic characteristics, heat transfer characteristics, solidification risk and thermal stress risk of the molten salt energy storage system, calculate the comprehensive operating condition coefficient Φ(t) at the current time t, characterize the comprehensive operating state of the molten salt energy storage system at time t, and use it to quantify the degree to which the system deviates from the ideal operating state; Step S2: Using the comprehensive operating condition coefficient Φ(t) as the core input, and based on the autoregressive exogenous model, predict the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt. Step S3: Using the sustainable output power SOP(t+Δt) at the future time t+Δt as the feedforward setpoint, and combining it with the comprehensive operating condition coefficient Φ(t) for feedback correction, calculate the optimal molten salt volumetric flow rate F that should be set at the current time t. opt (t), to achieve adaptive molten salt flow rate control; Step S4: After a complete charge-discharge cycle, calculate the actual cycle efficiency of the molten salt energy storage system. A multi-objective optimization function is constructed; the parameters of the autoregressive exogenous model are updated online using the multi-objective optimization function as the objective; the updated autoregressive exogenous model parameters are used in step S2 to predict the next cycle, forming closed-loop control.

[0007] Furthermore, the comprehensive operating condition coefficient As shown in Equation 1: Formula 1 In Equation 1, Let be the coefficient of friction of the pipe at time t; This is the reference friction coefficient for the molten salt energy storage system under design conditions; Let be the Nusselt number at time t. For reference Nusselt numbers; Let be the solidification risk factor at time t. The thermal stress sensitivity coefficient of the pipe material; Let t be the maximum temperature difference of the molten salt across the cross-section of the pipe.

[0008] Furthermore, in step S2, the molten salt energy storage system's sustainable output power at a future time t+Δt is... As shown in Equation 2: Formula 2 In Equation 2, Let be the measured sustainable output power at time t; α, β, γ, and δ are the weighting coefficients of the autoregressive exogenous model. The comprehensive operating condition coefficient at time t; Let t be the power demand of the external load. This refers to the system's rated output power. Let t be the inlet temperature of the molten salt energy storage system. , These represent the minimum and maximum permissible operating temperatures of the molten salt energy storage system, respectively; τ is the thermal inertia time constant of the molten salt energy storage system. Let be the rate of change of external load demand power at time t.

[0009] Furthermore, in step S3, the optimal molten salt volumetric flow rate F that should be set at the current time t is... opt (t) is shown in Equation 3: Formula 3 In Equation 3, For molten salt energy storage systems in the future The continuous output power at any given moment; Let be the density of the molten salt at time t at the operating temperature; Let be the isobaric specific heat capacity of the molten salt at time t; The rated temperature drop or rise of the molten salt set within the heat exchanger; K p For proportional gain; T i The integral time constant; To address the deviation in overall operating conditions The integral from the previous control cycle to the current time t; e(t) is the comprehensive operating condition deviation at time t. , Φ ideal Φ(t) is the ideal comprehensive working condition coefficient; Φ(t) is the comprehensive working condition coefficient at time t.

[0010] Furthermore, in step S4, the actual cycle efficiency η actual As shown in Equation 4: Formula 4 In Equation 4, This represents the total actual output energy within a single cycle. P represents the real-time power consumption of the molten salt pump. trace (τ) represents the real-time power consumption of the electric heat tracing system; Q loss This represents the static heat loss of the system during the cycle.

[0011] Furthermore, the online-updated autoregressive exogenous model parameters include weight coefficients α, β, γ, and δ, and their corresponding optimization objective is: ; Where, η actual η represents the actual cycle efficiency of the system. design The system is designed for loop efficiency; λ1, λ2, and λ3 are multi-objective optimization weight coefficients; The predicted power sequence SOP in this cycle is given by Equation 2. pred Mean square error between the actual power sequence SOPreal and the actual power sequence; The rated output power of the system is denoted as Var(Φ); Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle.

[0012] Furthermore, the online updating of the autoregressive exogenous model parameters with a multi-objective optimization function as the objective includes: Efficiency bias, prediction error, and operational stability are used as optimization objectives. An optimization objective function is constructed, and the weight coefficients α, β, γ, and δ of the autoregressive exogenous model are iteratively optimized using stochastic gradient descent or particle swarm optimization to achieve online updating of the autoregressive exogenous model parameters. The optimization objective function is shown in Equation 5. Formula 5 In Equation 5, η actual η represents the actual cycle efficiency of the system. design The system is designed for loop efficiency; λ1, λ2, and λ3 are multi-objective optimization weight coefficients; The predicted power sequence SOP in this cycle is given by Equation 2. pred With actual power sequence SOP real The mean square error; The rated output power of the system is denoted as Var(Φ); Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle.

[0013] Furthermore, the stochastic gradient descent method includes: First, the weight coefficients α, β, γ, and δ of the autoregressive exogenous model parameters in Equation 2 are used as the initial point parameters θ0. Secondly, calculate the optimization objective function. In the current parameter θ k The gradient vector is obtained by taking the partial derivative with respect to each weight coefficient. J(θ k ); Next, update the weight coefficients according to the following formula 6: θ (k+1) =θ k -η learn J(θ k Formula 6 In Equation 6, η learn Let θ be the learning rate. k For the current parameter, J(θ k ) represents the gradient vector; Then, repeat steps two and three until the preset iteration termination condition is met, and obtain the weight coefficients of the final optimized autoregressive exogenous model. Finally, the weight coefficients of the final optimized autoregressive exogenous model are updated in Equation 2.

