Reactive power compensation system for solving power factor electricity cost of water transfer pump station

Abstract

The invention relates to the technical field of reactive power compensation, in particular to a reactive power compensation system for a water transfer pump station to solve power rate electricity cost, which comprises a signal processing unit, a synchronous phasor measurement unit, a power supply unit, a power supply unit, a power supply unit and a power supply unit, and a power supply unit, and is characterized in that the signal processing unit is used for acquiring three-phase voltage signals and three-phase current signals of each node in a power distribution network of the water transfer pump station according to a synchronous phasor measurement unit; calculating the instantaneous reactive power of the node according to the three-phase voltage signal and the three-phase current signal of the node; and the feature extraction unit is used for determining the reactive power change rate of the node according to the instantaneous reactive power of the node. According to the invention, the reactive compensation putting model is constructed through the compensation putting unit, and the optimal putting decision is made according to the compensation capacity and power loss of the compensation equipment and the decision of the compensation branch, so that the power loss can be effectively reduced, the accurate putting of the compensation action can be ensured, and the power supply quality and the operation efficiency of the power distribution network are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of reactive power compensation, and particularly relates to a reactive power compensation system for a pumping station to solve the power factor electricity cost. Background Art

[0002] A low power factor will cause the power company to increase the costs of power transmission and transformation. Therefore, additional fees may be charged to users with a low power factor, and this kind of fee is the power factor electricity cost.

[0003] Traditional systems often rely on manual or experience-based reactive power adjustment, and cannot achieve real-time and accurate monitoring of voltage and current signals and calculation of reactive power. Due to the lack of real-time feedback, traditional systems often cannot accurately capture the fluctuations and changes of reactive power in the power grid, which leads to low efficiency of reactive power management and is likely to cause grid instability, especially in scenarios with large load fluctuations; moreover, the reactive power compensation decisions of traditional systems are usually based on set rules or empirical formulas, lacking intelligent analysis and optimization, so that accurate compensation decisions may not be made according to the real-time conditions of the power grid, which may lead to under-compensation or over-compensation and cannot optimally manage reactive power; and traditional systems lack a feedback mechanism based on real-time data, and the compensation investment often relies too much on the preset compensation capacity or empirical operation, which may lead to over-compensation or under-compensation of reactive power. When over-compensated, it will cause an increase in power loss and affect grid stability; while under-compensation may lead to voltage instability and reduce the reliability of the power grid; in addition, traditional systems usually lack the ability to optimize compensation investment, cannot flexibly adjust the compensation capacity and compensation method, resulting in large power losses and low compensation efficiency. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the above-mentioned disadvantages of the prior art and provide a reactive power compensation system for a pumping station to solve the power factor electricity cost.

[0005] The technical solution adopted to solve the above technical problem is: a reactive power compensation system for a pumping station to solve the power factor electricity cost, comprising:

[0006] A signal processing unit, which is used to obtain three-phase voltage signals and three-phase current signals of each node in the distribution network of the pumping station according to the synchronized phasor measurement unit, and calculate the instantaneous reactive power of the node according to the three-phase voltage signal and three-phase current signal of the node;

[0007] A feature extraction unit, which is used to determine the reactive power change rate of the node according to the instantaneous reactive power of the node, perform a fast Fourier transform on the three-phase current signal of the node to obtain the harmonic amplitude of the node, and determine the current harmonic distortion rate of the node according to the harmonic amplitude of the node;

[0008] A compensation decision-making unit, which is used to construct a reactive power compensation decision-making agent based on the instantaneous reactive power, reactive power change rate, and current harmonic distortion rate of each node in the distribution network, and solve the reactive power compensation decision-making agent according to the double deep Q network to obtain the compensation decisions of each compensation branch in the distribution network;

[0009] A compensation investment unit, which is used to construct a reactive power compensation investment model based on the compensation capacity of reactive power compensation equipment, the compensation decisions of each compensation branch in the distribution network, and the power losses of each compensation branch in the distribution network, and solve the reactive power compensation investment model to obtain the investment actions of each compensation branch in the distribution network;

[0010] A reactive power compensation unit, which is used to control the reactive power compensation equipment to perform reactive power compensation on each compensation branch in the distribution network according to the investment actions of each compensation branch in the distribution network.

