Control method, system, equipment and medium for participation of regional refrigeration system in secondary frequency modulation of power grid

Abstract

The invention discloses a control method, system, equipment and medium for a regional refrigeration system to participate in secondary frequency modulation of a power grid, and belongs to the technical field of power systems, and the method comprises the steps: constructing a system state model for describing the temperatures and cooling water flows of a plurality of regions; predicting the operation state of each region based on a state estimation unit; identifying a sensor fault through an anomaly detection module; under the fault condition, the observation value is replaced based on the prediction state to execute control, and state deviation judgment is carried out; when the deviation exceeds a set threshold value, control strategy switching is executed; and after the state of the fault area recovers and meets the confidence condition, frequency modulation control is accessed again. By combining multivariable Kalman filtering and deep learning, state estimation and adjustment control under an incomplete observation condition are realized, the robustness and fault-tolerant capability of a control system in a fault scene are improved, and the method is suitable for a high-reliability operation scene in which a multi-region cooling system participates in power grid auxiliary service.

Description

Technical Field

[0001] This invention relates to the field of power system technology, specifically to a control method, system, equipment, and medium for a regional cooling system to participate in the secondary frequency regulation of the power grid. Background Technology

[0002] In large building complexes, data centers, and district energy systems, district cooling systems are widely used for temperature control in multiple spatial areas. With increasing demands for grid regulation, district cooling systems are increasingly being incorporated into grid ancillary services, particularly secondary frequency control tasks. This type of control strategy, through adjusting cooling water flow and other means, enables the cooling system to meet indoor temperature control requirements while responding to grid frequency shifts, thereby improving the overall load flexibility and energy efficiency of the system.

[0003] Most existing studies assume an ideal system with complete data acquisition and properly functioning sensors in each zone, and construct temperature control and frequency modulation strategies accordingly. However, in actual engineering operations, zone cooling systems are highly susceptible to problems such as sensor failure, data delay, or observation errors. For example, a temperature sensor going offline will cause the regulation process to become open-loop control, losing the temperature feedback signal; data transmission delays or errors will directly damage control accuracy and reduce system stability. Especially in scenarios with strong multi-zone coupling and sensitive regulation time, these problems can lead to drastic fluctuations in zone temperature and even cause unstable operation of the control system.

[0004] Current technologies typically handle such faults conservatively, such as directly removing the affected area from secondary frequency regulation and restoring it to its basic cooling flow state. However, this approach results in a loss of regulation capacity, reducing the overall system's flexibility and response efficiency in grid interaction. Furthermore, there is currently no mature solution to maintain stable and reliable secondary frequency regulation control of the district cooling system in the event of sensor failure or data loss.

[0005] Therefore, there is an urgent need for a control method and system with higher robustness and fault tolerance, which can accurately estimate and stably control the regional temperature state under conditions of incomplete, delayed or abnormal sensor data, and ensure that the regional cooling system still has the ability to participate in grid interaction frequency regulation efficiently under non-ideal information conditions. Summary of the Invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the technical problem solved by this invention is: how to accurately estimate the system state and achieve stable and reliable secondary frequency regulation control of the power grid in the event that some sensors in a district cooling system are disconnected, data is delayed, or acquisition is incorrect, thereby improving the fault tolerance and control robustness of the system.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a control method for a district cooling system participating in the secondary frequency regulation of the power grid, comprising,

[0009] Construct a system state model to describe the temperature and cooling water flow rate in multiple zones;

[0010] Based on the system state model, the operating state of each region is predicted by the state estimation unit;

[0011] Within each control cycle, the anomaly detection module identifies whether there is a regional sensor malfunction based on the difference between the current observation data and the predicted state.

[0012] When a fault is identified in a certain area, control operations are performed based on the predicted state to replace the observed data, and the state deviation of the faulty area is continuously calculated.

[0013] If the state deviation exceeds the set threshold range, a control strategy switching operation is performed on the faulty area;

[0014] After determining that the state of the faulty area has recovered to the target temperature range and meets the preset signal conditions, the faulty area is re-involved in the secondary frequency regulation control process of the power grid.

[0015] As a preferred embodiment of the control method for a regional cooling system to participate in the secondary frequency regulation of the power grid as described in this invention, the state estimation unit constructs a state space model based on the system state model. The state space model includes temperature state variables and cooling water flow state variables of multiple regions, and the model parameters include thermal coupling coefficient and flow influence coefficient between regions.

[0016] In cases where observation data is missing in some areas, a recursive state estimation process is performed using the state transition matrix and the observation matrix to output a complete state estimate for control execution.

[0017] As a preferred embodiment of the control method for a regional cooling system participating in the secondary frequency regulation of the power grid as described in this invention, the anomaly detection module collects the state estimation residuals within multiple consecutive control cycles and constructs a sliding time window sequence.

[0018] The residual sequence within the time window is used as input features and input into the trained neural network model for analysis.

[0019] Based on the output of the neural network model, it is determined whether there are any abnormal states related to the regional temperature sensor during the control cycle, including data loss, data delay, or sudden changes in observation error.

[0020] As a preferred embodiment of the control method for a regional cooling system participating in the secondary frequency regulation of the power grid according to the present invention, the method for determining whether the state deviation exceeds the set range includes: squaring the state estimation residual within the current control cycle and accumulating it to form an error index.

[0021] The error index is compared with the error threshold set by the system;

[0022] If the cumulative error index over multiple consecutive control cycles exceeds the error threshold, a control strategy switching signal is generated to initiate subsequent adjustment operations.

[0023] As a preferred embodiment of the control method for a regional cooling system to participate in the secondary frequency regulation of the power grid as described in this invention, the control strategy switching operation includes: adjusting the cooling water flow rate of the fault area to the basic flow rate value of the initial setting stage of the fault area;

[0024] Remove the faulty area from the current secondary frequency modulation control task and stop issuing frequency modulation adjustment commands to it;

[0025] The base temperature maintenance control mode for the fault area is enabled, which is used only to maintain the local ambient temperature stability of the fault area.

[0026] As a preferred embodiment of the control method for a regional cooling system participating in the secondary frequency regulation of the power grid as described in this invention, after the control strategy switching operation is performed, a temperature response model of the fault area is constructed based on the historical operating data of the system. The temperature response model is used to characterize the temperature change process of the fault area over time under the condition of stable cooling water flow.