[0014] Furthermore, the iteration termination condition includes the following termination condition: Optimize objective function The change is less than the first preset threshold εJ; And / or, reaching the maximum number of iterations Kmax ; And / or, the norm of the gradient vector is less than a second preset threshold ε. .

[0015] The present invention also provides a system for performing the above-described molten salt energy storage system coordinated control method, comprising: The comprehensive operating condition coefficient Φ(t) acquisition module is used to couple the hydraulic characteristics, heat transfer characteristics, solidification risk and thermal stress risk of the molten salt energy storage system, calculate the comprehensive operating condition coefficient Φ(t) at the current time t, characterize the comprehensive operating state of the molten salt energy storage system at time t, and quantify the degree to which the system deviates from the ideal operating state. The module for obtaining the sustainable output power SOP(t+Δt) at the future time t+Δt is used to predict the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt based on an autoregressive exogenous model, with the comprehensive operating condition coefficient Φ(t) as the core input. The optimal molten salt volumetric flow rate F that should be set at present opt The (t) acquisition module is used to calculate the optimal molten salt volumetric flow rate F that should be set at the current time t, using the sustainable output power SOP(t+Δt) at a future time t+Δt as the feedforward setpoint and combined with the comprehensive operating condition coefficient Φ(t) for feedback correction. opt (t), to achieve adaptive molten salt flow rate control; The closed-loop control module is used to calculate the actual cycle efficiency of the molten salt energy storage system after a complete charge-discharge cycle. A multi-objective optimization function is constructed; the parameters of the autoregressive exogenous model are updated online using the multi-objective optimization function as the objective; the updated autoregressive exogenous model parameters are used in step S2 to predict the next cycle, forming closed-loop control.

[0016] Compared with existing technologies, the synergistic control method and system for molten salt energy storage systems proposed in this invention have the following beneficial effects: Enhanced safety and forward-looking capabilities: This invention dynamically quantifies the hydraulic, heat transfer, and safety (solidification, thermal stress) states within the system by using the comprehensive operating condition coefficient Φ(t) at the current time t. Combined with the prediction of the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt based on Φ(t), it can achieve advanced perception and early warning of potential operational risks, thereby adjusting control strategies in advance and effectively avoiding safety accidents such as solidification and excessive thermal stress.

[0017] Improved dynamic response speed: By employing feedforward control based on the sustainable output power SOP(t+Δt) at the future time t+Δt, the large inertia of the molten salt energy storage system can be overcome, and the required flow rate can be calculated and set in advance, significantly reducing the response lag to load changes. Combined with PI feedback control targeting the comprehensive operating coefficient Φ, the accuracy and robustness of the control are guaranteed, and the regulation lag caused by system inertia is greatly reduced.

[0018] Global operational economic optimization: Through periodic efficiency assessment and online optimization of multi-objective parameters, the system control model can evolve on its own, realize online optimization closed loop, and continuously find the highest efficiency operating point under the premise of ensuring safe and stable operation, thereby improving the overall economic efficiency of the system in the long term.

[0019] Strong adaptability and robustness: The complete closed loop consisting of four steps enables the system to have the ability of "perception-decision-execution-learning". The system can automatically adapt to complex situations such as molten salt aging, equipment performance degradation, and changes in external load characteristics, demonstrating strong adaptability and robustness. Attached Figure Description

[0020] Figure 1 A schematic flowchart of the molten salt energy storage system coordinated control method of the present invention. Detailed Implementation

[0021] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] The following describes the molten salt energy storage system collaborative control method and system according to embodiments of this application, with reference to the accompanying drawings.

[0024] The following is combined Figure 1This application provides a detailed description of the collaborative control method for molten salt energy storage systems. The method sequentially executes the following steps within a single control cycle to achieve logical progression and control closed loop.

[0025] Step S1: Couple the hydraulic characteristics, heat transfer characteristics, solidification risk and thermal stress risk of the molten salt energy storage system, calculate the comprehensive operating condition coefficient Φ(t) at the current time t, characterize the comprehensive operating state of the molten salt energy storage system at time t, and use it to quantify the degree to which the system deviates from the ideal operating state.

[0026] Specifically, the comprehensive operating condition coefficient As shown in Equation 1: Formula 1 In Equation 1, Let be the coefficient of friction of the pipe at time t; This is the reference friction coefficient for the molten salt energy storage system under design conditions; Let be the Nusselt number at time t. For reference Nusselt numbers; Let be the solidification risk factor at time t. The thermal stress sensitivity coefficient of the pipe material; Let t be the maximum temperature difference of the molten salt across the cross-section of the pipe.

[0027] Here, the parameters in Equation 1 are explained in detail: Φ(t): the comprehensive operating condition coefficient at the current time t, dimensionless, output quantity, characterizing the comprehensive operating state of the system at time t, the closer it is to 1, the closer it is to the ideal operating condition.