[0011] Preferably, the calculation formula for the instantaneous reactive power of the node is as follows:

[0012]

[0013] where q t represents the instantaneous reactive power of the node at time t, v a (t), v b (t) and v c (t) represent the a, b, and c phase voltage signals of the node at time t, and i a (t), i b (t) and i c (t) represent the a, b, and c phase current signals of the node at time t.

[0014] Preferably, the calculation formula for the reactive power change rate of the node is as follows:

[0015]

[0016] where rv t represents the reactive power change rate of the node at time t;

[0017] The calculation formula for the harmonic amplitude of the node is as follows:

[0018]

[0019] where, represents the harmonic amplitude of the node at time t, n represents the current sampling number, N represents the total number of sampling points, i abc represents the three-phase current signal, and h represents the order of the harmonic;

[0020] The calculation formula for the current harmonic distortion rate of the node is as follows:

[0021]

[0022] Among them, THD t represents the current harmonic distortion rate of the node at time t, H represents the total order of harmonics, represents the three-phase current amplitude of the fundamental wave.

[0023] Preferably, a reactive power compensation decision-making agent is constructed according to the instantaneous reactive power, reactive power change rate, and current harmonic distortion rate of each node in the distribution network, including:

[0024] A state space is constructed according to the instantaneous reactive power, reactive power change rate, current harmonic distortion rate, and power factor of each node in the distribution network, where the expression of the state space is as follows:

[0025] s t =[q t , rv t , THD t , cosφ];

[0026] Among them, s t represents the state space, and cosφ represents the power factor;

[0027] An action space is constructed, where the expression of the action space is as follows:

[0028] a t =[(0) n , (1) n ;

[0029] Among them, a t represents the action space, (0) n represents that the nth compensation branch does not require reactive power compensation, and a t represents that the nth compensation branch requires reactive power compensation;

[0030] A reward function is constructed according to the current harmonic distortion rate and power factor of each node in the distribution network, where the expression of the reward function is as follows:

[0031] r t =10(0.95 - |cosφ - 0.95|) - 0.1∑a t - 5THD t ;

[0032] Among them, r t represents the reward function.

[0033] Preferably, the reactive power compensation decision-making agent is solved according to the double deep Q network to obtain the compensation decisions of each compensation branch in the distribution network, including:

[0034] Initialize the parameters of the first deep Q network and the second deep Q network, and set the discount factor and learning rate;

[0035] At each time step, according to the current state, select an action and execute the action, observe the next state and reward, and store this pair of compensation decisions in the experience replay pool, where the expression of the compensation decision is as follows:

[0036] u t =(s t ,a t ,r t ,s t+1 );

[0037] Among them, u t represents the compensation decision, and s t+1 represents the next state;

[0038] Randomly sample a batch of data from the experience replay pool, and update the parameters of the first deep Q network and the second deep Q network according to the loss function, where the expression of the loss function is as follows:

[0039] L(θ)=(r t +γmax á Q2(s t+1 ,á)-Q1(s t ,a t )) 2 ;

[0040] Among them, L(θ) represents the loss function, and Q1 and Q2 represent the first deep Q network and the second deep Q network.

[0041] Preferably, a reactive power compensation investment model is constructed according to the compensation capacity of the reactive power compensation device, the compensation decisions of each compensation branch in the distribution network, and the power losses of each compensation branch in the distribution network, including:

[0042] Construct an objective function according to the compensation decisions of each compensation branch in the distribution network and the power losses of each compensation branch in the distribution network;

[0043] Determine the constraint conditions according to the compensation capacity of the reactive power compensation device;

[0044] Construct a reactive power compensation investment model according to the objective function and the constraint conditions.

[0045] Preferably, the expression of the objective function is as follows:

[0046]

[0047] Among them, L1 represents the objective function, and κ represents a preset weight, q req and q comp Determine the reactive power demand and the actually provided reactive power of the compensation branch that needs reactive power compensation according to the compensation decision, q nom represents the compensation capacity of the reactive power compensation device, P loss and P base represent the power loss and the reference power of the compensation branch that needs reactive power compensation determined according to the compensation decision.