[0027] The temperature response model is used to predict the temperature of the fault area, and the prediction result is compared with the target temperature range to determine whether the state of the fault area has returned to stability.

[0028] As a preferred embodiment of the control method for a regional cooling system participating in the secondary frequency regulation of the power grid as described in this invention, wherein: after the prediction results of the temperature response model indicate that the temperature of the fault area is close to the target temperature range, the actual observation confirmation step is performed.

[0029] During multiple consecutive control cycles, the actual temperature of the fault area is collected and compared with the set temperature tolerance range;

[0030] Once the temperature in the faulty area stabilizes within the tolerance range for a preset time threshold, its control state is automatically switched, and the faulty area is restored to participate in the secondary frequency regulation control process of the power grid.

[0031] Another objective of this invention is to provide a control system for a district cooling system to participate in secondary frequency regulation of the power grid.

[0032] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a control system for a regional cooling system participating in the secondary frequency regulation of the power grid, comprising: a system state modeling module, used to construct a system state model describing the temperature and cooling water flow of multiple regions, so as to reflect the dynamic interaction between the regions;

[0033] The state estimation module is used to predict the operating state of each region based on the system state model, and to estimate the missing state when monitoring data is missing or abnormal.

[0034] An anomaly detection module is used to identify whether there is a regional sensor fault based on the difference between the current observation data and the predicted state in each control cycle.

[0035] The control execution module is used to perform control operations to replace the observed data based on the predicted state when a fault is detected in a certain area, and to continuously calculate the state deviation of the fault area.

[0036] The strategy switching module is used to perform a control strategy switching operation on the fault area when the state deviation exceeds a set threshold range.

[0037] The state recovery judgment module is used to re-enable the fault area to participate in the secondary frequency regulation control process of the power grid when the state of the fault area recovers to the target temperature range and meets the preset information conditions.

[0038] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the control method for a regional cooling system to participate in the secondary frequency regulation of the power grid.

[0039] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the control method for a district cooling system participating in the secondary frequency regulation of the power grid.

[0040] The beneficial effects of this invention are as follows: In the control process of a district cooling system participating in the secondary frequency regulation of the power grid, it solves the problem of system control instability caused by temperature sensor disconnection, delay, or data error. By combining multivariate Kalman filtering and deep learning techniques, robust estimation of temperature and cooling flow status in multiple areas is achieved, maintaining the system's prediction accuracy and response capability even with incomplete observation data.

[0041] This invention designs an anomaly detection mechanism based on residual sequence analysis, which can promptly identify insignificant faults and improve the system's ability to judge sensor anomalies. A control strategy switching mechanism and confidence criteria are introduced to ensure that faulty regions automatically exit frequency modulation tasks when their state becomes uncontrollable, while avoiding regulation losses caused by premature switching. For regions that have exited, the system establishes a temperature recovery model based on historical operating data and automatically reintegrates them into frequency modulation control after confirming temperature stability.

[0042] This invention enhances the fault tolerance and continuity of multi-region systems under fault scenarios by modeling the thermal coupling relationship between regions and establishing a state compensation mechanism. It ensures the stable operation of frequency regulation control and energy regulation efficiency, and is suitable for high-reliability operation scenarios of large-scale multi-region cooling systems. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 The above is a flowchart of a control method for a regional cooling system to participate in the secondary frequency regulation of the power grid, provided as an embodiment of the present invention.

[0045] Figure 2 This is a temperature comparison chart of different regions over time, illustrating a control method for a regional cooling system participating in secondary frequency regulation of the power grid, provided as an embodiment of the present invention.

[0046] Figure 3 The experimental comparison diagrams for scenario 1 and scenario 2 of a control method for a regional cooling system participating in the secondary frequency regulation of the power grid, provided as an embodiment of the present invention.

[0047] Figure 4 This is another experimental comparison diagram of scenario 1 and scenario 2 of a control method for a regional cooling system participating in the secondary frequency regulation of the power grid, provided as an embodiment of the present invention. Detailed Implementation

[0048] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0049] Example 1, referring to Figures 1-4This is one embodiment of the present invention, which provides a control method for a district cooling system to participate in the secondary frequency regulation of the power grid, comprising:

[0050] Step 101: Construct a system state model to describe the temperature and cooling water flow rate in multiple regions, so as to reflect the dynamic interaction between the regions;

[0051] Step 102: Based on the system state model, the operating state of each region is predicted by the state estimation unit. The prediction includes the estimation of the missing state in the case of missing or abnormal monitoring data.

[0052] Step 103: Within each control cycle, based on the difference between the current observation data and the predicted state, the anomaly detection module identifies whether there is a regional sensor fault;

[0053] Step 104: When a fault is identified in a certain area, control operations are performed based on the predicted state to replace the observed data, and the state deviation of the faulty area is continuously calculated;

[0054] Step 105: If the state deviation exceeds the set threshold range, then perform a control strategy switching operation on the faulty area;

[0055] Step 106: After determining that the state of the fault area has recovered to the target temperature range and meets the preset signal conditions, the fault area is re-involved in the power grid secondary frequency regulation control process.

[0056] It should be noted that in large-scale district cooling systems, maintaining temperature control across multiple zones is crucial for ensuring energy efficiency and user comfort. Such systems typically rely on extensive networks of temperature sensors and control mechanisms, such as adjusting cooling water flow rates in response to environmental changes and varying heat loads.

[0057] However, due to sensor malfunctions, data transmission errors, and complex interdependencies between zones, district cooling systems face significant challenges in maintaining system stability and control accuracy. These issues can lead to decreased control precision, resulting in energy waste or reduced zone comfort.

[0058] This embodiment proposes a multivariate Kalman filter technique that provides robust state estimation in dynamic and uncertain environments. Unlike traditional univariate feedback control or static models, multivariate Kalman filtering can simultaneously consider multiple interdependent state variables, such as temperature and cooling water flow rate in multiple zones. This enables the system to more effectively predict, estimate, and control temperature, maintaining control accuracy and stability even in the event of sensor data errors or missing data.