[0028] Let be the pipe friction coefficient at time t, which is dimensionless. It is calculated using the Colebrook-White Equation based on the real-time Reynolds number Re(t) and the relative roughness of the pipe ε / D. The Reynolds number Re(t) = ρ(t)·v(t)·D / μ(t) is given by the equation: ρ(t) is the density of the molten salt at the operating temperature at time t; v(t) is the fluid velocity of the molten salt at time t, measured by a flow meter; D is the inner diameter of the pipe (design value); and μ(t) is the dynamic viscosity of the molten salt at time t.

[0029] The friction coefficient is a dimensionless reference coefficient for the molten salt energy storage system under design conditions. It is the friction coefficient calculated under the system's design conditions (rated flow rate and design temperature) and is a fixed constant.

[0030] Nu(t) is the Nusselt number at time t, which is dimensionless and represents the actual convective heat transfer intensity. The calculation formula is Nu(t) = h(t) · D / k(t), where h(t) is the measured convective heat transfer coefficient (which can be calculated by measuring the heat flux density and temperature difference), D is the inner diameter of the pipe (design value), and k(t) is the thermal conductivity of the molten salt at the temperature at time t (look up the property table).

[0031] The reference Nusselt number, dimensionless, is the Nusselt number corresponding to the target heat transfer intensity required to maintain efficient heat transfer and prevent solidification of the system, determined by the design convective heat transfer coefficient h. target The calculation yields the following results: = h target · D / k design h target To design the convective heat transfer coefficient, k design To design the thermal conductivity.

[0032] Let be the solidification risk factor at time t, which is dimensionless. The calculation formula is: = exp(-k s ·(T(t) -T f ), where T(t) is the real-time temperature (measured by thermocouple) of a key point (such as the outlet of the cold salt tank), T f K represents the freezing point temperature (physical property constant) of molten salt. s The empirical sensitivity coefficient (e.g., 0.1 °C) - ¹).

[0033] The thermal stress sensitivity coefficient of the pipe material, with dimensions [Θ]. - [¹] This is related to the coefficient of linear expansion and modulus of elasticity of the pipe material (such as 316L stainless steel), which can be found in the material handbook, for example, 1.2 × 10⁻⁶. -5 °C - ¹.

[0034] The maximum temperature difference of the molten salt on the cross-section of the pipe at time t is given by the dimension [Θ], which is obtained by measuring the difference through multiple thermocouples placed at different positions on the same cross-section of the pipe.

[0035] The origin and derivation of Formula 1 above is as follows: This invention creatively integrates the classical dimensionless criterion numbers of fluid mechanics and heat transfer. The original basis is the friction coefficient f (derived from the Darcy-Weisbach formula) and the Nusselt number Nu (derived from convective heat transfer criterion formulas, such as the Dietus-Belth formula). To couple the characterization of hydraulic and heat transfer states, this invention correlates the ratio of the actual friction coefficient to the reference value and the ratio of the actual heat transfer intensity (characterized by Nu) to the target value, and introduces an exponent of 1.8 to strengthen the weight of the heat transfer state in the comprehensive evaluation (this exponent originates from the square relationship of the Reynolds number exponent 0.8 in the typical turbulent heat transfer formula, and its influence has been amplified through engineering verification). Furthermore, the factor R, which characterizes the solidification risk, is... f Factors related to thermal stress risk ΔTgrad, as a multiplicative correction term, ultimately yields the comprehensive quantitative index Φ.

[0036] Step S2: Using the comprehensive operating condition coefficient Φ(t) as the core input, and based on the autoregressive exogenous model, predict the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt.

[0037] Specifically, the sustainable output power of the molten salt energy storage system at the future time t+Δt. As shown in Equation 2: Formula 2 In Equation 2, Let be the measured sustainable output power at time t; α, β, γ, and δ are the weighting coefficients of the autoregressive exogenous model. The comprehensive operating condition coefficient at time t; Let t be the power demand of the external load. This refers to the system's rated output power. Let t be the inlet temperature of the molten salt energy storage system. , These represent the minimum and maximum permissible operating temperatures of the molten salt energy storage system, respectively; τ is the thermal inertia time constant of the molten salt energy storage system. Let be the rate of change of external load demand power at time t.

[0038] Here, the parameters in Equation 2 are explained in detail. SOP(t+Δt) is the sustainable output power of the molten salt energy storage system at the future time t+Δt, with the dimension of [power] and the unit of W, representing the output quantity.

[0039] The measured continuous output power at time t is expressed in units of [power] and W. It is measured by a power sensor at the grid connection point.

[0040] α, β, γ, and δ are the weighting coefficients of the autoregressive exogenous model, and are dimensionless. They are calculated using historical operating data (SOP, Φ, P) from at least one complete quarter. demand , T in The parameters are obtained by least squares system identification of time series data, and are unique parameters obtained by the method of this application for specific systems.

[0041] The external load demand power at time t is expressed in units of [power] and W. This power is derived from power grid dispatch instructions or the host computer.

[0042] This is the system's rated output power, measured in units of [power] and W. It is a fixed value designed for this purpose.

[0043] Let t be the inlet temperature (usually the outlet temperature of the cold salt tank) of the molten salt energy storage system at time t, with dimensions [Θ] and units of °C. Measured by thermocouple.