[0048] Preferably, the expression of the constraint condition is as follows:

[0049]

[0050] Among them, q STATCOM represents the capacity allocated by the static synchronous compensator, q SVG represents the capacity allocated by the static var compensator, q FC represents the capacity allocated by the reactive power compensation capacitor, q STATCOM_MAX represents the maximum capacity of the static synchronous compensator, q SVG_MAX represents the maximum capacity of the static var compensator, q step represents the capacity step of the reactive power compensation capacitor, sat(q req -q STATCOM ,0,q SVG_MAX ) means to limit q req -q STATCOM between 0 and q SVG_MAX , and round represents the rounding function.

[0051] Preferably, solve the reactive power compensation investment model to obtain the investment actions of each compensation branch in the distribution network, including:

[0052] Randomly initialize the individual population, and the position of each individual represents the combination of each decision variable in the reactive power compensation investment model;

[0053] Calculate the output of the objective function according to the combination of each decision variable represented by each individual, and determine the fitness of the individual according to the output of the objective function;

[0054] Randomly generate a position update coefficient, which is a constant ranging from 0 to 1. Determine whether the current iteration number is greater than a preset iteration number threshold, and determine whether the position update coefficient is greater than 0.5. If the current iteration number is greater than the preset iteration number and the position update coefficient is greater than 0.5, update the position of the individual through the first position formula. If the current iteration number is greater than the preset iteration number and the position update coefficient is not greater than 0.5, update the position of the individual through the second position formula;

[0055] Obtain the optimal individual based on the updated positions of all individuals. If the iteration number reaches the maximum value, the iteration ends and the optimal combination of each decision variable in the reactive power compensation placement model is output. Otherwise, the iteration number is incremented by 1.

[0056] Preferably, the expression of the first position formula is as follows:

[0057]

[0058] Among them, X1(t + 1) represents the first position of the individual at the (t + 1)-th iteration, X best (t) represents the position of the current optimal individual, t represents the current iteration number, T represents the total iteration number, X M (t) represents the average position of the current individual at the t-th iteration, S rand represents the position update coefficient;

[0059] The expression of the second position formula is as follows:

[0060] X2(t + 1) = X best (t) * L(D) + X R (t) - S rand ;

[0061] Among them, X2(t + 1) represents the second position of the individual at the (t + 1)-th iteration, L(D) represents the Levy flight distribution function, X R (t) represents the worst position of the current individual at the t-th iteration.

[0062] The beneficial effects of the present invention are as follows: (1) The present invention monitors the voltage and current signals of each node in the distribution network in real time through the signal processing unit, calculates the instantaneous reactive power, and can accurately capture the fluctuations and changes of the reactive power. This real-time reactive power monitoring helps to effectively manage the reactive load in the power grid, avoid the impact of reactive power on the power system, especially in the case of high-power loads such as pumping stations; (2) The present invention extracts key parameters such as the reactive power change rate, harmonic amplitude, and current harmonic distortion rate of the node through the feature extraction unit, and combines the double deep Q-network algorithm in the compensation decision unit to automatically generate accurate reactive power compensation decisions. This method avoids traditional manual adjustment and experience-based compensation decisions, thereby improving the intelligence and accuracy of the decision-making, ensuring the stability and maximum benefit of the reactive power compensation system, and through the compensation investment unit, by constructing a reactive power compensation investment model, according to the compensation capacity, power loss of the compensation device, and the decision of the compensation branch, an optimal investment decision is made. In this way, not only can the power loss be effectively reduced, but also the accurate investment of the compensation action can be ensured, thereby improving the power supply quality and operation efficiency of the distribution network; (3) The present invention can accurately adjust the compensation capacity according to the actual operation state of the power grid, avoid excessive or insufficient reactive power compensation, and accurate compensation investment can avoid the waste of reactive power, reduce the stability problems brought by excessive or insufficient reactive power in the power grid, improve the operation reliability of the power grid, and at the same time can significantly reduce the power loss of the power system, achieving the effect of energy conservation. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the system architecture of the overall system in an embodiment proposed by the present invention.

