Multi-unit coordinated optimization control method and system for water supply pump based on demand response

Abstract

The application discloses a multi-unit water supplement pump coordinated optimization control method and system based on demand response, and belongs to the technical field of industrial process automation control. The application receives a demand response instruction, performs safety constraint pre-checking on a pipe network pressure margin, a liquid tank liquid level margin and a reducible power margin, determines an effective target response power, establishes a normalized double-target optimization scheduling model, and performs weighted summation on pipe network pressure deviation and response quantity deviation after normalizing the two into dimensionless quantities. A two-stage method is used to solve an optimal scheduling scheme. In a response execution process, a hyperbolic tangent function is used to adaptively adjust a weight coefficient according to a time period average response power deviation and to rollingly update the scheduling scheme, so that a response completion rate and pipe network pressure stability are dynamically balanced. An independent safety protection module at an execution layer and hard constraints at an optimization layer jointly constitute a double safety mechanism. The system adopts a three-layer architecture and is suitable for coordinated scheduling control of multi-unit variable frequency pump groups in an industrial fluid pipe network.

Description

Technical Field

[0001] This invention belongs to the field of industrial process automation control technology, specifically relating to a coordinated optimization control method and system for multi-unit water supply pumps based on demand response. Background Technology

[0002] Industrial pump groups, represented by variable frequency centrifugal pumps, are characterized by adjustable power and rapid response. They are important controllable load resources for demand response in industrial fluid transportation networks. In industrial fluid transportation networks, multiple variable frequency centrifugal pumps typically operate in parallel to jointly undertake the task of maintaining network pressure and replenishing fluid. The pump group adjusts its speed through frequency converters to maintain stable network pressure within a set range. The frequency converter drive allows for continuous adjustment of pump speed over a wide range, providing the physical basis for the pump group's participation in demand response.

[0003] However, existing pump group control technologies have several shortcomings in demand response, hindering the widespread application of industrial pump groups in the power demand response field. Regarding participation condition verification, existing technologies lack a systematic, multi-dimensional safety pre-inspection mechanism for pipeline pressure margin, storage tank level margin, and available power reduction margin. Under complex operating conditions such as low pipeline pressure, limited storage tank level margin, or insufficient available power reduction margin, it is difficult to accurately identify the actual participation capability of the pump group, posing a risk of blindly implementing power reduction under unsafe conditions. In terms of scheduling optimization, when participating in demand response, pump groups simultaneously face two mutually constraining control objectives: response power tracking and pipeline pressure stabilization. These two objectives have different dimensions. When existing technologies weight and merge these two objectives into a unified objective function, the lack of normalization often results in the larger objective term dominating the objective function in the optimization calculation. The weighting coefficients cannot truly regulate the relative priority of the two objectives, leading to a deterioration in pipeline pressure control or a low demand response completion rate. During response execution, existing technologies typically employ fixed-weight optimization strategies, failing to adaptively adjust the focus of scheduling decisions based on real-time changes in demand response progress. This leads to insufficient tracking when response progress lags, resulting in low response completion rates, and an overemphasis on response accuracy even when targets are met, compromising pipeline pressure stability. Regarding equipment protection, performance-oriented scheduling strategies often neglect constraints on frequent pump start-ups and shutdowns. Frequent switching of pump operating states in the scheduling scheme can cause sudden flow changes in the pipeline, triggering water hammer pressure surges and accelerating the wear of contactors and other electrical components, shortening equipment lifespan. Furthermore, existing safety protection mechanisms are usually coupled with the optimization scheduling process. When abnormal situations such as model estimation deviations or communication interruptions occur during optimization scheduling, there is a lack of independent safety fallback measures, making it difficult to guarantee the safety of the pipeline network under extreme operating conditions.

[0004] The aforementioned technical problems make it difficult for existing pump group control technologies to simultaneously ensure pipeline safety, response accuracy, and equipment protection when participating in demand response. There is an urgent need for a coordinated optimization control method that can systematically solve these problems. Summary of the Invention

[0005] To address the problems existing in the background art, the present invention provides a multi-unit makeup water pump coordinated optimization control method based on demand response, comprising the following steps:

[0006] S1: Receive the demand response command from the power grid side, obtain the target response power and response duration, and collect the operating status parameters of the multi-unit pump group in real time. The operating status parameters include the speed of each pump, the current total power of the pump group, the pressure of the main pipeline, and the liquid level of the storage tank.

[0007] S2: Based on the operating status parameters, a safety constraint pre-check is performed, and it is determined in sequence whether the pipeline pressure margin, the liquid level margin of the storage tank, and the power reduction margin meet the conditions for participating in demand response. After the pre-check is passed, the effective target response power is determined.

[0008] S3: Using the start-stop status and speed of each pump as decision variables, the weighted sum of the normalized pipeline pressure deviation and the normalized response deviation as the objective function, and the pipeline pressure constraint, liquid level constraint and start-stop number constraint as hard constraints, a dual-objective mixed integer nonlinear constraint optimization scheduling model is established, and the optimal scheduling scheme is solved by a two-stage method combining enumeration and sequential quadratic programming.

[0009] S4: During the demand response execution process, calculate the average response power up to the current time period. Based on the deviation between the average response power and the target response power, dynamically adjust the weight coefficients in the target function using the hyperbolic tangent function, and re-solve the optimized scheduling model according to the scheduling cycle to update the scheduling scheme on a rolling basis.

[0010] S5: Send the start / stop status and speed commands of each pump to the corresponding frequency converter for execution, and monitor the pressure of the main pipeline in real time. When the pressure of the main pipeline is lower than the emergency protection pressure, force exit the demand response mode and switch to constant pressure control mode.

[0011] Further, step S2 includes: S21: Determine whether the current main pipeline pressure is higher than the safe lower limit of the pipeline pressure. If the main pipeline pressure is already at the safe lower limit of the pipeline pressure, it is determined that the pressure conditions for participating in demand response are not met, a prohibition response flag is output and scheduling is terminated to realize pressure margin pre-check.

[0012] S22: Based on the difference between the current liquid level in the storage tank and the safe low liquid level, the effective cross-sectional area of ​​the storage tank, and the minimum flow rate required to maintain the safety of the pipeline pressure, calculate the maximum response duration that the liquid level can support. If the maximum response duration does not meet the requirement of the response duration, issue a liquid level insufficient warning to realize liquid level reserve pre-check.

[0013] S23: Calculate the difference between the current total power of the pump group and the minimum total power required to maintain the minimum flow rate as the power margin that can be reduced. If the power margin that can be reduced is zero, it is determined that there is no power condition to participate in demand response and the scheduling is terminated. If the power margin that can be reduced is less than the target response power, the power margin that can be reduced is used as the effective target response power to enter the partial response mode. Otherwise, the target response power is used as the effective target response power. Complete the safety constraint pre-check and pass the effective target response power to step S3.

[0014] Furthermore, step S3 includes: S31: Based on the centrifugal pump similarity law, establish the power-speed relationship of each pump according to the rated power, rated speed and actual speed of each pump; establish the flow-speed relationship of each pump according to the rated flow rate, rated speed and actual speed of each pump; based on the pipeline flow balance relationship, establish a pipeline pressure estimation model according to the current measured main pipe pressure, pipeline hydraulic resistance coefficient and the difference between the total injection flow rate of each operating pump and the current pipeline demand flow rate, thereby constructing the mapping relationship between decision variables and pipeline pressure and response power;

[0015] S32: Based on the start / stop state vectors of each pump and rotational speed vector For the decision variables, establish the following normalized biobjective optimization function:

[0016] ;

[0017] in, The objective function value; This is a start / stop status vector for each pump, where each element takes the value 0 or 1, where 0 indicates shutdown and 1 indicates operation. For each pump speed vector; This is the weighting coefficient for pipeline pressure deviation; For the response deviation weighting coefficient, and ; The output value (MPa) of the pipeline pressure estimation model established in step S31 under the current scheduling scheme. Set the pipeline pressure (MPa); The allowable pressure range (MPa) of the pipeline network is equal to the upper limit of the safe pressure of the pipeline network. With pipeline pressure safety lower limit The difference is used to normalize the pressure deviation to a dimensionless value; The actual response power (kW) of the dispatching scheme is equal to the baseline value of the total power of the pump group before the response. The difference between the power of each operating pump and the sum of the power of the pumps under the scheduling scheme; The effective target response power (kW) is used to normalize the response deviation to dimensionless.

[0018] S33: Set hard constraints, including: upper and lower limits of pipeline pressure safety, upper and lower limits of liquid level safety in storage tank, minimum number of operating pumps, upper and lower limits of pump speed, and upper limit of start-stop count in adjacent scheduling cycles. The upper limit of start-stop count restricts the number of pumps whose start-stop status changes between adjacent scheduling cycles from not exceeding a preset upper limit.

[0019] S34: The optimization scheduling model is solved using a two-stage method: In the first stage, all legal start-stop combinations that simultaneously satisfy the minimum number of operating units constraint and the upper limit of start-stop times constraint are traversed; In the second stage, the start-stop state is fixed for each legal start-stop combination, and the speed vector is used as a continuous decision variable. The sequential quadratic programming method is used to iteratively solve the problem under the speed constraint and pipeline pressure constraint; The scheme with the smallest objective function value among all legal start-stop combinations is selected as the optimal scheduling scheme output.