[0059] The application of Kalman filtering in district cooling systems is particularly suitable for scenarios where sensor data is unreliable due to faults, temporary interruptions, or noise. In these cases, the Kalman filter predicts the future state of the system and estimates sensor data by integrating historical data, physical models, and inter-regional dependencies. When a sensor in a certain region fails or provides erroneous data, the Kalman filter can utilize temperature data and historical flow information from adjacent regions to make reasonable state estimates, ensuring that the control system maintains predictive accuracy and control robustness even under incomplete or inaccurate measurement conditions.

[0060] Furthermore, complex thermal interactions typically exist between zones in a district cooling system. The temperature of a zone is not only controlled by its own temperature but also influenced by neighboring zones, piping systems, and external environmental factors. Kalman filtering can capture these dependencies between zones and incorporate them into a state-space model, thereby helping the system make more accurate cooling flow regulation decisions. For example, when a temperature sensor in a zone fails, Kalman filtering can use temperature data from neighboring zones to estimate the temperature of that zone, thus preventing a decline in control performance.

[0061] The adaptive nature of Kalman filtering also allows the system to dynamically adjust the control strategy based on real-time operating conditions. The process noise covariance (Q) and measurement noise covariance (R) can be dynamically adjusted according to the current operating conditions, ensuring that the system can cope with both short-term sensor outages (such as temporary sensor disconnection) and long-term changes (such as sensor drift or permanent failure). This makes Kalman filtering particularly robust in large-scale multi-zone refrigeration systems, providing excellent performance in fault tolerance and adaptability.

[0062] Kalman filtering is used to estimate the temperature and cooling flow rate of each region when sensor data is incomplete or unreliable.

[0063] By considering the thermal interactions between regions, control accuracy can be improved even in the event of local sensor failure. Multivariable Kalman filtering can simultaneously account for the interactions between multiple regions. For example, the temperature change in region 1 is not only affected by its own flow rate but may also be influenced by temperature feedback from regions 2 and 3. By incorporating these variables into the state equation, the system can more accurately predict temperature changes in each region.

[0064] A dynamic adjustment mechanism for the covariance of process noise and measurement noise is introduced to cope with different operating conditions, thereby enhancing the robustness of control.

[0065] The effectiveness of Kalman filtering is demonstrated through detailed simulation of a multi-zone refrigeration system, and compared with traditional control methods under various sensor failure scenarios.

[0066] When a sensor in one area goes offline, the system can continue to rely on observation data from other areas, maintaining predictions for the offline area through the synergistic effect of a multivariate Kalman filter. This fault tolerance is crucial for complex zone cooling systems.

[0067] In an optional embodiment of the present invention: the state estimation unit constructs a state-space model using a recursive calculation method based on a first-order differential state transition model. It represents the coupling relationship between the temperature state variables and the corresponding cooling water flow rate in each region of the system in matrix form, and iteratively calculates the predicted state within the control cycle. In typical applications, this method is suitable for cooling systems in small to medium-sized office buildings with weak inter-regional thermal coupling and relatively uniform sensor configuration, and can maintain basic predictive capabilities even when some sensors are temporarily offline.

[0068] In a preferred embodiment of the present invention: the state estimation unit constructs a state-space model based on a multivariate Kalman filter, integrates historical operating data, physical coupling modeling parameters, and currently available observation data, and performs recursive state estimation. This Kalman filter dynamically adjusts the prediction covariance and the observation covariance to adapt to changes in operating conditions such as missing regional sensor observations, data delays, or abnormal temperature fluctuations, making it suitable for campus-level or data center-level regional cooling systems.

[0069] The beneficial effects of this preferred technical solution are as follows: Compared with traditional static prediction or univariate estimation methods, this solution can significantly improve prediction accuracy and robustness in situations where multiple regions have complex conditions and sensors frequently observe abnormalities, ensuring that the control algorithm has fault tolerance and operational continuity, thereby maintaining the power grid's secondary frequency regulation response capability even under undesirable information conditions.

[0070] In an optional embodiment of the present invention: the anomaly detection module employs a fixed-window average residual analysis method, recording the residual between the state estimate and the observed value within each control cycle, and statistically analyzing the mean and standard deviation of the residual within a time window. When the absolute value of the residual exceeds twice the standard deviation, an anomaly is determined to exist in the observation for that cycle. In the cooling system of a public building with relatively stable temperature fluctuations, this method can initially achieve anomaly identification.

[0071] In a preferred embodiment of the present invention: the anomaly detection module constructs a state estimation residual sequence within a sliding time window and uses it as input features to input into a trained neural network model for discrimination and identification. The neural network model is trained under supervision using historical labeled data to identify anomaly patterns including data loss, delay, and abrupt changes in observation errors, and is suitable for large-scale regional cooling systems with strong multi-region thermal coupling and drastic signal fluctuations.

[0072] The beneficial effects of this preferred technical solution are as follows: the residual identification scheme based on deep learning has stronger nonlinear fitting ability and fault mode identification ability compared with traditional statistical methods. In particular, it can effectively improve the fault identification response speed, significantly reduce false alarm and false alarm rates, and improve the stability and reliability of frequency modulation control process in regional systems with large thermal inertia.

[0073] Example 2, an embodiment of the present invention, provides a control method for a district cooling system to participate in the secondary frequency regulation of the power grid, based on the previous embodiment, including:

[0074] First, the state vector and observation vector of the district cooling system are determined. In multivariate Kalman filtering, the state vector includes the temperature and flow rate of multiple regions, as well as feedback variables from adjacent regions. The system state is extended to multiple regions in combination. Kalman filtering is performed in conjunction with the physical characteristics of the district cooling system, mainly to enable more precise temperature observation and prediction of each region of the district cooling system.

[0075] Specifically, the state vector x(t) represents the state of the zone cooling system, which can include the temperature and flow rate of each zone. When the system is divided into n zones for regulation and control, the state vector matrix can be represented as:

[0076]

[0077] Among them, T i (t) is the temperature of the i-th region, q i (t) represents the flow rate in this region. The observation vector y(t) represents the sensor observations from n regions.

[0078]

[0079] Because some sensors in the system may be offline, resulting in data loss, when the temperature data for the i-th region is missing, the corresponding T... i (t) represents NaN, which means empty, indicating that there is no data here. represents NtaNumber (not a number), which is used to indicate missing or invalid data.