[0064] Reference numerals: 1, signal processing unit; 2, feature extraction unit; 3, compensation decision unit; 4, compensation investment unit; 5, reactive power compensation unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] Embodiment 1, as Figure 1 shown, a reactive power compensation system for solving the power factor electricity cost of a pumping station proposed by the present invention includes:

[0066] A signal processing unit 1, which is used to obtain the three-phase voltage signals and three-phase current signals of each node in the distribution network of the pumping station according to the synchronized phasor measurement unit, and calculate the instantaneous reactive power of the node according to the three-phase voltage signals and three-phase current signals of the node;

[0067] A feature extraction unit 2, which is used to determine the reactive power change rate of the node according to the instantaneous reactive power of the node, perform a fast Fourier transform on the three-phase current signal of the node to obtain the harmonic amplitude of the node, and determine the current harmonic distortion rate of the node according to the harmonic amplitude of the node;

[0068] A compensation decision-making unit 3, which is used to construct a reactive power compensation decision-making agent according to the instantaneous reactive power, reactive power change rate and current harmonic distortion rate of each node in the distribution network, and solve the reactive power compensation decision-making agent according to the double deep Q network to obtain the compensation decisions of each compensation branch in the distribution network;

[0069] A compensation investment unit 4, which is used to construct a reactive power compensation investment model according to the compensation capacity of the reactive power compensation equipment, the compensation decisions of each compensation branch in the distribution network and the power losses of each compensation branch in the distribution network, and solve the reactive power compensation investment model to obtain the investment actions of each compensation branch in the distribution network;

[0070] A reactive power compensation unit 5, which is used to control the reactive power compensation equipment to perform reactive power compensation on each compensation branch in the distribution network according to the investment actions of each compensation branch in the distribution network.

[0071] In the present invention, the synchronous phasor measurement unit is a device for real-time measurement of voltage and current in the power system, usually used to synchronously capture three-phase voltage signals and three-phase current signals in the power system. It obtains the corresponding power system state by measuring these signals, and is further used to calculate power and other electrical parameters; reactive power is the part of the power system that does not do work. It has nothing to do with the transmission and consumption of electricity, but is very important for the stability and efficiency of the power system. Instantaneous reactive power refers to the reactive power at a certain moment, and its change is closely related to the voltage and current waveforms of the system; the fast Fourier transform is a mathematical tool for analyzing the signal spectrum. By performing the fast Fourier transform on the three-phase current signal of the node, it can be converted into a frequency-domain signal to analyze its harmonic components; harmonics refer to the non-integer multiple frequency components that appear in the current or voltage signal; the double deep Q network is an algorithm based on reinforcement learning, which is used to select the optimal action through a deep neural network during the decision-making process; the reactive power compensation equipment is used to compensate reactive power in the power system.

[0072] Embodiment 2, a reactive power compensation system for a pumping station to solve the power factor electricity cost proposed by the present invention. Compared with Embodiment 1, this embodiment further includes: The calculation formula for the instantaneous reactive power of the node is as follows:

[0073]

[0074] where q t represents the instantaneous reactive power of the node at time t, and v a (t), v b (t) and v c (t) represent the a, b, and c phase voltage signals of the node at time t, and i a (t), i b (t) and ic (t) represents the phase a, b, and c current signals of the node at time t.

[0075] In an optional embodiment, the calculation formula for the reactive power change rate of the node is as follows:

[0076]

[0077] where rv t represents the reactive power change rate of the node at time t;

[0078] The calculation formula for the harmonic amplitude of the node is as follows:

[0079]

[0080] where, represents the harmonic amplitude of the node at time t, n represents the current sampling number, N represents the total number of sampling points, i abc represents the three-phase current signal, and h represents the order of the harmonic;

[0081] The calculation formula for the current harmonic distortion rate of the node is as follows:

[0082]

[0083] where THD t represents the current harmonic distortion rate of the node at time t, H represents the total order of the harmonics, represents the three-phase current amplitude of the fundamental wave.