[0020] Further, in step S31, the power-speed relationship of each pump is established based on the cubic law of centrifugal pump similarity, that is, the ratio of the power of each pump at the actual speed to its rated power is equal to the cube of the ratio of the actual speed to the rated speed; the flow-speed relationship of each pump is established based on the linear law of centrifugal pump similarity, that is, the ratio of the flow rate of each pump at the actual speed to its rated flow rate is equal to the ratio of the actual speed to the rated speed; the output value of the pipeline pressure estimation model is equal to the current measured main pipe pressure plus the product of the pipeline hydraulic resistance coefficient and the difference between the total injection flow rate and the pipeline demand flow rate, wherein the pipeline hydraulic resistance coefficient is calibrated by the pipeline filling test; in step S34, the second stage uses the current actual operating speed of each pump as the initial iteration point of the sequential quadratic programming method.

[0021] Furthermore, step S4 includes: S41: In each scheduling cycle, calculate the time-period average response power up to the current moment, that is, take the cumulative arithmetic average of the actual response power of each historical scheduling cycle;

[0022] S42: Based on the deviation between the average response power over the time period and the effective target response power, update the response deviation weighting coefficient according to the following formula:

[0023] ;

[0024] in, This is the response deviation weighting coefficient updated in the current scheduling cycle; This is a truncation function that restricts the calculation result of the first term within the parentheses to a lower bound. and upper limit between; To balance the basic weights of operating conditions; This represents the maximum adjustment range for the weights. It is the hyperbolic tangent function; The gain coefficient is the response sensitivity coefficient. Effective target response power (kW); As of the current moment The average response power over the time period (kW); This is the lower limit of the response deviation weight; The upper limit of the weight for the response deviation;

[0025] S43: Update the pipeline pressure deviation weighting coefficient ,in The updated pipeline pressure deviation weighting coefficient for the current scheduling cycle; the updated... and Substitute the objective function from step S3 into the solution, re-solve the optimized scheduling model, and output the optimal scheduling scheme for the current scheduling period to achieve rolling optimization and updating.

[0026] Furthermore, step S5 includes: S51: Converting the pump speed commands in the optimal scheduling scheme into analog signals or digital communication signals and sending them to the corresponding frequency converters; the operating status feedback signals of each pump frequency converter after execution are transmitted back to the operating status parameter acquisition link in step S1 in real time, forming a closed-loop control of acquisition, optimization, execution and feedback.

[0027] S52: During the entire demand response execution, independent of the optimization scheduling process in steps S3 and S4, the pressure of the main pipeline is monitored in real time. When the pressure of the main pipeline is lower than the emergency protection pressure, the demand response mode is immediately forcibly exited and switched to the constant pressure control mode. The emergency protection pressure is lower than the safety lower limit of the pipeline pressure set in step S3, which constitutes a safety fallback protection independent of the optimization scheduling.

[0028] The present invention also provides a demand-response-based multi-unit water supply pump coordination and optimization control system, including a perception layer, a decision optimization layer and an execution layer;

[0029] The sensing layer includes a signal acquisition and communication module; the signal acquisition and communication module is connected to the status output port signals of the pipeline pressure sensor, flow sensor, liquid storage tank level gauge and the frequency converter of each unit, respectively, for acquiring pipeline main pipe pressure, replenishment flow, liquid storage tank level and operating status parameters of each pump; the signal acquisition and communication module is also connected to the power grid demand response platform for receiving demand response commands;

[0030] The decision optimization layer includes a state preprocessing module and a multi-objective coordination and scheduling module. The state preprocessing module is connected to the data output terminal of the signal acquisition and communication module and is used to filter and process outliers in the acquired data before outputting a state estimate. The multi-objective coordination and scheduling module is connected to the output terminal of the state preprocessing module and the instruction output terminal of the signal acquisition and communication module, respectively, and is used to perform safety constraint pre-check, solve the normalized bi-objective optimization scheduling model, and perform adaptive weight adjustment, and output the start / stop status and speed instructions of each pump.

[0031] The execution layer includes a frequency converter control interface and a safety constraint protection module. The frequency converter control interface is connected to the output of the multi-objective coordinated scheduling module and to the control input of each pump frequency converter, used to convert speed commands into control signals and send them to each pump frequency converter. The safety constraint protection module is connected to the signal acquisition and communication module to obtain the main pipeline pressure. When the main pipeline pressure is lower than the emergency protection pressure, it sends a forced exit command to the frequency converter control interface, exits the demand response mode, and switches to the constant pressure control mode.

[0032] In the preferred embodiment, the inverter control interface supports analog signal output and digital communication signal output; the operating status feedback signals of each pump inverter are transmitted back to the decision optimization layer via the signal acquisition and communication module, forming a closed-loop control loop of perception layer acquisition, decision optimization layer solution, execution layer distribution, and feedback transmission; the safety constraint protection module operates independently of the multi-objective coordination and scheduling module, constituting an independent safety hard constraint fallback protection.

[0033] The beneficial effects achieved by this invention are as follows:

[0034] This invention designs a multi-unit makeup water pump coordinated optimization control method based on demand response. It employs a three-level safety constraint pre-check mechanism centered on pipeline pressure margin, storage tank level margin, and available power reduction margin. Before initiating optimization scheduling calculations, it verifies the basic conditions for the pump group to participate in demand response at each level. By introducing a partial response mode, the system can participate in the response at its maximum safe capacity even when the power margin is insufficient, avoiding safety accidents such as pipeline pressure instability or storage tank depletion caused by power reduction under conditions where participation is not possible. In addition to the optimization scheduling layer, this invention also independently sets up a safety fallback protection module with an emergency protection pressure as the trigger threshold. This module operates independently of the optimization scheduling module. When deviations occur during the optimization scheduling process due to abnormalities such as model errors, sensor delays, or communication interruptions, it can still forcibly exit the demand response mode and switch to a constant pressure control mode. This dual safety mechanism, with hard constraints at the optimization layer and independent protection at the execution layer complementing each other, comprehensively ensures the operational safety of the pump group throughout the entire demand response process.

[0035] This invention constructs a normalized bi-objective optimization framework that integrates minimizing pipeline pressure deviation and demand response power deviation into a single objective function. By normalizing the two objective terms using the allowable pressure range of the pipeline and the effective target response power, the pressure and power deviations, which have different dimensions, are converted into dimensionless relative deviations. This eliminates the problem of weight parameters lacking clear physical meaning due to inconsistent dimensions, allowing the magnitude of the weight coefficients to directly and reliably reflect the relative priority of the two optimization objectives, providing an intuitive reference benchmark for engineering tuning. Simultaneously, this invention employs a two-stage solution method combining enumeration and sequential quadratic programming. In the start-stop combination enumeration stage, start-stop constraints significantly compress the candidate solution space. In the speed optimization stage, the current actual operating speed is used as the initial iteration point to accelerate convergence. Thus, without relying on dedicated computing hardware, the mixed-integer nonlinear constraint optimization problem is efficiently solved within each scheduling cycle, yielding an optimal scheduling scheme that simultaneously considers pipeline pressure stability and demand response accuracy.

[0036] This invention designs an adaptive weight dynamic adjustment mechanism that uses the deviation between the time-period average response power and the effective target response power as the driving signal and the hyperbolic tangent function as the mapping core. During the demand response execution process, the weight coefficients in the dual-objective optimization function are automatically updated in each scheduling cycle. Combined with a rolling optimization strategy, this achieves continuous adaptive control of the relative priority between the pipeline pressure stability target and the response completion target. The smooth and bounded characteristics of the hyperbolic tangent function ensure that the weights change continuously with the response deviation without abrupt changes, avoiding control oscillations caused by weight step jumps. Upper and lower limit truncation further ensures that both objectives are balanced to a certain extent under any operating condition. When the response progress lags, the weights automatically shift towards the response target to increase tracking intensity; when the response approaches the target, the weights automatically return to equilibrium to maintain pipeline pressure stability. Therefore, compared with the fixed weight method, this significantly improves the power tracking convergence speed and overall response completion rate of demand response, while effectively suppressing the fluctuation amplitude of pipeline pressure.

[0037] This invention incorporates the upper limit of pump start-stop times in adjacent scheduling cycles as a hard constraint into the optimization scheduling model. This eliminates candidate schemes that frequently switch pump operating states at the source of optimization scheme generation, ensuring that power regulation is primarily achieved through continuous, gradual changes in speed rather than frequent switching of the number of pumps. This effectively suppresses water hammer pressure surges caused by sudden changes in flow rate in the pipeline, reduces the frequent operation of contactors and other electrical components, extends equipment lifespan, and ensures the safe and stable operation of the pump group during demand response. Furthermore, this invention implements the above method using a three-layer architecture: a perception layer, a decision optimization layer, and an execution layer. Each layer has a clear functional division and defined interfaces. The perception layer is responsible for data acquisition and communication, the decision optimization layer is responsible for state preprocessing and multi-objective coordinated scheduling, and the execution layer is responsible for command issuance and independent safety protection. These layers form a complete closed-loop control circuit, and the overall system can operate stably on a standard industrial computer, demonstrating good engineering feasibility and cross-industry applicability. Attached Figure Description

[0038] Figure 1 The graph shows a comparison of the response power tracking performance of Example 1, Example 2 and Comparative Example 1. (a) is a tracking curve of the response power of each scheme changing with the scheduling cycle, and (b) is a curve of the adaptive weight coefficient dynamic adjustment process.

[0039] Figure 2 The diagram shows a comparison of the network pressure stability of Example 1 and Comparative Example 2. (a) is a time-series curve of the network main pipe pressure after normalization in Example 1, and (b) is a time-series curve of the network main pipe pressure in Comparative Example 2.