[0044] , These are the minimum and maximum permissible operating temperatures of the molten salt, respectively, with dimensions [Θ] and units of °C. They are determined by the chemical stability of the molten salt and the equipment design.

[0045] τ is the system's thermal inertia time constant, with dimensions of [time] and units of s. It is obtained through step response testing or estimated based on tank capacity and molten salt flow rate, for example, 1800 s.

[0046] Let be the rate of change of external load demand power at time t, with the dimension [power / time] and unit W / s. (via...) - (t-Δt s )) / Δt s Calculate Δt s The sampling period.

[0047] The source and derivation of Formula 2 above is as follows: This invention is based on the Autoregressive Exogenous Model (ARX) in system identification theory. The standard ARX model is in the form A(z)y(t) = B(z)u(t) + e(t). This invention uses SOP as the output y, and Φ, The model's weighting coefficients α, β, γ, and δ are obtained by training on historical operating data and fitting the model. The invention's innovation lies in using the comprehensive operating condition coefficient Φ as the core variable of the exogenous input, enabling the prediction model to perceive the dynamic impact of the system's internal health status on its output capability, rather than solely relying on external demands.

[0048] Step S3: Using the sustainable output power SOP(t+Δt) at the future time t+Δt as the feedforward setpoint, and combining it with the comprehensive operating condition coefficient Φ(t) for feedback correction, calculate the optimal molten salt volumetric flow rate F that should be set at the current time t. opt (t) enables adaptive molten salt flow rate control.

[0049] Specifically, the optimal molten salt volumetric flow rate F should be set at the current time t. opt (t) is shown in Equation 3: Formula 3 In Equation 3, For molten salt energy storage systems in the future The continuous output power at any given moment; Let be the density of the molten salt at time t at the operating temperature; Let be the isobaric specific heat capacity of the molten salt at time t; The rated temperature drop or rise of the molten salt set within the heat exchanger; K p For proportional gain; T i The integral time constant; To address the deviation in overall operating conditions The integral from the previous control cycle to the current time t; e(t) is the comprehensive operating condition deviation at time t. , Φ ideal Φ(t) is the ideal comprehensive working condition coefficient; Φ(t) is the comprehensive working condition coefficient at time t.

[0050] Here, the parameters in Equation 3 are explained in detail. The optimal molten salt volumetric flow rate setpoint calculated at time t, with dimensions of [volumetric flow rate] and units of m³ / s. The output is sent to the frequency converter of the molten salt pump.

[0051] For molten salt energy storage systems in the future The continuous output power at any given time, measured in units of [power] and W.

[0052] Let be the density of the molten salt at time t under the operating temperature, with dimensions of [mass / volume] and units of kg / m³. This density is obtained by referring to the molten salt property fitting formula or table based on the real-time temperature.

[0053] Let be the isobaric specific heat capacity of the molten salt at time t, with dimensions [mass / volume] and units of kg / m³. This can be obtained by referring to the fitting formula or table of molten salt properties based on the real-time temperature.

[0054] The rated temperature drop or rise of the molten salt set within the heat exchanger, dimensioned as [Θ], unit °C. Design fixed value.

[0055] Kp This is the proportional gain, dimensionless. It is obtained through the Ziegler-Nichols tuning method or based on model simulation tuning.

[0056] T i Let K be the integration time constant, with dimensions of [time] and units of seconds (s). The tuning method is the same as for K. p .

[0057] e(t) is the overall operating condition deviation at time t, which is dimensionless.

[0058] Φ ideal This is the ideal comprehensive working condition coefficient, dimensionless. The theoretical optimal value is 1.

[0059] To address the deviation in overall operating conditions The integral from the previous control cycle to the current time t is expressed in units of time (s); t0 is the start time of the integration in the current control cycle, which is the end time of the previous control cycle. A time variable used exclusively for calculus operations; it has no independent physical meaning and is only used to characterize the integration interval. Any point in time within the range.

[0060] The origin and derivation of Formula 3 above is as follows: This formula is based on an innovative combination of the law of conservation of energy and classical proportional-integral (PI) control theory. The feedforward term is directly derived from the predicted power through Q = m·C. p The energy balance relationship of ΔT (m=ρ·F) is derived to ensure rapid response. The feedback term adopts the form of a standard PI controller, but its control objective is to stabilize the comprehensive operating condition coefficient Φ calculated by Equation 1 at the ideal value Φ. ideal It uses nearby, rather than traditional temperature or pressure, to achieve safe closed-loop control of the overall system status.

[0061] Step S4: After a complete charge-discharge cycle, calculate the actual cycle efficiency of the molten salt energy storage system. A multi-objective optimization function is constructed; the parameters of the autoregressive exogenous model are updated online using the multi-objective optimization function as the objective; the updated autoregressive exogenous model parameters are used in step S2 to predict the next cycle, forming closed-loop control.

[0062] Specifically, the actual cycle efficiency η actual As shown in Equation 4: Formula 4 In Equation 4, This represents the total actual output energy within a single cycle. P represents the real-time power consumption of the molten salt pump. trace (τ) represents the real-time power consumption of the electric heat tracing system; Q loss This represents the static heat loss of the system during the cycle.