[0080] Before fault diagnosis, the raw sensor data needs to be preprocessed to ensure the quality and format of the data input into the deep learning model.

[0081] In an optional embodiment of the invention, data synchronization and alignment are performed to ensure that data from different sensors (such as temperature and flow rate) are time-aligned. If sensor delays exist, interpolation methods are used for time alignment.

[0082] y aligned(k)=Interpolate(y(k),timestamps)

[0083] In the formula: the input is the original observation vector y(k), which may contain asynchronous data with different timestamps; timestamps: timestamps representing the time series aligned to the target (e.g., uniformly spaced). The output y aligned (k): The aligned observation vector, where all data points are located at a unified timestamp. Interpolate indicates that interpolation is used. In data alignment and interpolation, "Interplate" refers to estimating the value of an unknown time point using known discrete data points to solve the problem of asynchronous or missing sensor data.

[0084] In an optional embodiment of the present invention, the data is normalized to make it range between [0,1] or [-1,1] in order to speed up the training process and improve model performance.

[0085]

[0086] Where y(k) represents the observed value of the raw sensor data (such as temperature, flow rate) at time step. min This represents the minimum value in the data sequence. max This represents the maximum value in the data sequence. norm (k): The normalized value that satisfies 0 ≤ y norm (k)≤1.

[0087] In an optional embodiment of the present invention, time series data is divided into fixed-length windows using windowing to capture time series features. Assuming the window size is n, each input sample can be represented as:

[0088] Z(k)={y norm (k-n+1),y norm (k-n+2),…,y norm (k)}

[0089] Among them, y norm Z(k) represents the normalized observation (e.g., temperature, flow rate) at time step k. n represents the window length (number of historical time steps). Z(k) represents the window vector constructed at time step k, containing n data points from k-n+1 to k.

[0090] Furthermore, by combining the state estimation with the Kalman filter, the system state difference is extracted as the input feature of the deep learning model.

[0091] State estimation difference:

[0092]

[0093] Here, e(k) represents the residual, which is the difference between the observed value at time k and the state estimate obtained through prediction. It represents the deviation between the observation and the prediction result. Kalman filtering updates the system state estimate by continuously correcting this residual. y(k) represents the actual observed value, that is, the data obtained from the sensor at time k, such as temperature, flow rate, etc. C represents the observation matrix, which maps the state vector from the state space to the observation space. It defines which state variables affect the observed values ​​and is usually a matrix used to represent the relationship between the observed variables and the state variables. The predicted state estimate represents the predicted value of the system state at time k. This estimate is obtained based on the data at time k-1 and the state transition equation obtained by Kalman filtering, and is usually called the "prior estimate".

[0094] Input features: Combine the state differences from the past m time steps into a feature vector:

[0095] Z(k)=[e(k-m+1),e(k-m+2),…,e(k)]

[0096] This formula represents the residual sequence Z(k), which is a vector of multiple residual values ​​at time k.

[0097] In multivariable Kalman filtering, the evolution of the system is described by state equations and observation equations:

[0098] The state equation is defined as follows:

[0099] x(t+1)=Ax(t)+Bu(t)+w(t)

[0100] Here, A represents the system's state transition matrix, describing the interactions between different regions. The temperature of one region is affected by the temperatures of its neighboring regions; therefore, A encompasses the relationships between the regions.

[0101] B represents the control input matrix, which indicates the effect of flow rate u(t) on temperature.

[0102] w(t) represents process noise, reflecting the uncertainty of the model.

[0103] The flow rate u(t) is specifically represented as:

[0104]

[0105] Where q1(t), q2(t), ..., q n (t) represent the cold water flow rates from region 1 to region n.

[0106] Furthermore,

[0107] The state transition matrix A reflects how the system transitions from the current moment to the next moment when there is no control input. For a district cooling system, the state transition matrix represents how the temperature of each district changes over time, while also taking into account the mutual corroboration of temperature changes in adjacent districts.

[0108] The physical characteristics of a district cooling system are well-defined, and the A matrix can be obtained through physical modeling. For a district cooling system, temperature changes can be modeled using the following physical relationships:

[0109] Heat balance equation: According to the laws of thermodynamics, the temperature change of a system is related to parameters such as heat capacity and heat transfer coefficient. For example, for each region i, the temperature change can be expressed as:

[0110]

[0111] Where, α i It is the heat transfer coefficient, β i T is the flow coefficient. out It is the outside temperature, q i This is the cold water flow rate. Based on this type of equation, the discrete-time state transition matrix A can be derived.

[0112] The control input matrix B reflects how control inputs (such as cold water flow rate) affect the temperature of each zone.

[0113]

[0114] Where: β ij This represents the effect of the cold water flow rate in region j on the temperature in region i.

[0115] Thermodynamic models can be used to determine the effect of cold water flow rate on temperature. Typically, this relationship can be modeled using heat transfer equations (such as heat conduction, convection, etc.). Cold water flow rate q i The effect on temperature changes is relatively clear, and each element in the B matrix can be directly calculated using physical formulas.

[0116] In a multivariate Kalman filter model, w(t) represents process noise, which is typically used to describe uncertainties in the model and disturbances that are not accurately modeled in actual operation. For example, there may be external factors affecting temperature changes in the system that are not fully captured by the model, and w(t) is used to represent these random disturbances.

[0117] Specifically, w(t) is a vector representing noise among multiple state variables. In the application of district cooling systems, process noise w(t) includes: the influence of external temperature fluctuations; temperature regulation deviations caused by equipment aging and untimely maintenance; accuracy issues of flow controllers; and unknown interference between different zones.

[0118] Assuming the system has n regions, the process noise w(t) can be written as a column vector:

[0119]

[0120] Where w1(t), w2(t), ..., w n (t) represents the process noise from region 1 to region n.

[0121] Specific implementation methods:

[0122] Typically, w(t) is assumed to be Gaussian white noise with zero mean, and its covariance matrix is ​​Q, i.e.:

[0123] w(t)~N(0,Q)

[0124] Here, Q is the process noise covariance matrix, which describes the correlation and intensity of process noise in each region. The covariance matrix Q can be estimated by analyzing historical data or by setting empirical parameters.