[0040] Figure 3 The chart shows the comparison of start-stop frequency and equipment protection between Example 1, Example 3 and Comparative Example 3. (a) is a grouped bar chart of the cumulative start-stop switching times and the maximum continuous switching times of a single pump for each scheme, and (b) is a bar chart of the maximum transient pressure impact on the pipeline network for each scheme.

[0041] Figure 4 The bar charts are the comprehensive performance indicators of Examples 1-3 and Comparative Examples 1-3, where (a) is the bar chart of the response completion rate of each scheme, (b) is the bar chart of the standard deviation of the pipeline pressure of each scheme, and (c) is the bar chart of the cumulative number of start-stop switching of each scheme.

[0042] Figure 5 This is a flowchart of the demand-response-based multi-unit water supply pump coordinated optimization control method of the present invention. Detailed Implementation

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] The multi-unit makeup water pump coordinated optimization control method based on demand response provided by this invention uses the power grid demand response signal as the external driver, the pipeline pressure safety and the liquid level safety of the storage tank as hard constraints, and a normalized bi-objective weighted optimization algorithm as the decision-making basis. Through an adaptive weight adjustment mechanism, it dynamically balances the response completion rate and pipeline pressure stability during the response execution process. (Refer to...) Figure 5 The method of the present invention includes five main steps, S1 to S5. The specific implementation of each step is described in detail below.

[0045] Step S1 involves receiving the demand response command and collecting the pump group's operating status parameters. During the normal operation of the industrial fluid transport network, the multi-unit make-up pump group undertakes the task of maintaining pressure and replenishing the network, with each pump operating at a certain speed driven by a frequency converter. When the power grid issues a demand response command, the host computer control system receives the command through the communication interface and parses it to obtain two key parameters: the target response power. and response duration The target response power is the total power reduction required by the power grid for the pump group during the response period, and the response duration is the duration for which the power grid requires the pump group to maintain the reduction state. Communication between the host computer and the power grid demand response platform can use industry standard communication protocols such as Modbus / TCP or IEC61968.

[0046] In step S1, the host computer simultaneously collects real-time operating status parameters of the multi-unit pump group through the signal acquisition and communication module of the sensing layer. The collected parameters include the current speed of each pump and the current total power of the pump group. Main pipeline pressure and the liquid level in the storage tank The speed of each pump can be obtained from the speed feedback signal built into the frequency converter; the current total power of the pump group can be obtained by summing the power detection values ​​built into the frequency converters of each pump, or by measuring it through an external energy meter; the pressure of the main pipeline is collected by a pressure sensor installed on the main pipeline; the liquid level of the storage tank is collected by a level gauge. After the above operating status parameters are filtered and outlier removed by the status preprocessing module, a reliable status estimate is output, which provides input data for the safety constraint pre-check in step S2 and the establishment of the optimization scheduling model in step S3.

[0047] Step S2 involves a safety constraint pre-check based on operating status parameters. The purpose of this safety constraint pre-check is to determine whether the current operating conditions meet the basic requirements for participating in demand response before performing optimized scheduling calculations, thus avoiding safety accidents such as pipeline pressure instability or insufficient liquid level in the storage tank caused by performing power reduction operations under unsafe conditions. Step S2 comprises three sub-steps: S21, S22, and S23, which respectively pre-check the pipeline pressure margin, the storage tank level margin, and the available power reduction margin.

[0048] Step S21 is a pressure margin pre-check. The control system determines the current pressure of the main pipeline. Is it higher than the lower limit of safe pressure in the pipeline network? The lower safety limit of pipeline pressure is the minimum pressure value that must be maintained to ensure the normal operation of the pipeline network, and it is determined by the pipeline network design parameters. If the current pressure of the main pipeline is already at the lower safety limit, it means that there is no room for further pressure reduction. Implementing power reduction at this point would lead to a decrease in pump output and a reduction in the amount of fluid injected into the pipeline, causing the pipeline pressure to drop below the lower safety limit. Therefore, the control system outputs a prohibition signal and terminates the scheduling process. Only when the pressure of the main pipeline is higher than the lower safety limit is the pipeline considered to have sufficient pressure margin to continue subsequent pre-inspections.

[0049] Step S22 is a pre-check of the liquid level. The control system checks the current liquid level in the storage tank. With safe low liquid level The difference between the effective cross-sectional area of ​​the liquid storage tank and the minimum flow rate required to maintain safe pipeline pressure. Calculate the maximum response duration that the liquid level can support. The calculation principle is as follows: when the pump group operates at reduced power, the injection flow rate will decrease, and in extreme cases, the liquid level in the storage tank will continue to drop; the larger the liquid level margin, the longer the pump group can operate continuously at low power. The calculated maximum response duration... Response duration that does not meet grid-side requirements If required, the control system will issue a low liquid level warning, which can be followed. Renegotiate the response time with the power grid, or wait for the liquid level conditions to improve before participating in the response.

[0050] Step S23 is a preliminary check of the power reduction margin. The control system calculates the current total power of the pump group. Minimum total power required to maintain minimum flow rate The difference is defined as the power reduction margin. The minimum total power required to maintain the minimum flow rate refers to the minimum operating power that the pump group needs to maintain under the premise of ensuring that the pipeline network does not lose pressure. It is usually taken as the sum of the power of the pumps maintaining the minimum number of operating pumps running at the minimum allowable speed. If the power reduction margin is 0, it means that the current operating state of the pump group is at the safety bottom line and does not have the power conditions to participate in demand response, and the control system terminates the scheduling. This judgment also ensures that the normalization formulas involved in subsequent steps S3 and S4 will not have a denominator of 0, ensuring the legality of the mathematical operation. If the power reduction margin is greater than 0 but less than the target response power, If the power reduction margin is not less than the target response power, the control system will enter partial response mode, using the available power reduction margin as the effective target response power, meaning it will achieve the reduction target as much as possible within the safe limits allowed by the pump group. Then the target response power As the effective target response power. After completing the safety constraint pre-check of the above three sub-steps, the control system transmits the determined effective target response power to step S3 to start the optimization scheduling calculation.

[0051] Step S3 involves establishing a bi-objective mixed-integer nonlinear constraint optimization scheduling model and solving for the optimal scheduling scheme. Step S3 is the core innovation of this invention. It establishes a normalized bi-objective optimization framework, incorporating the minimization of pipeline pressure deviation and response deviation into a unified objective function, and outputs the optimal scheduling scheme using hard constraints and a two-stage solution method. Step S3 sequentially includes four sub-steps: S31, S32, S33, and S34.

[0052] Step S31 involves establishing the power-speed relationship, flow-speed relationship, and pipeline pressure estimation model for each pump, and constructing the mapping relationship between decision variables and pipeline pressure and response power.

[0053] Regarding the power-speed relationship, the centrifugal pump similarity law is a fundamental law in fluid machinery engineering, describing the proportional relationship between the performance parameters of geometrically similar centrifugal pumps operating at different speeds. According to the cubic law within the centrifugal pump similarity law, the pump's shaft power is proportional to the cube of its rotational speed. Therefore, the... The pump at actual speed Operating power It can be represented as

[0054] ;

[0055] in, For the first The pump at the speed Operating power (kW); For the first Rated power (kW) of the pump; For the first The actual rotational speed (rpm) of the pump; For the first The rated speed (rpm) of the pump; In this formula, the shaft power of the variable frequency centrifugal pump changes with the cube of the speed. When the speed decreases, the power will decrease rapidly in a cube-shaped proportion. This means that the pump group can generate a considerable power reduction by appropriately reducing the speed, which provides a physical basis for participating in demand response.

[0056] Regarding the flow-speed relationship, according to the linear law in the similarity law of centrifugal pumps, the pump's output flow rate is directly proportional to its speed. Therefore, the first... The pump at actual speed Output flow rate It can be represented as

[0057] ;

[0058] in, For the first The pump at the speed Output flow rate (m) 3 / s); For the first Rated flow rate of the pump (m³) 3 / s). This formula shows that the pump flow rate is linearly related to the pump speed, and the total injection flow rate after adjusting the pump speed can be directly calculated based on this.

[0059] In terms of pipeline pressure estimation, a pipeline pressure estimation model is established based on the pipeline flow balance relationship. Within a scheduling cycle, when the total flow injected into the pipeline by the pump group exceeds the pipeline's demand flow, liquid will accumulate in the pipeline, causing the pipeline pressure to rise; conversely, when the total injected flow is less than the demand flow, the pipeline pressure will decrease. Based on the above flow balance principle, the estimated pipeline pressure under the scheduling scheme is... It can be represented as:

[0060] ;

[0061] in, This represents the estimated pipeline pressure (MPa) under the dispatching scheme. The measured pressure of the main pipe is currently measured in MPa. The hydraulic resistance coefficient of the pipeline network (MPa·s / m) 3 This characterizes the change in static pressure in the pipeline network caused by a unit excess injection volume, and is obtained through pipeline filling test calibration. The total injection flow rate (m³) of each operating pump under the scheduling scheme 3 / s), equal to the start / stop status of each pump. With corresponding traffic Summing of products; Current pipeline demand flow (m³) 3 The pressure ( / s) can be estimated from real-time pressure change trends and historical operating data. This model is a steady-state linear approximation and is suitable for pressure trend prediction within a scheduling cycle of 1 minute. Pipeline hydraulic resistance coefficient These are inherent physical parameters of the pipeline network itself. Their values ​​depend on the pipe diameter, pipe length, pipe material, and topology of the pipeline network. They are basically stable after the pipeline network is built and can be used for a long time after being calibrated once through a liquid filling test.