[0084] It should be noted that the reactive power change rate is the speed of the reactive power change of a certain node (such as a substation, a load point, etc.) in the power system at a specific moment. The reactive power change rate reflects the dynamic change of the reactive power, and it is usually related to the stability of the power grid; harmonics refer to the frequency components that appear in the current or voltage waveform, usually integer multiples of the fundamental frequency. The harmonic amplitude describes the intensity or amplitude of these harmonic components; the harmonic order refers to the individual frequency components in the current or voltage signal, usually integer multiples of the fundamental frequency; the current harmonic distortion rate is an index used to quantify the degree of distortion or aberration caused by the harmonic components in the current waveform. The current harmonic distortion rate represents the influence of the harmonic components on the overall shape of the current signal; the fundamental wave refers to the main frequency component in the current or voltage signal, which is the basic oscillation frequency of the system.

[0085] In an optional embodiment, a reactive power compensation decision-making intelligent agent is constructed according to the instantaneous reactive power, reactive power change rate, and current harmonic distortion rate of each node in the distribution network, including:

[0086] Construct a state space based on the instantaneous reactive power, reactive power change rate, current harmonic distortion rate, and power factor of each node in the distribution network. The expression of the state space is as follows:

[0087] s t =[q t ,rv t ,THD t ,cosφ];

[0088] Among them, s t represents the state space, and cosφ represents the power factor;

[0089] Construct an action space. The expression of the action space is as follows:

[0090] a t =[(0) n ,(1) n ;

[0091] Among them, a t represents the action space, (0) n means that the nth compensation branch does not require reactive power compensation, and a t means that the nth compensation branch requires reactive power compensation;

[0092] Construct a reward function based on the current harmonic distortion rate and power factor of each node in the distribution network. The expression of the reward function is as follows:

[0093] r t =10(0.95 - |cosφ - 0.95|) - 0.1∑a t - 5THD t ;

[0094] Among them, r t represents the reward function.

[0095] In an alternative embodiment, solve the reactive power compensation decision-making agent according to the double deep Q network to obtain the compensation decisions of each compensation branch in the distribution network, including:

[0096] Initialize the parameters of the first deep Q network and the second deep Q network, and set the discount factor and learning rate;

[0097] At each time step, according to the current state, select an action and execute this action, observe the next state and reward, and store this pair of compensation decisions in the experience replay pool. The expression of the compensation decision is as follows:

[0098] u t =(s t ,a t ,r t ,st+1 );

[0099] wherein, u t represents the compensation decision, and s t+1 represents the next state;

[0100] Randomly sample a batch of data from the experience replay pool, and update the parameters of the first deep Q-network and the second deep Q-network according to the loss function. The expression of the loss function is as follows:

[0101] L(θ) = (r t + γ max á Q2(s t+1 , á) - Q1(s t , a t )) 2 ;

[0102] wherein, L(θ) represents the loss function, and Q1 and Q2 represent the first deep Q-network and the second deep Q-network.

[0103] It should be noted that the deep Q-network is an algorithm that combines deep learning and reinforcement learning. It approximates the Q-value function through a neural network, thereby solving the problem of dealing with high-dimensional state spaces in traditional Q-learning. In reinforcement learning, the Q-value (state-action value function) represents the expected reward that can be obtained by taking a certain action in a given state; random sampling is to randomly select a batch of sample data (i.e., a quadruple of a group of states, actions, rewards, and the next state) from the experience replay pool for training the Q-network. This helps to break the temporal correlation between data and makes the training more stable.

[0104] In an optional embodiment, a reactive power compensation investment model is constructed based on the compensation capacity of the reactive power compensation device, the compensation decisions of each compensation branch in the distribution network, and the power losses of each compensation branch in the distribution network, including:

[0105] Construct an objective function based on the compensation decisions of each compensation branch in the distribution network and the power losses of each compensation branch in the distribution network;

[0106] Determine the constraint conditions according to the compensation capacity of the reactive power compensation device;

[0107] Construct a reactive power compensation investment model based on the objective function and the constraint conditions.