[0063] Here, the parameters in Equation 4 are explained in detail. The actual total output energy within one cycle, expressed in units of energy (J). This represents the measured SOP from the power sensor. real Perform time integration.

[0064] This represents the real-time power consumption of the molten salt pump, measured in units of [power] (W). It is obtained from a power meter for the pump drive motor.

[0065] P trace (τ) represents the real-time power consumption of the electric heat tracing system, measured in units of [power] and W. It is obtained from the power meter of the heat tracing circuit.

[0066] Q loss This represents the static heat loss of the system during the cycle, with the dimension [energy] and unit J. It is estimated based on insulation design, average ambient temperature, and empirical formulas, and can be considered a constant.

[0067] In a specific embodiment, the online updated autoregressive exogenous model parameters include weight coefficients α, β, γ, and δ, and the corresponding optimization objective is:

[0068] Where ηactual is the actual loop efficiency of the system; ηdesign is the design loop efficiency of the system; and λ1, λ2, and λ3 are the multi-objective optimization weight coefficients. The mean square error between the predicted power sequence SOPpred and the actual power sequence SOPreal in this cycle is given by Equation 2. The rated output power of the system is denoted as Var(Φ); Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle.

[0069] Specifically, this means finding an optimal set of parameters (α, β, γ, and δ) that minimizes the weighted sum of the following three objectives: First objective: Efficiency deviation term

[0070] Meaning: This is an evaluation of the system; it measures the actual operating efficiency of the system. ) and ideal design efficiency ( The smaller the gap, the more efficient the system is and the less energy is wasted.

[0071] λ1: This is a weighting coefficient that represents the degree of importance attached to "high efficiency" during optimization.

[0072] Second objective: Prediction error term

[0073] Meaning: "IQ" measures the accuracy of the model we previously used to predict power (the ARX model in step S2). ) and actual power ( The mean square error (MSE) of the power is calculated and normalized by dividing it by the square of the rated power to obtain a dimensionless error value. The smaller this value, the more accurate the prediction model and the more reliable the basis for control decisions.

[0074] λ2: Represents the degree of importance attached to "prediction accuracy".

[0075] Third objective: operational stability

[0076] Meaning: This is an assessment of the system's safety and stability. Φ(t) is the "comprehensive operating condition coefficient," reflecting the system's internal thermal, hydraulic, and risk conditions. If Φ(t) fluctuates drastically, it indicates that the system's internal state is unstable and may pose safety hazards. Therefore, the variance (Var(Φ)) of Φ(t) within a cycle is calculated; the smaller the variance, the more stable and safer the system operation.

[0077] λ3: Represents the degree of importance attached to "operational stability".

[0078] The source and derivation of Formula 4 above is as follows: The above formula is based on the efficiency definition of the first law of thermodynamics. The corresponding optimization objective is based on multi-objective optimization theory. This invention creatively uses efficiency deviation, prediction error (normalization), and operational stability (variance of Φ) as optimization objectives. It uses optimization algorithms such as stochastic gradient descent or particle swarm optimization to iteratively optimize the parameters of Formula 2, enabling the system to adaptively learn and approach the global optimal operating point.

[0079] Specifically, the parameters of the autoregressive exogenous model are updated online with a multi-objective optimization function as the objective, including: Efficiency bias, prediction error, and operational stability are used as optimization objectives. An optimization objective function is constructed, and the weight coefficients α, β, γ, and δ of the autoregressive exogenous model are iteratively optimized using stochastic gradient descent or particle swarm optimization to achieve online updating of the autoregressive exogenous model parameters. The optimization objective function is shown in Equation 5. Formula 5 In Equation 5, η actual η represents the actual cycle efficiency of the system. design The system is designed to improve loop efficiency; λ1, λ2, and λ3 are the multi-objective optimization weight coefficients. The predicted power sequence SOP in this cycle is given by Equation 2. pred With actual power sequence SOPreal The mean square error is denoted as Φ; Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle.

[0080] Here, the parameters in Equation 5 are explained in detail, η actual The actual cycle efficiency of the system is dimensionless and represents the output quantity.

[0081] η design The system design cycle efficiency is dimensionless and given in the design document.

[0082] λ1, λ2, and λ3 are dimensionless weighting coefficients for multi-objective optimization. They are set by the operator; for example, λ1=0.6, λ2=0.3, and λ3=0.1 represent different emphases on efficiency, prediction accuracy, and stationarity.

[0083] The predicted power sequence SOP in this cycle is given by Equation 2. pred With actual power sequence SOP real The mean square error is expressed in units of [power²] and W².

[0084] P rated : System rated power, used to make MSE dimensionless.

[0085] Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle, and is dimensionless. It is used to measure operational stability.

[0086] In another specific embodiment, the update process uses gradient descent for iterative optimization, and the specific steps are as follows: First, the weight coefficients α, β, γ, and δ of the autoregressive exogenous model in Equation 2 are used as the initial point parameters θ0, i.e., θ0 = [α0, β0, γ0, δ0]. T ; Secondly, calculate the optimization objective function. In the current parameter θ k The gradient vector is obtained by taking the partial derivative with respect to each weight coefficient. J(θ k ); Next, update the weight coefficients according to the following formula 6: θ (k+1) =θk-η learn J(θ k Formula 6 In Equation 6, η learn For efficient learning, θ is a positive decimal (e.g., 0.01) used to control the update step size. k For the current parameter, J(θ k) is the gradient vector.