[0062] With the establishment of the above three models, the mapping relationship between decision variables and objective variables in the optimized scheduling model is now complete. Given any set of start-stop state vectors... and rotational speed vector This allows for the calculation of the corresponding total power of the pump group, total injection flow rate, estimated pipeline pressure, and actual response power, laying the foundation for establishing the objective function in the subsequent step S32.

[0063] Step S32 is to use the start / stop state vector of each pump and rotational speed vector For the decision variables, a normalized bi-objective optimization function is established. In the scheduling problem of multiple pump groups participating in demand response, there are two conflicting optimization objectives. The first objective is to minimize the pipeline pressure deviation, that is, to ensure that the pipeline pressure under the scheduling scheme is as close as possible to the set value to ensure the safe and stable operation of the pipeline network. The second objective is to minimize the response deviation, that is, to ensure that the actual power reduction of the pump group is as close as possible to the target response power required by the grid side. There is an inherent contradiction between these two objectives, because increasing the power reduction will reduce the pump group output, resulting in a decrease in the injection flow rate, and the pipeline pressure will deviate from the set value. In addition, the dimension of pipeline pressure deviation is MPa, while the dimension of response deviation is kW. The two have different dimensions. If they are directly weighted and summed without normalization, the weight coefficients do not have a clear physical meaning, and there is no reference benchmark when adjusting the project.

[0064] To address the aforementioned issue of dimensional inconsistency, this invention normalizes the two objective function terms separately, transforming them into dimensionless relative deviations, and then sums them using weighted coefficients. The normalized bi-objective optimization function is as follows:

[0065] ;

[0066] in, The objective function value is a dimensionless quantity. This is the start / stop state vector for each pump. Each element The value can be 0 or 1, where 0 indicates shutdown and 1 indicates operation. This represents the total number of pumps in the pump group. For the speed vector of each pump, ; This is the weighting coefficient for pipeline pressure deviation; For the response deviation weighting coefficient, and ; The output value (MPa) of the pipeline pressure estimation model established in step S31 under the current scheduling scheme; The setpoint for the pipeline pressure (MPa) is usually taken as the rated pressure of the pipeline. The allowable pressure range (MPa) of the pipeline network is equal to the upper limit of the safe pressure of the pipeline network. With pipeline pressure safety lower limit The difference is used to normalize the pressure deviation to a dimensionless value; The actual response power (kW) of the scheduling scheme; The effective target response power (kW) determined in step S2 is used to normalize the response deviation to dimensionless; in the above formula, the actual response power... The calculation method is based on the total power of the pump group before the demand response begins. As a baseline value, subtract the sum of the power of all operating pumps under the current scheduling scheme, that is:

[0067] ;

[0068] in, To respond to the reference value of the total power of the pump group (kW) before the response, it is locked at the start of the response and remains unchanged throughout the entire response duration; This is a summation operation over all pumps; For the first The start / stop status of the pump; For the first The pump at the speed Operating power (kW) at the following rate. When the pump stops. The power of this pump is not included in the sum of the operating pump power; when the pump is running The pump power is included in the summation.

[0069] After normalization, the first term of the objective function This represents the proportion of the pipeline pressure deviation relative to the allowable pressure range of the pipeline network. Its value range is typically between -1 and 1, and its squared value is between 0 and 1. (The second term of the objective function) This represents the ratio of the response power deviation to the target response power, and is also a dimensionless value. After normalizing both items to the same order of magnitude, the weighting coefficient... and The numerical value directly reflects the relative priority of the two targets. When At times, the optimizer focuses more on maintaining stable pipeline pressure; when At this time, the optimizer focuses more on achieving the response objective. This normalization process gives the engineering tuning of the weight parameters an intuitive physical meaning.

[0070] Step S33 involves setting hard constraints. Hard constraints are rigid conditions that cannot be violated during the optimization process; any scheduling scheme that does not satisfy the hard constraints is considered an infeasible solution and is excluded. The hard constraints set in this invention include the following five categories.

[0071] Category 1 involves upper and lower limits of pipeline pressure safety constraints, requiring the estimated pipeline pressure under the dispatching scheme to be within acceptable limits. Not lower than the safe lower limit of pipeline pressure And not exceeding the safe upper limit of pipeline pressure. This constraint ensures that no scheduling scheme will cause the pipeline pressure to exceed the safe range.

[0072] Category 2 is the upper and lower limit constraint of the liquid level in the storage tank, which requires the current liquid level to be within safe limits. Not lower than the safe low liquid level And not higher than the safe high liquid level This constraint prevents the reservoir from overflowing or emptying.

[0073] The third type is the minimum number of operating pumps constraint, which requires that the number of operating pumps, i.e., the sum of the start and stop states of all pumps, not be less than the preset minimum number of operating pumps. This constraint ensures that the pipeline network has enough pumps running at all times to maintain basic replenishment functions.

[0074] Category 4 consists of upper and lower speed limits for each pump, requiring that the speed of each operating pump not be lower than the minimum permissible speed. And not exceeding the maximum permissible speed. When the pump is stopped, its speed is 0. The minimum allowable speed is set because variable frequency centrifugal pumps may experience insufficient flow and poor bearing lubrication when operating at excessively low speeds, making them unsuitable for long-term operation.

[0075] Category 5 is the upper limit constraint on the number of start-stop cycles in adjacent scheduling cycles, requiring that the number of pumps whose start-stop status changes between two adjacent scheduling cycles does not exceed a preset upper limit. The engineering significance of this constraint lies in suppressing frequent pump start-ups and shutdowns. Frequent start-ups and shutdowns accelerate the wear of contactors and electrical components, shortening equipment lifespan. Simultaneously, rapid pump start-ups and shutdowns cause sudden flow changes in pipelines, generating water hammer effects and jeopardizing pipeline safety. By actively constraining the number of start-ups and shutdowns at the optimization level, schemes requiring frequent switching of pump operating states can be avoided during the scheduling scheme generation phase.

[0076] Step S34 involves solving the above optimization scheduling model using a two-stage method. Since the decision variables include both integer and continuous variables, this optimization problem is a mixed-integer nonlinear programming problem. Start / Stop State Vector Each element in the vector is an integer variable, taking the value 0 or 1; the rotational speed vector. Each element in the equation is a continuous variable, taking values ​​continuously within the allowable speed range. The objective function includes a cubic term representing the speed; therefore, it is a nonlinear and non-convex function, and cannot be directly solved using linear programming or convex optimization methods.

[0077] This invention employs a two-stage method combining enumeration and sequential quadratic programming to solve the problem. The first stage involves enumerating start-stop combinations, traversing all legal start-stop combinations that simultaneously satisfy the minimum number of operating pumps constraint and the upper limit constraint on the number of start-stop cycles. Since the upper limit constraint on the number of start-stop cycles limits the number of pumps whose start-stop status changes between adjacent scheduling cycles, a large number of start-stop combinations that do not satisfy this constraint are excluded, resulting in a number of legal candidate schemes that is far less than the total number of combinations, and the computational complexity of the enumeration is controllable.

[0078] The second stage is speed optimization. For each valid start-stop combination obtained in the first stage, the start-stop state is fixed. Unchanged, only the speed vector For continuous decision variables, under constraints of upper and lower speed limits and pipeline pressure, a sequential quadratic programming method is used for iterative solution. Sequential quadratic programming is a mature numerical method in the field of nonlinear constrained optimization. Its basic idea is to approximate the original nonlinear optimization problem at the current iteration point using a second-order Taylor expansion in each iteration, construct and solve a quadratic programming subproblem, obtain the search direction and step size, update the iteration point, and repeat the above process until the following conditions are met:

[0079] The Karush-Kuhn-Tucker optimality condition. Sequential quadratic programming exhibits superlinear convergence speed. For the pump group scale involved in this invention, using the current actual operating speed of each pump as the initial iteration point, it typically converges to a local optimum satisfying the optimality condition after several iterations. The sequential quadratic programming method can be implemented using mature open-source numerical optimization libraries, requiring no dedicated computing hardware.

[0080] After traversing all legal start-stop combinations and optimizing the speed for each combination, the control system compares the objective function values ​​corresponding to each combination and selects the scheme with the smallest objective function value as the optimal scheduling scheme for the current scheduling cycle. This scheme includes the optimal start-stop state and optimal speed command for each pump.

[0081] Step S4 involves dynamically adjusting the weight coefficients and updating the scheduling scheme based on the response progress during the demand response execution process. After the optimal scheduling scheme is output in the initial scheduling cycle in step S3, the pump group executes according to this scheme. However, during the continuous execution of the demand response, operating parameters such as pipeline demand flow and storage tank level may change, causing the actual response power to deviate from the target. If a fixed weight coefficient is always used, it may not be possible to increase the tracking effort of the response target in time when the response progress lags behind, and it may overemphasize response accuracy and ignore the stability of pipeline pressure when the response progress exceeds the target. Therefore, this invention designs an adaptive weight adjustment mechanism based on the hyperbolic tangent function, which dynamically adjusts the dual-target weight coefficients according to the current response progress in each scheduling cycle. Step S4 includes three sub-steps: S41, S42, and S43.

[0082] Step S41 involves calculating the time-period average response power up to the current moment. In each scheduling cycle, the control system calculates the cumulative arithmetic mean of the actual response power for each historical scheduling cycle to obtain the power up to the current moment. Average response power over time The time-period average response power reflects the overall response completion level from the start of the response to the current moment. Using a cumulative arithmetic average instead of an instantaneous value can smooth the fluctuations in individual scheduling cycles and provide a stable driving signal for weight adjustment.