[0108] In an optional embodiment, the expression of the objective function is as follows:

[0109]

[0110] wherein, L1 represents the objective function, and κ represent preset weights, q reqand q comp Determine the reactive power demand and the actually provided reactive power of the compensation branch that requires reactive power compensation according to the compensation decision, q nom Represents the compensation capacity of the reactive power compensation device, P loss and P base Represents the power loss and the reference power of the compensation branch that requires reactive power compensation determined according to the compensation decision.

[0111] It should be noted that power loss refers to the energy loss generated due to the resistance of wires, electrical equipment, etc. in the power system. In the power grid, power loss usually refers to the part that is converted into heat due to the resistance effect when current passes through transmission lines or transformers.

[0112] In an alternative embodiment, the expression of the constraint condition is as follows:

[0113]

[0114] where q STATCOM Represents the capacity allocated by the static synchronous compensator, q SVG Represents the capacity allocated by the static var compensator, q FC Represents the capacity allocated by the reactive power compensation capacitor, q STATCOM_MAX Represents the maximum capacity of the static synchronous compensator, q SVG_MAX Represents the maximum capacity of the static var compensator, q step Represents the capacity step of the reactive power compensation capacitor, sat(q req -q STATCOM ,0,q SVG_MAX ) means restricting q req -q STATCOM between 0 and q SVG_MAX , and round represents the rounding function.

[0115] In an alternative embodiment, solve the reactive power compensation investment model to obtain the investment actions of each compensation branch in the distribution network, including:

[0116] Randomly initialize the individual population, and the position of each individual represents the combination of each decision variable in the reactive power compensation investment model;

[0117] Calculate the output of the objective function according to the combination of each decision variable represented by each individual, and determine the fitness of the individual according to the output of the objective function;

[0118] Randomly generate a position update coefficient, which is a constant with a range of 0 to 1. Determine whether the current iteration number is greater than the preset iteration number threshold, and determine whether the position update coefficient is greater than 0.5. If the current iteration number is greater than the preset iteration number and the position update coefficient is greater than 0.5, update the position of the individual through the first position formula. If the current iteration number is greater than the preset iteration number and the position update coefficient is not greater than 0.5, update the position of the individual through the second position formula;

[0119] Obtain the optimal individual based on the updated positions of all individuals. If the iteration number reaches the maximum value, the iteration ends and the optimal combination of each decision variable in the reactive power compensation investment model is output. Otherwise, the iteration number is incremented by 1.

[0120] It should be noted that in the evolutionary algorithm or particle swarm optimization algorithm, the population is composed of multiple individuals. Each individual represents a possible solution and evolves continuously through operations such as selection, crossover, and mutation during the iteration process; position is an important concept in the evolutionary algorithm. The position of each individual in the solution space represents the specific values of a set of decision variables.

[0121] In an alternative embodiment, the expression of the first position formula is as follows:

[0122]

[0123] where X1(t + 1) represents the first position of the individual at the (t + 1)-th iteration, X best (t) represents the position of the current optimal individual, t represents the current iteration number, T represents the total iteration number, X M (t) represents the average position of the current individual at the t-th iteration, S rand represents the position update coefficient;

[0124] The expression of the second position formula is as follows:

[0125] X2(t + 1) = X best (t)*L(D) + X R (t) - S rand ;

[0126] where X2(t + 1) represents the second position of the individual at the (t + 1)-th iteration, L(D) represents the Levy flight distribution function, X R (t) represents the worst position of the current individual at the t-th iteration.

[0127] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the knowledge scope of those skilled in the relevant art.