[0125] Then, the observation equations are established:

[0126] y(t) = Cx(t) + v(t)

[0127] Here, C represents the observation matrix, describing the mapping from the state vector to the observation vector. Since sensor disconnection is possible, the state of not all areas may be observable; therefore, the C matrix may represent partial observations. v(t) represents the observation noise, indicating the error in the sensor measurement.

[0128] The observation matrix C represents the mapping from states to observations. In a zone cooling system, sensors cannot observe the state of all areas, so the C matrix is ​​usually a sparse matrix, indicating which states are observable.

[0129] Sensor placement: In a district cooling system, the sensors within a given area are known in advance, and the observation matrix C can be directly determined through simple geometric or positional relationships. If temperature sensors are installed in certain areas, the corresponding rows in the observation matrix C will be non-zero. When a temperature sensor is offline, the corresponding row will be zero.

[0130] Assuming there are 10 regions, and the sensor in region 3 is offline, then the observation matrix C can be represented as:

[0131]

[0132] Each row represents an observable state variable. For example, the first row [1,0,0,0,...] indicates that we can observe the temperature of the first region. The third row corresponds to region 3; when the sensor is offline, the temperature of this region cannot be observed, so this row is all 0. The other rows continue to represent the observed temperatures of other regions.

[0133] Furthermore, multivariate Kalman filtering is divided into a prediction step and an update step:

[0134] Prediction steps: Predict the state vector and error covariance matrix at the next time step:

[0135]

[0136] P(t|t-1)=AP(t-1|t-1)A T +Q

[0137] in, P(t|t-1) represents the predicted state vector at time t, P(t|t-1) is the prediction error covariance matrix, representing the uncertainty of the predicted state, and Q is the process noise covariance matrix, reflecting the random disturbance of the system state.

[0138] Update steps: Update the state estimate based on the observed data:

[0139]

[0140] r(t)=y(t)-y pred (t)

[0141] S(t)=CP(t|t-1)C T +R

[0142] Calculate the Kalman gain:

[0143] K(t)=P(t|t-1)C T S(t) -1

[0144] Update the state vector and error covariance matrix:

[0145]

[0146] P(t|t)=(IK(t)C)P(t|t-1)

[0147] Obtaining the process noise covariance matrix Q and the measurement noise covariance matrix R: The process noise covariance matrix Q reflects the uncertainty of the system model, such as random disturbances caused by external environmental temperature, equipment aging, etc. When Q cannot be directly obtained, an empirical setting is used: In many practical applications, the initial Q value can be set empirically. For example, a small process noise is initially assumed, and then gradually adjusted based on the filtering effect.

[0148] Measurement noise covariance matrix R: reflects the error of sensor measurement. For example, the accuracy of a temperature sensor may cause a difference between the observed value and the true value. R is estimated based on the noise parameters provided by the sensor manufacturer: R is set according to the measurement accuracy and noise range provided in the sensor's technical specifications.

[0149] After initializing Q and R in the previous step, Q and R are adaptively adjusted during runtime to continuously improve filter performance.

[0150] Adaptive Kalman Filtering: Online adjustment of Q and R matrices, based on residual adjustment.

[0151] Q and R are dynamically adjusted by monitoring the residuals of the Kalman filter (i.e., the difference between observed and predicted values). When the residuals increase, the process noise covariance Q can be increased to reflect the increased uncertainty of the system; when the residuals decrease, the measurement noise covariance R can be gradually decreased to increase the confidence in the observed data.

[0152] Adaptive filtering formula:

[0153] Q(t+1)=Q(t)+α·(r(t)r T (t)-P(t|t))

[0154] R(t+1)=R(t)+β·(r(t)r T (t)-S(t))

[0155] Where r(t) is the residual, P(t|t) is the updated error covariance, and α and β are adjustment coefficients.

[0156] In practical applications, the relationship between regional temperature and flow rate may be non-linear and may be influenced by historical data. Therefore, the predictive power of multivariate Kalman filtering can be further enhanced by incorporating historical data models. For example:

[0157] Constructing a temperature-flow rate model: Using historical data, construct a nonlinear response model of flow rate changes to temperature, such as:

[0158] T(t) = f(q(t), q(t-1), ...)

[0159] Here, f represents the use of a neural network method. This model can provide a direct response of temperature to changes in flow rate, and is corrected by combining previous multivariate Kalman filtering.

[0160] Prediction Correction: When the response between temperature and flow rate predicted by the Kalman filter does not match, the prediction results can be further corrected using historical data models.

[0161] It should be noted that the LSTM deep learning model was chosen to process time series data. The LSTM network consists of multiple LSTM units, each containing an input gate, a forget gate, and an output gate, which are used to capture long-term dependencies.

[0162] f t =σ(W f ·[h t-1 [,Z(t)]+b f )

[0163] i t =σ(W i ·[h t-1 [,Z(t)]+b i )

[0164] C t =f t ⊙C t-1 +i t ⊙tanh(W C ·[h t-1 [,Z(t)]+b C )

[0165] o t =σ(W o ·[h t-1 [,Z(t)]+b o )

[0166] h t =o t ⊙tanh(C t )

[0167] Where σ is the Sigmoid activation function, ⊙ represents element-wise multiplication, and W and b are the weights and bias parameters.

[0168] Output layer: Usually a fully connected layer, outputting the probability of fault determination or the binary classification result (normal / fault).

[0169]

[0170] During the training process, label generation is performed: based on historical data and fault events, fault labels f(k) are generated, where f(k) = 1 indicates a fault and f(k) = 0 indicates normal operation.

[0171] Loss function: In each training iteration, the binary cross-entropy loss function is used to update and optimize the model parameters.

[0172]

[0173] Optimization method: Use the Adam optimizer or other advanced optimization algorithms to minimize the loss function.

[0174]

[0175] Where θ represents the model parameters and η is the learning rate.

[0176] Regularization: To prevent overfitting, regularization terms such as L2 regularization or Dropout can be introduced.

[0177]

[0178] When the cumulative error E(k) exceeds the threshold E max Time processing, E max It is a pre-set cumulative error threshold, when E(k) > E max At this point, the system determines that the control deviation has exceeded the acceptable range and emergency measures are required. This is determined through simulation and historical data, based on the system's specific requirements and tolerance levels.