[0083] Step S42 involves updating the response deviation weighting coefficient based on the deviation between the average response power over the time period and the effective target response power, using the following formula:

[0084] ;

[0085] in, This is the response deviation weighting coefficient updated in the current scheduling cycle; This is a truncation function that restricts the calculation result of the first term within the parentheses to the lower bound. and upper limit When the calculation result is below the lower limit, the lower limit value is output; when the calculation result is above the upper limit, the upper limit value is output; when the calculation result is between the lower limit and the upper limit, the original value is output. To balance the basic weights of the operating conditions, this represents the weight value when the response progress is exactly equal to the target. This limits the maximum range of weight deviation from the base value, representing the maximum adjustment range for the weight. It is a hyperbolic tangent function, whose output range is a continuous smooth curve between -1 and 1. When the independent variable is 0, the output is 0; when the independent variable approaches positive infinity, the output approaches 1; and when the independent variable approaches negative infinity, the output approaches -1. This is the response sensitivity gain coefficient, used to adjust the sensitivity of the weights to response deviations; The effective target response power (kW); As of the current moment The average response power (kW) over the time period; This is the lower limit of the response deviation weight; This represents the upper limit of the response deviation weight.

[0086] The driving logic of this formula is as follows. When the response progress is lagging, i.e., the average response power over the time period... Below the effective target response power Difference If the value is positive, after mapping with the hyperbolic tangent function, the output value is positive, such that... If the error is increased, the optimizer will focus more on reducing the response deviation in the next scheduling cycle, i.e., increasing the power reduction to catch up with the response progress. When the response progress is normal, i.e., the average response power over the period is close to the target, the difference is close to 0, and the hyperbolic tangent function output is close to 0. Basic weights for near-equilibrium operating conditions The two objectives remain in equilibrium. When the response exceeds the target, i.e., the average response power over the period is higher than the target, the difference is negative, the hyperbolic tangent function outputs a negative value, and this results in... Reduce, and correspondingly increase the weight of pipeline pressure deviation. With increased weights, the optimizer shifts its focus to maintaining stable pipeline pressure, appropriately relaxing its pursuit of response accuracy. The smoothing characteristic of the hyperbolic tangent function ensures that the weights change continuously with the deviation without abrupt changes, avoiding control oscillations caused by weight step jumps. The truncation of the upper and lower limits of the weights prevents any objective from being completely ignored, ensuring that both pipeline pressure and response objectives are taken into account to a certain extent under any operating condition.

[0087] Regarding the values ​​of each adjustable parameter in the above formula, the basic weight of the balanced working condition. A value between 0.3 and 0.5 is recommended, indicating that in an equilibrium state, the response objective and stress stability each account for approximately half the weight. Maximum weight adjustment range. It is recommended to use a value between 0.2 and 0.4 to ensure that the weight adjustment range is appropriate. Response sensitivity gain coefficient. A value between 1.0 and 3.0 is recommended. A value that is too small will result in sluggish weight adjustment, while a value that is too large may cause drastic changes in weight even with small deviations. Lower limit of response deviation weight. It is recommended to use a value greater than 0, which is the upper limit of the response deviation weight. It is recommended to use values ​​less than 1 to ensure that neither objective is completely ignored. The specific values ​​of the above parameters can be determined through offline simulation based on the actual pipeline network characteristics and response requirements.

[0088] Step S43 involves updating the pipeline pressure deviation weighting coefficients and resolving the optimization scheduling model. Pipeline pressure deviation weighting coefficients ,in This is the pipeline pressure deviation weighting coefficient updated for the current scheduling cycle; 1 is a constant. This is a subtraction operation. The updated... and Substituting the objective function from step S32, the two-stage solution process of step S34 is re-executed to output the optimal scheduling scheme for the current scheduling cycle. Steps S41 to S43 are repeated in each scheduling cycle, forming a rolling optimization and update mechanism that allows the scheduling scheme to continuously adapt to changes in operating conditions. It is important to note that regardless of how the weighting coefficients are adjusted, the hard constraints on the upper and lower safety limits of the pipeline pressure set in step S33 must never be violated, forming the first line of defense for pipeline safety.

[0089] Step S5 involves distributing and executing the scheduling plan and providing a safety fallback. Step S5 comprises two sub-steps, S51 and S52.

[0090] Step S51 involves converting the pump speed commands in the optimal scheduling scheme into analog or digital communication signals and sending them to the corresponding frequency converters. Analog signals typically use 4-20mA current signals, emitted from the analog output channel of the control system and transmitted via signal cables to the analog input port of the frequency converter. Digital communication signals typically use Modbus / TCP or other industrial fieldbus protocols, sent by the control system to the frequency converter via a communication network. Upon receiving the speed commands, each pump frequency converter adjusts its output frequency to drive the pump at the target speed. The operational status feedback signals from each pump frequency converter, including actual speed, actual current, and fault signals, are transmitted back in real-time to the operational status parameter acquisition stage in step S1, forming a closed-loop control system of acquisition-optimization-execution-feedback. This closed-loop structure allows the control system to obtain the latest pump group operational status in each scheduling cycle, ensuring that subsequent scheduling optimizations are based on actual operating conditions.

[0091] Step S52 serves as a safety fallback. Throughout the entire demand response execution, the control system independently monitors the main pipeline pressure in real time, separate from the optimization scheduling processes in steps S3 and S4. When the main pipeline pressure falls below the emergency protection pressure, the control system immediately forces an exit from demand response mode and switches to constant pressure control mode. Constant pressure control mode refers to the operating mode of the pump group with the sole control objective of maintaining constant pipeline pressure. It typically employs a PID controller to adjust the pump speed based on pipeline pressure feedback. The emergency protection pressure falls below the safe lower limit of pipeline pressure set in step S33. As the last line of defense for safety, the hard constraint in step S33 should ensure that the pipeline pressure remains within a safe range during normal operation. The safety fallback protection in step S52 is to cope with abnormal situations such as sensor delays, model estimation errors, or communication interruptions, ensuring that even if unexpected deviations occur during the optimization scheduling process, the pipeline pressure will not continue to drop to a level that endangers system safety. This dual safety mechanism of hard constraint and fallback protection fully guarantees the safety of the pump group when participating in demand response.

[0092] The following describes the specific implementation of the demand-response-based multi-unit water supply pump coordination and optimization control system provided by the present invention.

[0093] The control system of this invention adopts a three-layer architecture to implement the above method, including a perception layer, a decision optimization layer, and an execution layer. The three-layer architecture performs data acquisition, decision calculation, and instruction execution functions from bottom to top, with clear hierarchical division and data interaction between each layer through explicit interfaces.

[0094] The sensing layer includes a signal acquisition and communication module. This module connects to the network pressure sensors, flow sensors, liquid level gauges in the storage tanks, and the status output ports of the pump inverters for each unit. The network pressure sensors are installed on the main network pipe to measure the pressure in real time. The flow sensors are installed on the pump group outlet pipe to measure the total replenishment flow. The liquid level gauges are installed on the storage tanks to measure the current liquid level. The status output ports of each pump inverter provide operating parameters such as the actual speed, actual current, and fault status of each pump. The signal acquisition and communication module aggregates the signals from the sensors and inverters, completing analog signal acquisition, digital signal acquisition, and communication protocol parsing. The module also establishes a communication connection with the grid demand response platform, receiving demand response commands from the grid side, including the target response power and response duration. The signal acquisition and communication module simultaneously performs bidirectional communication functions, transmitting acquired data to the upper layer and forwarding grid commands to the lower layer, serving as the hub for data flow throughout the system.

[0095] The decision optimization layer comprises a state preprocessing module and a multi-objective coordination and scheduling module. The state preprocessing module is connected to the data output of the signal acquisition and communication module, receiving the raw data. Sensor signals in industrial settings are often affected by electromagnetic interference, ambient temperature changes, and other factors, potentially resulting in noise, spikes, and occasional outliers in the raw data. The state preprocessing module performs digital filtering on the acquired data to remove high-frequency noise, performs outlier detection and removal to exclude data points that significantly deviate from the normal range, and outputs a state estimate after quality assessment. The state preprocessing module is placed in the decision optimization layer rather than the perception layer because data quality assessment and filtering algorithms belong to the data processing and decision support stages, and are more closely related to decision calculation. The preprocessed state estimate is then transmitted to the multi-objective coordination and scheduling module. The multi-objective coordination and scheduling module is connected to both the output of the state preprocessing module and the command output of the signal acquisition and communication module. After receiving preprocessed pump group operating status data and power grid demand response commands, the multi-objective coordinated scheduling module sequentially executes steps S2 (safety constraint pre-check), S3 (normalized bi-objective optimization scheduling model solution), and S4 (adaptive weight adjustment and rolling optimization update), ultimately outputting the start / stop status and speed commands for each pump. The multi-objective coordinated scheduling module is the decision-making center of the entire system, and its operations can be performed on an industrial control computer or a programmable logic controller (PLC) with corresponding computing capabilities.