[0087] Then, repeat steps two and three until the preset iteration termination condition is met, and obtain the weight coefficients of the final optimized autoregressive exogenous model. Finally, the weight coefficients of the final optimized autoregressive exogenous model are updated in Equation 2. That is, the final optimized parameter is θ. * = [α * , β * , γ * , δ * ] T Replace the old parameters in Formula 2 for power prediction in the next running cycle.

[0088] Through the above update process, the model parameters α, β, γ, δ can be adaptively adjusted according to the actual operating performance of the system (efficiency, prediction error, stability), so that the prediction model can continuously approach the true dynamic characteristics of the system, thereby achieving continuous performance optimization.

[0089] Here, the gradient can be calculated using automatic differentiation tools or the finite difference method. For example, for the parameter α, its partial derivative is approximately:

[0090] in, It is a very small positive number (such as 1e-6).

[0091] Specifically, the preset iteration termination conditions include the following termination conditions: Optimize objective function The change is less than the first preset threshold εJ; And / or, reaching the maximum number of iterations K max (e.g., 100 times); And / or, the norm of the gradient vector is less than a second preset threshold ε. (e.g., 1e-3).

[0092] Based on the above-mentioned synergistic control method for molten salt energy storage systems, and in conjunction with an embodiment of a molten salt energy storage system applied in a 50MWh solar thermal power plant, the present invention will be further explained.

[0093] In this embodiment, the molten salt is Solar Salt (60% NaNO3 + 40% KNO3), with a freezing point T. f =240°C, maximum allowable temperature T max =565°C, the pipe is made of 321 stainless steel, and the inner diameter D=0.3m.

[0094] Step 1: Initialization and Data Acquisition Before system startup, the initial parameters for Formula 2 are obtained through training using historical data: α=0.85, β=0.12, γ=0.05, δ=0.25. Δt=300s, τ=1800s, Φ ideal =1, K p =0.5, T i =600s, ΔT set =100°C. Install and calibrate all sensors: flow meter, pipeline multi-point thermocouple, tank temperature sensor, power transmitter.

[0095] Step 2: Real-time control loop (taking time t as an example) 1. Calculate Φ(t): The controller reads the flow velocity v(t) = 1.2 m / s at time t, the mainstream molten salt temperature T(t) = 285°C, the pipe wall temperature, and the cross-sectional temperature difference ΔT. grad (t) = 5°C, etc. From the property table, we get ρ(t) = 1900 kg / m³, μ(t) = 0.0032 Pa·s, k(t) = 0.52 W / (m·K). We calculate Re(t), and then use the Colebrook equation to calculate f(t) = 0.018. We then calculate h(t) in reverse, obtaining Nu(t) = 320. We take f... ref =0.017, Nu ref =350, calculate R. f (t) = 0.011. Let C be... ts =1.5×10 -5 °C - ¹, Substituting into Formula 1, we get Φ(t) = 0.92.

[0096] 2. Predict SOP(t+Δt): Read the current SOP(t) = 35 MW. =38 MW, =290°C, =250°C, =550°C, calculate t = 2 MW / min = 0.033 MW / s. Substituting the above data and Φ(t) = 0.92 into Formula 2, we calculate SOP(t+300s) = 36.8 MW.

[0097] 3. Regulate flow Given SOP(t+300s) = 36.8 MW and ρ(t) = 1900 kg / m³, C p Substituting (t) = 1550 J / (kg·°C) into the feedforward term of Formula 3, we obtain the basic flow rate. Calculate e(t) = 1 - 0.92 = 0.08, and combine this with K... p , T iCalculate the PI feedback correction factor. Finally, calculate... =0.65 m³ / s. This value is then sent to the molten salt pump frequency converter.

[0098] Step 3: Cycle optimization (after one charge-discharge cycle) After the cycle ends, the data recording system summarizes all the data for this cycle.

[0099] 1. Calculate actual efficiency: Apply formula 4 to calculate the actual efficiency of this cycle. Assume the total output energy obtained by integration is 1.08 × 10⁻⁶. 11 J, the total pump consumption and heat tracing energy consumption is 2.1×10 9 J, heat loss Q loss =5×10 8 J, then η actual = 0.978 (97.8%), Design efficiency η design =0.982.

[0100] 2. Calculate the optimization objective term: Calculate the mean square error (MSE) between the predicted power and the actual power in this cycle: MSE = 1.2 × 10⁻⁶ 6 W², P rated =50 MW, therefore the normalized prediction error term is 4.8 × 10 -7 The variance Var(Φ) of the current cyclic Φ sequence is calculated to be 0.002.

[0101] 3. Construct and optimize the objective function: Set weights λ1=0.6, λ2=0.3, λ3=0.1, and construct the objective function J. With the current parameters θ0= [0.85, 0.12, 0.05, 0.25] T Starting with η, we use gradient descent for optimization. Let the learning rate be η. learn =0.01, maximum number of iterations 100.

[0102] 4. Execution parameter update: First iteration: Calculate J(θ0) and gradient. J(θ0), update the parameters to get θ.