[0179] Reaching threshold E max Then the temperature needs to be restored.

[0180] During the recovery process in the fault area, the pipeline flow rate is restored to its initial value u0, and temperature recovery is predicted based on historical experience. (Alternatively, deep learning methods can be used to determine the duration of regulation based on post-fault predictions; once a certain time has elapsed, the initial temperature recovery process begins.) Assume the post-recovery temperature dynamics satisfy a first-order transfer function:

[0181]

[0182] Where T(t) is the region temperature at time t, T0 is the initial temperature (the initial temperature after flow recovery), and T ∞ It is the final stable temperature, and τ is the system's time constant.

[0183] Using the above model, it can be predicted that the temperature in the area will recover to 25 degrees within a certain period of time (e.g., 1 hour), with a confidence level of over 95% within the error range (±0.5 degrees).

[0184] When the confidence level reaches 95% or higher, we can consider the faulty region to have recovered to a safe state and can be reinstated for secondary frequency modulation. At this point, the system returns to the Kalman filter + deep learning detection and adjustment loop, continuously monitoring the system status and repeating the following steps: fault diagnosis and detection; control strategy adjustment; cumulative error monitoring; recovery and readjustment.

[0185] In a preferred embodiment of the invention, even if a sensor in a certain area goes offline, the Kalman filter can still infer the state of the offline area by dynamically adjusting the observation matrix C and using data from other areas. This robustness is particularly useful in cases of sensor instability or data loss, enabling the system to maintain high-precision state estimation and control.

[0186] The Kalman filter itself is designed to maintain robustness and high accuracy in state estimation even when sensors are offline or data is lost. This is achieved through Kalman filtering state estimation, dynamic adjustment of the observation matrix C, and adaptive adjustment to process and observation noise.

[0187] The Kalman filter's design inherently enables it to maintain robustness and high-accuracy state estimation even in the event of sensor disconnection or data loss. This is achieved through Kalman filtering state estimation, dynamic adjustment of the observation matrix C, and adaptive adjustment to process and observation noise. Specifically, the robustness of the Kalman filter is manifested in the following aspects:

[0188] The observation matrix C defines how the system state maps to the observed values; that is, it describes the relationship between the system state and the sensor observations. If a sensor in a certain area goes offline, the Kalman filter will dynamically adjust the C matrix to ignore the observations in that area.

[0189] For example, suppose there is a refrigeration system with multiple zones, each with a temperature sensor. When a sensor in a certain zone goes offline, the Kalman filter detects the problem and modifies the observation matrix C (e.g., setting the corresponding row for the offline zone to zero). In this way, the Kalman filter no longer relies on the measurement data of that zone, but instead relies on data from other zones and historical conditions for prediction and estimation.

[0190] When a sensor in a certain area goes offline, the corresponding row of the observation matrix is ​​set to zero (or the row is deleted directly) so that invalid data is ignored in the update step.

[0191] The original observation matrix C contains the mapping relationship of all sensors (e.g., one row for each sensor).

[0192] After a sensor goes offline, the adjusted C' only retains rows with valid sensors, for example: (The second sensor failed.)

[0193] In the update equation of the Kalman filter, the Kalman gain K is automatically adjusted according to C', and the state estimate is corrected only by the effective sensor.

[0194] The observation noise covariance R of the failed sensor can be set to a maximum value (or the corresponding rows and columns can be deleted) to further reduce its impact on the results.

[0195] Even if some sensors fail, the Kalman filter can still predict the system state using state transition equations and process noise. The Kalman filter uses historical data and observations from neighboring areas (if available) to predict the state of the offline region. This prediction relies on the system's physical model, such as the heat transfer equations for temperature changes. Through prediction update steps, even with missing observations, the Kalman filter can still estimate the state of the offline region based on available information. The system uses temperature data and flow information from other areas to refine the prediction, thus maintaining the overall system accuracy as much as possible.

[0196] State equation: x(k+1)=Ax(k)+Bu(k)+w(k)

[0197] Physical meaning: The system state x (such as temperature, flow rate) is dynamically transmitted through the model matrix A (thermodynamic coupling relationship) and the input matrix B (control input).

[0198] Robustness of the prediction step: Even if some sensors fail, the system can still provide prior values ​​x(k|k-1) for state estimation through model prediction.

[0199] Models A and B implicitly contain physical relationships between regions (such as energy conservation and mass conservation), allowing the state of unobserved regions to be indirectly inferred from observation data of adjacent regions.

[0200] Kalman filtering is adaptive, adjusting its dependence on measurements based on their reliability. By adjusting the Kalman gain, the Kalman filter dynamically adjusts its state estimation according to actual measurement noise and prediction errors. Specifically, when a sensor in a certain area is offline or the data is unstable, the Kalman gain reduces its dependence on observations in that area and increases its confidence in observations in other areas.

[0201] Predicted covariance: P(k|k-1)=AP(k-1)A T +Q

[0202] Function: The prediction error covariance P(k|k-1) reflects the model uncertainty and guides the weight allocation of observation data in the update step.

[0203] Update covariance: P(k|k)=(IK(k)C′)P(k|k-1)

[0204] Effect: When a sensor fails, the adjusted C' causes the Kalman gain K to redistribute the trust weights, relying more on model predictions and other effective sensors.

[0205] Kalman filters can consider the interdependencies between different regions by expanding the system state and observation matrix into a multivariate form. For example, when a sensor in a certain region goes offline, the Kalman filter can use temperature data and flow information from other regions to estimate the temperature of the offline region. Based on the physical relationships of heat transfer and supported by observation data from other regions, the system can continue to maintain an estimate of the state of the faulty region.

[0206] When a sensor fails, open-loop control relies entirely on observation; failure directly interrupts control. Simple interpolation cannot utilize the system model, resulting in large extrapolation errors. Static observers, with fixed weight allocation, cannot dynamically adapt to sensor failure. Kalman filtering, on the other hand, maintains robustness through model prediction and dynamic weight allocation.

[0207] Example 3, referring to Figures 2-4 This invention provides a control method for a regional cooling system to participate in the secondary frequency regulation of the power grid. To verify the beneficial effects of this invention, scientific demonstration is carried out through experiments.