[0096] The execution layer includes a frequency converter control interface and a safety constraint protection module. The frequency converter control interface connects to the output of the multi-objective coordinated scheduling module, receiving pump speed commands from the optimal scheduling scheme. Simultaneously, the frequency converter control interface connects to the control input of each pump frequency converter, converting the speed commands into corresponding control signals and sending them to each pump frequency converter. The frequency converter control interface supports analog signal output and digital communication signal output to adapt to frequency converters of different models and communication protocols. The operating status feedback signals of each pump frequency converter are transmitted back to the decision optimization layer via the signal acquisition and communication module, forming a closed-loop control loop of perception layer acquisition, decision optimization layer solution, execution layer distribution, and feedback transmission. The safety constraint protection module connects to the signal acquisition and communication module to acquire real-time pipeline main pipe pressure data. The safety constraint protection module operates independently of the multi-objective coordinated scheduling module and is unaffected by the state of the optimized scheduling algorithm. When the pipeline main pipe pressure is lower than the emergency protection pressure, the safety constraint protection module sends a forced exit command to the frequency converter control interface, exiting the demand response mode and switching to constant pressure control mode. The independent operation of the safety constraint protection module ensures that even if the multi-objective coordination and scheduling module experiences software anomalies or calculation timeouts, the pipeline pressure safety is still guaranteed, forming a safety hard constraint fallback protection independent of the optimized scheduling.

[0097] Example 1: This example uses a circulating water makeup pipeline network as the application object. The network consists of four variable frequency centrifugal makeup water pumps of different models: three small pumps P1 to P3 and one large pump P4, numbered P1 to P4, operating in parallel. The rated parameters of each pump are as follows: P1 rated power 1.1kW, rated speed 2850r / min, rated flow rate 0.0025m³ / min. 3 / s; P2 rated power 1.5kW, rated speed 2850r / min, rated flow rate 0.0032m³ / s; 3 / s; P3 rated power 2.2kW, rated speed 2850r / min, rated flow rate 0.0045m³ / s; 3 / s; P4 rated power 4.0kW, rated speed 2900r / min, rated flow rate 0.0080m³ / s; 3 / s. Pipeline pressure setpoint The maximum safe pressure is 0.80 MPa. The lower safety limit is 1.00 MPa. It is 0.60 MPa. The maximum operating pressure is 0.40 MPa, and the emergency protection pressure is 0.55 MPa. The minimum number of operating units is limited to 2, and the maximum number of start-stop cycles is [not specified]. The value is 1. The scheduling period is 60 seconds, and the response duration is [missing information]. The timeframe is 30 minutes. Adaptive weight parameters: It is 0.40. It is 0.30. It is 2.0. It is 0.15. It is 0.85.

[0098] According to step S1, the host computer receives the power grid demand response command and obtains the target response power. It has a power output of 3.0kW and a response duration of [duration missing]. The duration is 30 minutes. Simultaneously, the initial operating status of the pump group was collected: all four pumps were running, with P1, P2, and P3 rotating at approximately 2400 r / min, and P4 rotating at approximately 2400 r / min. The total power of the pump group was... The pressure of the main pipeline is 5.13kW. The pressure is 0.82 MPa, and the liquid level in the storage tank is... It is 2.8m.

[0099] Perform a safety constraint pre-check according to step S2. In step S21, the current main pipeline pressure of 0.82 MPa is higher than the lower safety limit of 0.60 MPa, and the pressure margin pre-check passes. In step S22, determine the maximum response duration that the liquid level can support. The time limit is 56 minutes, which is greater than 30 minutes, and the liquid level margin pre-check is passed. In step S23, the current total power of the pump group is 5.13kW. Based on the minimum total power required for P3 and P4 to run at the minimum allowable speed of 1500r / min, which is approximately 0.87kW, the power margin can be reduced by 4.26kW, which is greater than the target of 3.0kW. 3.0kW is taken as the effective target response power.

[0100] A normalized bi-objective optimization scheduling model is established according to step S3. In step S31, the power-speed relationship of each pump P1 to P4 is established based on the cubic law of centrifugal pumps, and the flow-speed relationship of each pump is established based on the linear law. A pipeline pressure estimation model is also established. Since the four pumps are of different models, their respective rated power, rated speed, and rated flow parameters are used in the power-speed and flow-speed relationships, and they are not mixed. In step S32, the allowable pressure range of the pipeline is used. The objective functions are established by normalizing the two objective terms, namely 0.40 MPa and the effective target response power of 3.0 kW. In step S33, constraints are set for pipeline pressure, liquid level, minimum number of operating pumps (≥2), upper and lower limits for pump speed, and upper limit for start / stop cycles. There are a total of 5 types of hard constraints. In step S34, a two-stage method is used to solve the problem.

[0101] According to step S4, the average response power of each scheduling period is calculated in step S41, the weight coefficients are dynamically adjusted through the hyperbolic tangent function in step S42, the pipeline pressure deviation weight is updated in step S43, and the optimization model is solved again to achieve rolling update.

[0102] According to step S5, the scheduling plan is issued and executed in step S51, and the safety constraint protection module independently monitors the pressure of the main pipeline in step S52.

[0103] After 30 minutes of operation across 30 scheduling cycles, the average response power of the pump group converged to 2.93kW, with a response completion rate of 97.8%. The pressure of the main pipeline remained within the range of 0.72 to 0.88MPa, with a pressure standard deviation of 0.025MPa. A total of 3 pump start-stop switching operations were recorded.

[0104] Example 2: In this example, the pipeline network consists of six identical variable frequency centrifugal water supply pumps operating in parallel. Each pump has a rated power of 2.2kW, a rated speed of 2850r / min, and a rated flow rate of 0.004m³ / min. 3 / s. Initial operating status: All 6 pumps are running, each pump speed is approximately 2450 r / min, and the total power of the pump group is approximately 8.38 kW. Pipeline pressure setpoint: 0.60 MPa, upper safety limit: 0.75 MPa, lower safety limit: 0.45 MPa. The initial pressure is 0.30 MPa, and the emergency protection pressure is 0.40 MPa. The target response power is 5.0 kW, the response duration is 20 minutes, and the scheduling cycle is 60 seconds. Adaptive weight parameters: It is 0.50. It is 0.20. It is 1.5. It is 0.20. The value is 0.80. Minimum number of running units: 3; maximum number of start / stop cycles: [not specified]. The value is 2.

[0105] In step S23, based on the minimum total power of approximately 0.96 kW from the operation of three pumps at 1500 r / min, a power reduction margin of approximately 7.42 kW is obtained, which is greater than the target of 5.0 kW. 5.0 kW is then used as the effective target response power. The remaining steps are the same as in Example 1.

[0106] After 20 minutes of operation across 20 scheduling cycles, the average response power was 4.83 kW, the response completion rate was 96.5%, the pipeline pressure was maintained between 0.54 and 0.66 MPa, the pressure standard deviation was 0.020 MPa, and there were a total of 2 start-stop switching operations.

[0107] Example 3: This example uses the same pipeline system and pump group configuration as Example 1: P1: 1.1kW; P2: 1.5kW; P3: 2.2kW; P4: 4.0kW, with the three smaller pumps and one larger pump operating in parallel. However, the target response power required by the power grid is 5.5kW, exceeding the pump group's power reduction margin of 4.26kW. Following step S23, the effective target response power is set at the power reduction margin of 4.26kW, and the system enters partial response mode. Adaptive weighting parameters: It is 0.45. It is 0.35. It is 2.5. It is 0.10. It is 0.90.

[0108] After 30 minutes of operation, the average response power was 4.06kW, with a response completion rate of 95.2% relative to the effective target. The pipeline pressure was maintained between 0.65 and 0.90 MPa, with a pressure standard deviation of 0.030 MPa. A total of 4 start-stop switching operations were conducted.

[0109] Comparative Example 1: Fixed weighting method; using the exact same pump group configuration as Example 1: P1: 1.1kW; P2: 1.5kW; P3: 2.2kW; P4: 4.0kW, with three small pumps and one large pump, pipeline parameters, and response targets. The power is 3.0kW, the difference being that the adaptive weight adjustment mechanism in step S4 is not used; instead, a fixed weight is used throughout the entire response process. 0.50 The value remains unchanged at 0.50. The remaining steps are the same as in Example 1.

[0110] After 30 minutes of operation, the average response power was 2.66kW, the response completion rate was 88.6%, the pipeline pressure fluctuation range was 0.68~0.92MPa, the pressure standard deviation was 0.045MPa, and there were a total of 3 start-stop switching operations.

[0111] Comparative Example 2: Non-normalized weighted method; using the same pump group configuration as in Example 1: P1: 1.1kW; P2: 1.5kW; P3: 2.2kW; P4: 4.0kW and response target. The value is 3.0kW, the difference being that: in step S32, normalization is not performed, and the original squared deviations in MPa and kW are directly weighted and summed. The adaptive weight adjustment mechanism is retained, and the remaining steps are the same.

[0112] After 30 minutes of operation, the average response power was 2.47kW, the response completion rate was 82.3%, the pipeline pressure fluctuated violently, with a fluctuation range of 0.62 to 0.98MPa and a pressure standard deviation of 0.065MPa. There were two scheduling cycles in which the pipeline pressure approached the safety limit, and a total of 5 start-stop switchings were carried out.

[0113] Comparative Example 3: No start-stop constraint method; using the same pump group configuration as Example 1: P1: 1.1kW; P2: 1.5kW; P3: 2.2kW; P4: 4.0kW, response target The optimization framework for 3.0kW and normalized dual objectives was adopted, the difference being that the upper limit constraint on the number of start-stop cycles was removed in step S33. The remaining steps were the same. After 30 minutes of operation, the average response power was 2.90kW, the response completion rate was 96.5%, the pipeline pressure was maintained between 0.71 and 0.89 MPa, and the pressure standard deviation was 0.028 MPa, but the cumulative number of start-stop switching was as high as 14.