[0103] Repeat the iterations, assuming the termination condition (|ΔJ|<1e-4) is met after the 20th iteration, to obtain the optimized parameters θ. * =[0.872, 0.113, 0.048, 0.262] T .

[0104] Parameter replacement: Update Equation 2 with new parameters α=0.872, β=0.113, γ=0.048, δ=0.262 for power prediction in the next running cycle.

[0105] This completes a full "monitoring-prediction-control-optimization" cycle. The system will continue to run this cycle to achieve safe, fast, and efficient adaptive and coordinated regulation.

[0106] This application also provides a system for implementing the above-described molten salt energy storage system coordinated control method, comprising: The comprehensive operating condition coefficient Φ(t) acquisition module is used to couple the hydraulic characteristics, heat transfer characteristics, solidification risk and thermal stress risk of the molten salt energy storage system, calculate the comprehensive operating condition coefficient Φ(t) at the current time t, characterize the comprehensive operating state of the molten salt energy storage system at time t, and quantify the degree to which the system deviates from the ideal operating state. The module for obtaining the sustainable output power SOP(t+Δt) at the future time t+Δt is used to predict the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt based on an autoregressive exogenous model, with the comprehensive operating condition coefficient Φ(t) as the core input. The optimal molten salt volumetric flow rate F that should be set at present opt The (t) acquisition module is used to calculate the optimal molten salt volumetric flow rate F that should be set at the current time t, using the sustainable output power SOP(t+Δt) at a future time t+Δt as the feedforward setpoint and combined with the comprehensive operating condition coefficient Φ(t) for feedback correction. opt (t), to achieve adaptive molten salt flow rate control; The closed-loop control module is used to calculate the actual cycle efficiency of the molten salt energy storage system after a complete charge-discharge cycle. A multi-objective optimization function is constructed; the parameters of the autoregressive exogenous model are updated online using the multi-objective optimization function as the objective; the updated autoregressive exogenous model parameters are used in step S2 to predict the next cycle, forming closed-loop control.

[0107] Here, those skilled in the art will understand that the specific methods of the above-mentioned molten salt energy storage system collaborative control system have been described in detail in the description of the molten salt energy storage system collaborative control method in reference 1 above.

[0108] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. The scope of patent protection of the present invention shall be determined by the claims. Similarly, any equivalent structural changes made based on the description and drawings of the present invention shall also be included within the scope of protection of the present invention.

Claims

1. A method for coordinated control of a molten salt energy storage system, characterized in that, Includes the following steps: The method executes the following steps sequentially within a single control cycle to achieve logical progression and control closed loop: Step S1: Couple the hydraulic characteristics, heat transfer characteristics, solidification risk and thermal stress risk of the molten salt energy storage system, calculate the comprehensive operating condition coefficient Φ(t) at the current time t, characterize the comprehensive operating state of the molten salt energy storage system at time t, and use it to quantify the degree to which the system deviates from the ideal operating state; Step S2: Using the comprehensive operating condition coefficient Φ(t) as the core input, and based on the autoregressive exogenous model, predict the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt. Step S3: Using the sustainable output power SOP(t+Δt) at the future time t+Δt as the feedforward setpoint, and combining it with the comprehensive operating condition coefficient Φ(t) for feedback correction, calculate the optimal molten salt volumetric flow rate F that should be set at the current time t. opt (t), to achieve adaptive molten salt flow rate control; Step S4: After a complete charge-discharge cycle, calculate the actual cycle efficiency of the molten salt energy storage system. A multi-objective optimization function is constructed; the parameters of the autoregressive exogenous model are updated online using the multi-objective optimization function as the objective; the updated autoregressive exogenous model parameters are used in step S2 to predict the next cycle, forming closed-loop control.

2. The method for coordinated control of a molten salt energy storage system according to claim 1, characterized in that, In step S1, the comprehensive operating condition coefficient As shown in Equation 1: Formula 1 In Equation 1, Let be the coefficient of pipe friction at time t; This is the reference friction coefficient for the molten salt energy storage system under design conditions; Let be the Nusselt number at time t. For reference Nusselt numbers; Let be the solidification risk factor at time t. The thermal stress sensitivity coefficient of the pipe material; Let t be the maximum temperature difference of the molten salt across the cross-section of the pipe.

3. The method for coordinated control of a molten salt energy storage system according to claim 1, characterized in that, In step S2, the sustainable output power of the molten salt energy storage system at a future time t+Δt is... As shown in Equation 2: Formula 2 In Equation 2, Let be the measured sustainable output power at time t; α, β, γ, and δ are the weighting coefficients of the autoregressive exogenous model. The comprehensive operating condition coefficient at time t; Let t be the power demand of the external load. This refers to the system's rated output power. Let t be the inlet temperature of the molten salt energy storage system. , These represent the minimum and maximum permissible operating temperatures of the molten salt energy storage system, respectively; τ is the thermal inertia time constant of the molten salt energy storage system. Let be the rate of change of external load demand power at time t.