[0208] The system is controlling a cooling system with four zones, and at some point, the sensor in zone 3 goes offline. There are four zones, each with a temperature T. i It is necessary to adjust the cold water flow rate q i To control this. The state equation and observation equation are as follows:

[0209] State vector x(t): contains temperature and flow rate of 4 regions.

[0210] x(t)=[T1(t) T2(t) T3(t) T4(t) q1(t) q2(t) q3(t) q4(t)] T

[0211] Control input: Represents flow rate u(t) = [q1(t) q2(t) q3(t) q4(t)] T

[0212] Observation vector: Only observes the temperature of each region.

[0213] y(t)=[T1(t) T2(t) T3(t) T4(t)] T

[0214] Kalman filter settings

[0215] Equations of state:

[0216] x(t+1)=A·x(t)+B·u(t)+w(t)

[0217] Where A is the state transition matrix, reflecting the dynamics of temperature change and the interaction between regions; B is the control input matrix, describing the effect of flow rate q(t) on temperature T(t); and w(t) is the process noise, usually assumed to be zero-mean Gaussian noise with covariance Q.

[0218] Suppose A and B are in the following forms:

[0219]

[0220]

[0221] Observation equation:

[0222] y(t) = C·x(t) + v(t)

[0223] Here, C is the observation matrix, used to map the state vector x(t) to the observation vector y(t). Under normal circumstances, the C matrix is:

[0224]

[0225] Assume the observed noise v(t) is zero-mean Gaussian noise with covariance R.

[0226] Implementation steps

[0227] 1. Prediction Steps

[0228] First, predict the state at the next time step based on the current state x(t) and the control input u(t):

[0229]

[0230] And calculate the prediction error covariance matrix P(t+1|t):

[0231] P(t+1|t)=A·P(t)·A T +Q

[0232] 2. Observation Update

[0233] Suppose that at time t, the temperature sensor in region 3 goes offline, meaning the temperature of region 3 cannot be observed. In this case, the observation matrix C needs to be modified by setting the third row to 0:

[0234]

[0235] Next, calculate the Kalman gain K(t):

[0236] K(t)=P(t|t-1)·C T ·(C·P(t|t-1)·C T +R) -1

[0237] Update the state estimate based on the observation data:

[0238]

[0239] And update the error covariance matrix:

[0240] P(t|t)=(IK(t)·C)·P(t|t-1)

[0241] Reference Figure 2 The diagram illustrates the temperature changes over time in four regions. In region 3, a sensor malfunction occurs at 60 minutes, causing the temperature to remain constant until it recovers at 80 minutes and re-engages in temperature regulation. After this recovery point, the temperature in region 3 gradually returns to its initial state, accompanied by minor fluctuations. This meets the control requirements of the constraints.

[0242] Reference Figure 3 and Figure 4 After the sensor failure, the residuals increased significantly, reflecting an increase in model prediction error. Once the threshold was reached, the system began to recover temperature. During the recovery process, the residuals gradually decreased, indicating improved prediction accuracy after the system regained control. The dynamic changes in the residuals visually illustrate the robustness of the model during the failure and recovery processes.

[0243] Example 4, refer to Figure 2 and Figure 4 As one embodiment of the present invention, this embodiment provides a control system for a district cooling system participating in secondary frequency regulation of the power grid, comprising:

[0244] The system state modeling module is used to construct a system state model that describes the temperature and cooling water flow rate of multiple regions, so as to reflect the dynamic interaction between the regions.

[0245] The state estimation module is used to predict the operating state of each region based on the system state model, and to estimate the missing state when monitoring data is missing or abnormal.

[0246] An anomaly detection module is used to identify whether there is a regional sensor fault based on the difference between the current observation data and the predicted state in each control cycle.

[0247] The control execution module is used to perform control operations to replace the observed data based on the predicted state when a fault is detected in a certain area, and to continuously calculate the state deviation of the fault area.

[0248] The strategy switching module is used to perform a control strategy switching operation on the fault area when the state deviation exceeds a set threshold range.

[0249] The state recovery judgment module is used to re-enable the fault area to participate in the secondary frequency regulation control process of the power grid when the state of the fault area recovers to the target temperature range and meets the preset information conditions.

[0250] It should be noted that when the district cooling system participates in the grid interaction for secondary frequency regulation, some sensors in the district may go offline (this will cause the control to become open-loop control and lose the temperature feedback signal); at the same time, there may be excessive sensor signal transmission delay; and the sensor may collect incorrect temperature data.

[0251] Current methods for enabling district cooling systems to participate in grid interaction for secondary frequency regulation assume ideal, normal system data acquisition. There is currently no specific research addressing how to allow district cooling systems to participate in grid interaction for secondary frequency regulation even when some sensors are offline or data is erroneous.

[0252] These situations can render the method of district cooling systems participating in grid interaction for secondary frequency regulation unreliable, and may even lead to catastrophic consequences during the regulation process. Therefore, highly specialized, precise, and safe control methods are required.

[0253] Generally, if a sensor problem occurs in a region, the simplest approach is to restore the initial pipeline flow rate for that region, thus removing it from the secondary frequency regulation process. However, this will result in a loss of some regulation potential, leading to a decrease in regulation efficiency and capability.

[0254] Therefore, to address this problem, this embodiment proposes a method combining Kalman filtering, deep learning, and physical modeling to enhance the robustness of the control and regulation method for regional cooling systems to participate in secondary frequency regulation through grid interaction. This enables the regional cooling system to effectively participate in secondary frequency regulation even in the event of data loss or data errors.

[0255] Step 1. The first step is to determine if the operating district cooling system is malfunctioning. A temperature sensor disconnection is easily identified. However, determining if the sensor temperature delay exceeds the threshold tolerance range or if the temperature data acquisition is directly erroneous is difficult. This is because district cooling systems have thermal inertia, making it difficult for traditional methods to quickly determine if there is a problem with the transmitted temperature data. Therefore, a Kalman filter combined with deep learning is proposed for this purpose, learning from past curves.

[0256] Step 2. After determining that a sensor fault has occurred in the system, change the original control strategy. When a fault is detected in zone 2, accurately determine the time of the fault occurrence, use the temperature and flow data before the fault as the standard for subsequent prediction, and adjust the zone cooling system accordingly.