[0114] Experiment Example 1: Comparative Experiment of Response Power Tracking Performance; This experiment compares the power tracking performance of Example 1, Example 2, and Comparative Example 1 during the response execution process. The following indicators are used: the ratio of average response power to effective target response power during the response completion rate period, the steady-state convergence time of the response, the moment when the response power first enters the range of ±5% of the target value and remains there, and the standard deviation of the response power.

[0115] The experiment was conducted under the same initial conditions, and the actual response power and adaptive weights were recorded for each scheduling cycle. The process of change. Experimental results are as follows: Figure 1 As shown.

[0116] Figure 1In Figure (a), the response power tracking curves for each scheme are shown. The horizontal axis represents the scheduling cycle in minutes, with each scheduling cycle being 60 seconds (1 minute). Example 1 converges to the target value ±5% range (2.85–3.15 kW) within the 8th minute and maintains it thereafter. The final average response power for the time period is 2.93 kW, with a power standard deviation of 0.12 kW. Example 2 has a response duration of 20 minutes, so its curve only covers the interval from the 1st to the 20th minute. It converges to the target ±5% range (4.75–5.25 kW) within the 6th minute, with an average response power of 4.83 kW and a standard deviation of 0.10 kW. Comparative Example 1, due to its fixed weights, cannot adjust the tracking strength according to the response progress and fluctuates between 2.44 and 2.82 kW with a standard deviation of approximately 0.20 kW. It fails to converge to the target ±5% range (2.85–3.15 kW) within 30 minutes. Figure 1 In (b), the weights of Example 1 and Example 2 are given. The curve is dynamically adjusted, and the fixed-weight reference line of Comparative Example 1 is used, with the horizontal axis representing the scheduling period (min). In the initial response phase, because the actual response power is lower than the target, the input of the hyperbolic tangent function in step S42 is adjusted. If it is a positive value, the mapping will make As the base value increases to the 0.60–0.65 range, the optimizer is driven to increase its power reduction; once the response stabilizes... It falls back to near the baseline value. Similarly, the weight curve of Example 2 only covers up to the 20th minute.

[0117] from Figure 1 It can be seen that the response completion rates of Examples 1 and 2, which adopt the adaptive weight adjustment mechanism, are 9.2 and 7.9 percentage points higher than those of Comparative Example 1, respectively, with significantly improved convergence speed and significantly reduced tracking fluctuations. In step S4 of this invention, the hyperbolic tangent function maps the normalized deviation between the time-period average response power and the target into a continuous smooth signal between -1 and 1. This signal, superimposed on the basic weights, changes the relative priority of the two optimization objectives in the objective function. When the response lags, the weights automatically bias towards the response objective, prompting the optimizer to select a more aggressive power reduction scheme. When the response tends to meet the target, the weights automatically return to equilibrium, avoiding excessive reduction that could cause pipeline pressure instability.

[0118] Experiment Example 2: Comparative Experiment on Pipeline Pressure Stability; This experiment compares the pipeline pressure stability of Example 1 and Comparative Example 2 during the response execution process. The following indicators were used: the maximum absolute value of the difference between the measured pressure and the set value, the standard deviation of pipeline pressure, the number of pressure exceeding limits, and the number of scheduling cycles in which the pressure exceeded the safe range.

[0119] Experimental results are as follows Figure 2 As shown. Figure 2In Figure (a), the pressure time-series curve of the main pipeline in Example 1 shows that the pressure fluctuates steadily between 0.72 and 0.88 MPa, with a maximum deviation of 0.08 MPa, a standard deviation of 0.025 MPa, and 0 pressure exceedances. The entire curve falls within the safety zone formed by the red dashed line of the upper safety limit and the magenta dotted line of the lower safety limit. Figure 2 Figure (b) shows the time-series pressure curve of the main pipeline in Comparative Example 2. Due to the lack of normalization, the squared value of the response deviation in kW is much larger than the squared value of the pressure deviation in MPa. The optimizer actually prioritizes the response deviation and ignores the pipeline pressure deviation, leading to a deterioration in pressure control performance. The fluctuation range is 0.62–0.98 MPa, with a maximum deviation of 0.18 MPa and a standard deviation of 0.065 MPa. Around min 11 and min 22, the pipeline pressure rises to approximately 0.97 MPa and 0.98 MPa, respectively, approaching the safety limit of 1.00 MPa, indicating two instances of pressure exceeding the limit, marked with red inverted triangles in the figure.

[0120] from Figure 2 It can be seen that the maximum deviation and standard deviation of pipeline pressure in Example 1 were reduced by 55.6% and 61.5% respectively compared with Comparative Example 2, and no pressure exceedance occurred, while Comparative Example 2 showed two pressure exceedance risks approaching the safety upper limit. Normalization, by dividing by their respective reference ranges, maps the two items to dimensionless relative deviations of the same order of magnitude, making the weighting coefficients truly have the physical meaning of regulating the relative priority of the two objectives.

[0121] Experiment Example 3: Comparison Experiment of Start-Stop Frequency and Equipment Protection; This experiment compares the pump start-stop switching behavior during the response execution process with Example 1, Example 3, and Comparative Example 3. Frequent start-stop cycles accelerate contactor electrical wear and cause water hammer impact in the pipeline. The following indicators were used: cumulative number of start-stop switching cycles, maximum number of consecutive switching cycles for a single pump, and maximum transient pressure impact in the pipeline network.

[0122] Experimental results are as follows Figure 3 As shown. Figure 3 (a) is a grouped bar chart showing the cumulative number of start-stop switching and the maximum number of consecutive switching for a single pump for each scheme. Example 1 has a cumulative total of 3 times and a maximum of 1 time per pump; Example 3 has a cumulative total of 4 times and a maximum of 2 times per pump; and Comparative Example 3 has a cumulative total of 14 times and a maximum of 5 times per pump. Figure 3 (b) is a bar chart of the maximum transient pressure impact of each scheme's pipeline network. Example 1 is 0.03 MPa, Example 3 is 0.04 MPa, and Comparative Example 3 is 0.09 MPa.

[0123] from Figure 3It can be seen that the cumulative number of start-stop cycles in Examples 1 and 3 is reduced by 78.6% and 71.4% respectively compared to Comparative Example 3, and the maximum transient pressure shock is reduced by 66.7% and 55.6% respectively. The start-stop constraint of this invention eliminates candidate combinations where the number of pumps with different start-stop states between adjacent scheduling cycles exceeds the upper limit during the first stage of enumeration in step S34, thus avoiding frequent pump switching from the source of the optimization scheme generation. Rapid pump start-stop causes a step change in flow rate in the pipeline. This sudden change in flow rate is converted into a pressure shock, i.e., water hammer effect, through the hydraulic resistance of the pipeline network. The shock amplitude is proportional to the rate of flow rate change. After limiting the number of start-stop cycles, the flow rate change within a single scheduling cycle is achieved through a continuous gradual change in rotational speed, significantly reducing the pipeline network pressure shock.

[0124] Experiment Example 4: Comprehensive Performance Comparison Experiment; This experiment compares the comprehensive performance of all 3 implementation examples and 3 comparative examples, and presents 3 original indicators: response completion rate (%), standard deviation of pipeline pressure, and cumulative number of start-stop switching.

[0125] Experimental results are as follows Figure 4 As shown. Figure 4 Three subgraphs are used to show the comparison of each scheme on three indicators. Figure 4 (a) is a bar chart showing the response completion rate of all 6 schemes. The 3 implementation schemes are represented by blue bars, and the 3 comparative schemes are represented by red bars. The values ​​are marked at the top of each bar. Figure 4 (b) is a bar chart of the standard deviation of pipeline pressure for all 6 schemes, with color coding consistent with (a); Figure 4 (c) is a bar chart showing the cumulative start-stop switching counts for all six schemes, with consistent color coding. Black dashed lines are used to separate the embodiments from the comparative examples in each sub-chart for easy identification.

[0126] from Figure 4 It can be seen that, Figure 4 In (a), the response completion rates of the three embodiments all exceeded 95%, while Comparative Example 1 and Comparative Example 2 were only 88.6% and 82.3% respectively. Although Comparative Example 3 reached 96.5%, it was insufficient in other indicators. Figure 4 In (b), the pressure standard deviation of the three examples did not exceed 0.030 MPa, while that of Comparative Example 1 and Comparative Example 2 was 0.045 and 0.065 MPa, respectively, which were significantly higher than those of the examples. Figure 4In Example (c), the cumulative number of start-stop cycles for Examples 1-3 are 3, 2, and 4, respectively, while Comparative Example 3 has as many as 14, far exceeding other schemes. A comprehensive analysis of the three sub-figures shows that Comparative Example 1 suffers from a significant decrease in response completion rate due to the lack of adaptive weight adjustment; Comparative Example 2 suffers from the worst response completion rate and pressure stability due to the lack of normalization processing; and Comparative Example 3 suffers from a much higher number of start-stop cycles due to the lack of start-stop constraints. This indicates that the normalized bi-objective optimization framework, adaptive weight adjustment mechanism, and start-stop cycle constraints of this invention synergistically improve the overall system performance, achieving an optimal balance between pipeline safety, response accuracy, and equipment protection.