4. The method for coordinated control of a molten salt energy storage system according to claim 1, characterized in that, In step S3, the optimal molten salt volumetric flow rate F that should be set at the current time t is... opt (t) is shown in Equation 3: Formula 3 In Equation 3, For molten salt energy storage systems in the future The continuous output power at any given moment; Let be the density of the molten salt at time t at the operating temperature; Let be the isobaric specific heat capacity of the molten salt at time t; The rated temperature drop or rise of the molten salt set within the heat exchanger; K p For proportional gain; T i The integral time constant; To address the deviation in overall operating conditions The integral from the previous control cycle to the current time t; e(t) is the comprehensive operating condition deviation at time t. , Φ ideal Φ(t) is the ideal comprehensive working condition coefficient; Φ(t) is the comprehensive working condition coefficient at time t.

5. The method for coordinated control of a molten salt energy storage system according to claim 3, characterized in that, In step S4, the actual cycle efficiency η actual As shown in Equation 4: Formula 4 In Equation 4, The total actual output energy within a cycle; P represents the real-time power consumption of the molten salt pump. trace (τ) represents the real-time power consumption of the electric heat tracing system; Q loss This represents the static heat loss of the system during the cycle.

6. The method for coordinated control of a molten salt energy storage system according to claim 5, characterized in that, The parameters of the autoregressive exogenous model updated online include weight coefficients α, β, γ, and δ, and the corresponding optimization objective is: ； Where, η actual η represents the actual cycle efficiency of the system. design The system is designed for loop efficiency; λ1, λ2, and λ3 are multi-objective optimization weight coefficients; The predicted power sequence SOP in this cycle is given by Equation 2. pred With actual power sequence SOP real The mean square error; The rated output power of the system is denoted as Var(Φ); Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle.

7. The method for coordinated control of a molten salt energy storage system according to claim 6, characterized in that, The method of updating the parameters of the autoregressive exogenous model online with a multi-objective optimization function as the objective includes: Efficiency bias, prediction error, and operational stability are used as optimization objectives. An optimization objective function is constructed, and the weight coefficients α, β, γ, and δ of the autoregressive exogenous model are iteratively optimized using stochastic gradient descent or particle swarm optimization to achieve online updating of the autoregressive exogenous model parameters. The optimization objective function is shown in Equation 5. Formula 5 In Equation 5, η actual η represents the actual cycle efficiency of the system. design The system is designed for loop efficiency; λ1, λ2, and λ3 are multi-objective optimization weight coefficients; The predicted power sequence SOP in this cycle is given by Equation 2. pred With actual power sequence SOP real The mean square error; The rated output power of the system is denoted as Var(Φ); Var(Φ) is the variance of the comprehensive operating condition coefficient Φ sequence within this cycle.

8. The method for coordinated control of a molten salt energy storage system according to claim 7, characterized in that, The stochastic gradient descent method includes: First, the weight coefficients α, β, γ, and δ of the autoregressive exogenous model in Equation 2 are used as the initial point parameters θ0. Secondly, calculate the optimization objective function. In the current parameter θ k The gradient vector is obtained by taking the partial derivative with respect to each weight coefficient. J(θ k ); Next, update the weighting coefficients according to the following formula 6: i (k+1) =θ k -or learn J(θ k ) formula 6 In Equation 6, η learn Let θ be the learning rate. k For the current parameter, J(θ k ) represents the gradient vector; Then, repeat steps two and three until the preset iteration termination condition is met, and obtain the weight coefficients of the final optimized autoregressive exogenous model. Finally, the weight coefficients of the final optimized autoregressive exogenous model are updated in Equation 2.

9. The method for coordinated control of a molten salt energy storage system according to claim 8, characterized in that, The preset iteration termination conditions include the following termination conditions: Optimize objective function The change is less than the first preset threshold εJ; And / or, reaching the maximum number of iterations K max ; And / or, the norm of the gradient vector is less than a second preset threshold ε. .

10. A system for performing the coordinated control method of a molten salt energy storage system as described in any one of claims 1-9, characterized in that, include: The comprehensive operating condition coefficient Φ(t) acquisition module is used to couple the hydraulic characteristics, heat transfer characteristics, solidification risk and thermal stress risk of the molten salt energy storage system, calculate the comprehensive operating condition coefficient Φ(t) at the current time t, characterize the comprehensive operating state of the molten salt energy storage system at time t, and quantify the degree to which the system deviates from the ideal operating state. The module for obtaining the sustainable output power SOP(t+Δt) at the future time t+Δt is used to predict the sustainable output power SOP(t+Δt) of the molten salt energy storage system at the future time t+Δt based on an autoregressive exogenous model, with the comprehensive operating condition coefficient Φ(t) as the core input. The optimal molten salt volumetric flow rate F that should be set at present opt The (t) acquisition module is used to calculate the optimal molten salt volumetric flow rate F that should be set at the current time t, using the sustainable output power SOP(t+Δt) at a future time t+Δt as the feedforward setpoint and combined with the comprehensive operating condition coefficient Φ(t) for feedback correction. opt (t), to achieve adaptive molten salt flow rate control; The closed-loop control module is used to calculate the actual cycle efficiency of the molten salt energy storage system after a complete charge-discharge cycle. A multi-objective optimization function is constructed; the parameters of the autoregressive exogenous model are updated online using the multi-objective optimization function as the objective; the updated autoregressive exogenous model parameters are used in step S2 to predict the next cycle, forming closed-loop control.

Images