[0257] Step 3. When using predicted data to adjust the temperature of the faulty area, this leads to a cumulative error. Over time, the temperature error increases, eventually exceeding the tolerance range, which is unacceptable. Therefore, we propose a new control approach: first, we use deep learning to determine the reliability; then, during the adjustment process, if we believe there is a 50% probability that the temperature cannot be safely controlled within the permissible range, we proceed to the next adjustment step.

[0258] Step 4. For the faulty area, begin restoring the original pipeline flow supply. Based on prior experience, it is known that after restoring the original pipeline flow supply, for example, after 1 hour, the temperature in this area will return to its initial temperature, such as 25 degrees Celsius, with a very small error (e.g., ±0.5 degrees Celsius). This is equivalent to restoring the area. At this point, the area can be reused for secondary frequency regulation. Repeat steps 2, 3, and 4 to ultimately achieve complete temperature regulation.

[0259] This embodiment also provides an electronic device applicable to a control method for a district cooling system participating in secondary frequency regulation of the power grid, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the control method for a district cooling system participating in secondary frequency regulation of the power grid as proposed in the above embodiment.

[0260] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a control method for a regional cooling system to participate in secondary frequency regulation of the power grid, as proposed in the above embodiments.

[0261] The storage medium proposed in this embodiment and the control method for implementing a regional cooling system to participate in the secondary frequency regulation of the power grid proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0262] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (RM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0263] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A control method for a district cooling system participating in secondary frequency regulation of the power grid, characterized in that: include, Construct a system state model to describe the temperature and cooling water flow rate in multiple zones; Based on the system state model, the operating state of each region is predicted by the state estimation unit; Within each control cycle, the anomaly detection module identifies whether there is a regional sensor malfunction based on the difference between the current observation data and the predicted state. When a fault is identified in a certain area, control operations are performed based on the predicted state to replace the observed data, and the state deviation of the faulty area is continuously calculated. If the state deviation exceeds the set threshold range, a control strategy switching operation is performed on the faulty area; After determining that the state of the faulty area has recovered to the target temperature range and meets the preset signal conditions, the faulty area is re-involved in the secondary frequency regulation control process of the power grid.

2. The control method for a regional cooling system participating in secondary frequency regulation of the power grid as described in claim 1, characterized in that: The state estimation unit constructs a state space model based on the system state model. The state space model includes temperature state variables and cooling water flow state variables for multiple regions. The model parameters include the thermal coupling coefficient and flow influence coefficient between regions. In cases where observation data is missing in some areas, a recursive state estimation process is performed using the state transition matrix and the observation matrix to output a complete state estimate for control execution.

3. The control method for a regional cooling system participating in secondary frequency regulation of the power grid as described in claim 2, characterized in that: The anomaly detection module collects the state estimation residuals over multiple consecutive control cycles and constructs a sliding time window sequence. The residual sequence within the time window is used as input features and input into the trained neural network model for analysis. Based on the output of the neural network model, it is determined whether there are any abnormal states related to the regional temperature sensor during the control cycle, including data loss, data delay, or sudden changes in observation error.

4. The control method for a regional cooling system participating in secondary frequency regulation of the power grid as described in claim 3, characterized in that: Determining whether the state deviation exceeds the set range includes: squaring the state estimation residuals within the current control cycle and accumulating them to form an error index; The error index is compared with the error threshold set by the system; If the cumulative error index over multiple consecutive control cycles exceeds the error threshold, a control strategy switching signal is generated to initiate subsequent adjustment operations.

5. The control method for a regional cooling system participating in secondary frequency regulation of the power grid as described in claim 4, characterized in that: The control strategy switching operation includes: adjusting the cooling water flow rate of the fault area to the basic flow rate value of the initial setting stage of the fault area; Remove the faulty area from the current secondary frequency modulation control task and stop issuing frequency modulation adjustment commands to it; The base temperature maintenance control mode for the fault area is enabled, which is used only to maintain the local ambient temperature stability of the fault area.

6. The control method for a regional cooling system participating in secondary frequency regulation of the power grid as described in claim 5, characterized in that: After the control strategy switching operation is performed, a temperature response model of the fault area is constructed based on the system's historical operating data. The temperature response model is used to characterize the temperature change process of the fault area over time under stable cooling water flow conditions. The temperature response model is used to predict the temperature of the fault area, and the prediction result is compared with the target temperature range to determine whether the state of the fault area has returned to stability.

7. The control method for a regional cooling system participating in secondary frequency regulation of the power grid as described in claim 6, characterized in that: After the prediction results of the temperature response model indicate that the temperature in the fault area is close to the target temperature range, the actual observation confirmation step is performed. During multiple consecutive control cycles, the actual temperature of the fault area is collected and compared with the set temperature tolerance range; Once the temperature in the faulty area stabilizes within the tolerance range for a preset time threshold, its control state is automatically switched, and the faulty area is restored to participate in the secondary frequency regulation control process of the power grid.

8. A control system for a district cooling system participating in secondary frequency regulation of the power grid, comprising the control method for a district cooling system participating in secondary frequency regulation of the power grid as described in any one of claims 1 to 7, characterized in that, include: The system state modeling module is used to construct a system state model that describes the temperature and cooling water flow rate of multiple regions, so as to reflect the dynamic interaction between the regions. The state estimation module is used to predict the operating state of each region based on the system state model, and to estimate the missing state when monitoring data is missing or abnormal. An anomaly detection module is used to identify whether there is a regional sensor fault based on the difference between the current observation data and the predicted state in each control cycle. The control execution module is used to perform control operations to replace the observed data based on the predicted state when a fault is detected in a certain area, and to continuously calculate the state deviation of the fault area. The strategy switching module is used to perform a control strategy switching operation on the fault area when the state deviation exceeds a set threshold range. The state recovery judgment module is used to re-enable the fault area to participate in the secondary frequency regulation control process of the power grid when the state of the fault area recovers to the target temperature range and meets the preset information conditions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the control method for a regional cooling system participating in the secondary frequency regulation of the power grid, as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method for a regional cooling system participating in the secondary frequency regulation of the power grid, as described in any one of claims 1 to 7.

Images