[0127] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A coordinated optimization control method for multi-unit makeup water pumps based on demand response, characterized in that, Includes the following steps: S1: Receive the grid-side demand response command, obtain the target response power and response duration, and collect the operating status parameters of the multi-unit pump group in real time. The operating status parameters include the speed of each pump, the current total power of the pump group, the pressure of the main pipeline, and the liquid level of the storage tank. S2: Based on the operating status parameters, a safety constraint pre-check is performed, and it is determined in sequence whether the pipeline pressure margin, the liquid level margin of the storage tank, and the power reduction margin meet the conditions for participating in demand response. After the pre-check is passed, the effective target response power is determined. S3: Using the start-stop status and speed of each pump as decision variables, the weighted sum of the normalized pipeline pressure deviation and the normalized response deviation as the objective function, and the pipeline pressure constraint, liquid level constraint and start-stop number constraint as hard constraints, a dual-objective mixed integer nonlinear constraint optimization scheduling model is established, and the optimal scheduling scheme is solved by a two-stage method combining enumeration and sequential quadratic programming. S4: During the demand response execution process, calculate the average response power up to the current time period. Based on the deviation between the average response power and the target response power, dynamically adjust the weight coefficients in the target function using the hyperbolic tangent function, and re-solve the optimized scheduling model according to the scheduling cycle to update the scheduling scheme on a rolling basis. S5: Send the start / stop status and speed commands of each pump to the corresponding frequency converter for execution, and monitor the pressure of the main pipeline in real time. When the pressure of the main pipeline is lower than the emergency protection pressure, force exit the demand response mode and switch to constant pressure control mode.

2. The method according to claim 1, characterized in that, Step S2 includes: S21: Determine whether the current main pipeline pressure is higher than the safe lower limit of the pipeline pressure. If the main pipeline pressure is already at the safe lower limit of the pipeline pressure, it is determined that the pressure conditions for participating in demand response are not met, a prohibition response flag is output and scheduling is terminated to realize pressure margin pre-check. S22: Based on the difference between the current liquid level in the storage tank and the safe low liquid level, the effective cross-sectional area of ​​the storage tank, and the minimum flow rate required to maintain the safety of the pipeline pressure, calculate the maximum response duration that the liquid level can support. If the maximum response duration does not meet the requirement of the response duration, issue a liquid level insufficient warning to realize liquid level reserve pre-check. S23: Calculate the difference between the current total power of the pump group and the minimum total power required to maintain the minimum flow rate as the power margin that can be reduced. If the power margin that can be reduced is zero, it is determined that there is no power condition to participate in demand response and the scheduling is terminated. If the power margin that can be reduced is less than the target response power, the power margin that can be reduced is used as the effective target response power to enter the partial response mode. Otherwise, the target response power is used as the effective target response power. Complete the safety constraint pre-check and pass the effective target response power to step S3.

3. The method according to claim 1, characterized in that, Step S3 includes: S31: Based on the centrifugal pump similarity law, establish the power-speed relationship of each pump according to its rated power, rated speed and actual speed; establish the flow-speed relationship of each pump according to its rated flow rate, rated speed and actual speed; based on the pipeline flow balance relationship, establish a pipeline pressure estimation model according to the current measured main pipe pressure, pipeline hydraulic resistance coefficient and the difference between the total injection flow rate of each operating pump and the current pipeline demand flow rate, thereby constructing the mapping relationship between decision variables and pipeline pressure and response power; S32: Based on the start / stop state vectors of each pump and rotational speed vector For the decision variables, establish the following normalized biobjective optimization function: ； in, The objective function value; This is a start / stop status vector for each pump, where each element takes the value 0 or 1, where 0 indicates shutdown and 1 indicates operation. For each pump speed vector; This is the weighting coefficient for pipeline pressure deviation; For the response deviation weighting coefficient, and ; The output value of the pipeline pressure estimation model established in step S31 under the current scheduling scheme; Set the pipeline pressure value; The allowable pressure range of the pipeline network is equal to the safe upper limit of the pipeline network pressure. With pipeline pressure safety lower limit The difference is used to normalize the pressure deviation to a dimensionless value; The actual response power of the scheduling scheme is equal to the baseline value of the total power of the pump group before the response. The difference between the power of each operating pump and the sum of the power of the pumps under the scheduling scheme; The effective target response power is used to normalize the response deviation to dimensionless; S33: Set hard constraints, including: upper and lower limits of pipeline pressure safety, upper and lower limits of liquid level safety in storage tank, minimum number of operating pumps, upper and lower limits of pump speed, and upper limit of start-stop count in adjacent scheduling cycles. The upper limit of start-stop count restricts the number of pumps whose start-stop status changes between adjacent scheduling cycles from not exceeding a preset upper limit. S34: The optimization scheduling model is solved using a two-stage method: In the first stage, all legal start-stop combinations that simultaneously satisfy the minimum number of operating units constraint and the upper limit of start-stop times constraint are traversed; In the second stage, the start-stop state is fixed for each legal start-stop combination, and the speed vector is used as a continuous decision variable. The sequential quadratic programming method is used to iteratively solve the problem under the speed constraint and pipeline pressure constraint; The scheme with the smallest objective function value among all legal start-stop combinations is selected as the optimal scheduling scheme output.

4. The method according to claim 3, characterized in that: In step S31, the power-speed relationship of each pump is established based on the cubic law of centrifugal pump similarity, that is, the ratio of the power of each pump at the actual speed to its rated power is equal to the cube of the ratio of the actual speed to the rated speed; the flow-speed relationship of each pump is established based on the linear law of centrifugal pump similarity, that is, the ratio of the flow rate of each pump at the actual speed to its rated flow rate is equal to the ratio of the actual speed to the rated speed; the output value of the pipeline pressure estimation model is equal to the current measured main pipe pressure plus the product of the pipeline hydraulic resistance coefficient and the difference between the total injection flow rate and the pipeline demand flow rate, wherein the pipeline hydraulic resistance coefficient is calibrated by the pipeline filling test; in step S34, the second stage uses the current actual operating speed of each pump as the initial iteration point of the sequential quadratic programming method.

5. The method according to claim 1, characterized in that, Step S4 includes: S41: In each scheduling cycle, calculate the time-averaged response power up to the current moment, that is, take the cumulative arithmetic average of the actual response power of each historical scheduling cycle. S42: Based on the deviation between the average response power over the time period and the effective target response power, update the response deviation weighting coefficient according to the following formula: ； in, This is the response deviation weighting coefficient updated for the current scheduling cycle; This is a truncation function that restricts the calculation result of the first term within the parentheses to a lower bound. and upper limit between; To balance the basic weights of operating conditions; This represents the maximum adjustment range for the weights. It is the hyperbolic tangent function; The gain coefficient is the response sensitivity coefficient. Effective target response power; As of the current moment The average response power over the time period; This is the lower limit of the response deviation weight; The upper limit of the weight for the response deviation; S43: Update the pipeline pressure deviation weighting coefficient ,in The updated pipeline pressure deviation weighting coefficient for the current scheduling cycle; the updated... and Substitute the objective function from step S3 into the solution, re-solve the optimized scheduling model, and output the optimal scheduling scheme for the current scheduling period to achieve rolling optimization and updating.

6. The method according to claim 1, characterized in that, Step S5 includes: S51: After converting the pump speed commands in the optimal scheduling scheme into analog signals or digital communication signals, the commands are sent to the corresponding frequency converters. The operating status feedback signals of each pump frequency converter after execution are transmitted back to the operating status parameter acquisition link in step S1 in real time, forming a closed-loop control of acquisition, optimization, execution and feedback. S52: During the entire demand response execution, independent of the optimization scheduling process in steps S3 and S4, the pressure of the main pipeline is monitored in real time. When the pressure of the main pipeline is lower than the emergency protection pressure, the demand response mode is immediately forcibly exited and switched to the constant pressure control mode. The emergency protection pressure is lower than the safety lower limit of the pipeline pressure set in step S3.

7. A demand-response-based multi-unit makeup water pump coordination and optimization control system, used to implement the method described in any one of claims 1-6, characterized in that, It includes a perception layer, a decision optimization layer, and an execution layer; among which: The sensing layer includes a signal acquisition and communication module. The signal acquisition and communication module is connected to the status output ports of the pipeline pressure sensor, flow sensor, liquid storage tank level gauge and pump frequency converter of each unit to collect the pipeline main pressure, replenishment flow, liquid storage tank level and pump operating status parameters. The signal acquisition and communication module is also connected to the power grid demand response platform to receive demand response commands. The decision optimization layer includes a state preprocessing module and a multi-objective coordinated scheduling module. The state preprocessing module is connected to the data output of the signal acquisition and communication module and is used to filter and process outliers in the acquired data before outputting a state estimate. The multi-objective coordinated scheduling module is connected to the output of the state preprocessing module and the instruction output of the signal acquisition and communication module, respectively. It is used to perform safety constraint pre-check, solve the normalized bi-objective optimization scheduling model, and perform adaptive weight adjustment, and output the start / stop status and speed instructions of each pump. The execution layer includes a frequency converter control interface and a safety constraint protection module. The frequency converter control interface is connected to the output of the multi-target coordination and scheduling module and to the control input of each pump frequency converter, used to convert speed commands into control signals and send them to each pump frequency converter. The safety constraint protection module is connected to the signal acquisition and communication module to obtain the pressure of the main pipeline. When the pressure of the main pipeline is lower than the emergency protection pressure, it sends a forced exit command to the frequency converter control interface, exits the demand response mode and switches to the constant pressure control mode.

8. The system according to claim 7, characterized in that, The inverter control interface supports analog signal output and digital communication signal output; the operating status feedback signals of each pump inverter are transmitted back to the decision optimization layer through the signal acquisition and communication module, forming a closed-loop control loop of perception layer acquisition, decision optimization layer solution, execution layer distribution, and feedback transmission; the safety constraint protection module operates independently of the multi-objective coordination and scheduling module.

Images