Claims

1. A reactive power compensation system for solving the power factor electricity cost of a pumping station, characterized in that, Including: A signal processing unit (1) for obtaining three-phase voltage signals and three-phase current signals of each node in the distribution network of the pumping station according to a synchronized phasor measurement unit, and calculating the instantaneous reactive power of the node according to the three-phase voltage signal and the three-phase current signal of the node; A feature extraction unit (2) for determining the reactive power change rate of the node according to the instantaneous reactive power of the node, performing a fast Fourier transform on the three-phase current signal of the node to obtain the harmonic amplitude of the node, and determining the current harmonic distortion rate of the node according to the harmonic amplitude of the node; A compensation decision unit (3) for constructing a reactive power compensation decision agent according to the instantaneous reactive power, reactive power change rate and current harmonic distortion rate of each node in the distribution network, and solving the reactive power compensation decision agent according to a double deep Q network to obtain the compensation decision of each compensation branch in the distribution network; A compensation delivery unit (4) for constructing a reactive power compensation delivery model according to the compensation capacity of the reactive power compensation device, the compensation decision of each compensation branch in the distribution network and the power loss of each compensation branch in the distribution network, and solving the reactive power compensation delivery model to obtain the delivery action of each compensation branch in the distribution network; A reactive power compensation unit (5) for controlling the reactive power compensation device to perform reactive power compensation on each compensation branch in the distribution network according to the delivery action of each compensation branch in the distribution network.

2. A reactive power compensation system for solving the power factor electricity cost of a pumping station according to claim 1, characterized in that, The calculation formula for the instantaneous reactive power of the node is as follows: Among them, q t represents the instantaneous reactive power of the node at time t, v a (t), v b (t) and v c (t) represent the phase a, b, and c voltage signals of the node at time t, i a (t), i b (t) and i c (t) represent the phase a, b, and c current signals of the node at time t.

3. A reactive power compensation system for solving the power factor electricity cost of a pumping station according to claim 2, characterized in that, The calculation formula for the reactive power change rate of the node is as follows: Among them, rv t represents the reactive power change rate of the node at time t; The calculation formula for the harmonic amplitude of the node is as follows: Among them, represents the harmonic amplitude of the node at time t, n represents the current sampling number, N represents the total number of sampling points, and i abc represents the three-phase current signal, and h represents the order of the harmonic; The calculation formula for the current harmonic distortion rate of the node is as follows: Among them, THD t represents the current harmonic distortion rate of the node at time t, H represents the total order of harmonics, represents the three-phase current amplitude of the fundamental wave.

4. A reactive power compensation system for solving the power factor electricity cost of a pumping station according to claim 3, characterized in that, Constructing a reactive power compensation decision agent according to the instantaneous reactive power, reactive power change rate and current harmonic distortion rate of each node in the distribution network includes: Constructing a state space according to the instantaneous reactive power, reactive power change rate, current harmonic distortion rate and power factor of each node in the distribution network, where the expression of the state space is as follows: s t = [q t , rv t , THD t , cosφ]; where s t represents the state space, and cosφ represents the power factor; Constructing an action space, where the expression of the action space is as follows: a t ＝[(0) n ,(1) n ]； Among them, a t represents the action space, (0) n indicates that the nth compensation branch does not require reactive power compensation, a t indicates that the nth compensation branch requires reactive power compensation; Constructing a reward function according to the current harmonic distortion rate and power factor of each node in the distribution network, where the expression of the reward function is as follows: r t = 10(0.95 - |cosφ - 0.95|) - 0.1∑a t - 5THD t ; Among them, r t represents the reward function.

5. A reactive power compensation system for solving the power factor electricity cost of a regulating pump station according to claim 4, characterized in that, Solving the reactive power compensation decision agent according to a double deep Q network to obtain the compensation decision of each compensation branch in the distribution network includes: Initializing the parameters of the first deep Q network and the second deep Q network, and setting the discount factor and the learning rate; At each time step, according to the current state, select an action and execute the action, observe the next state and the reward, and store this pair of compensation decisions in the experience replay pool, where the expression of the compensation decision is as follows: u t = (s t , a t , r t , s t+1 ); Among them, u t represents the compensation decision, and s t+1 represents the next state; Randomly sample a batch of data from the experience replay pool, and update the parameters of the first deep Q network and the second deep Q network according to the loss function, where the expression of the loss function is as follows: L(θ) = (r t + γ maxá Q2(s t+1 , á) - Q1(s t , a t )) 2 ; Among them, \(L(\theta)\) represents the loss function, and \(Q_1\) and \(Q_2\) represent the first deep Q-network and the second deep Q-network.

6. A reactive power compensation system for solving power factor electricity charges in a regulated water pump station according to claim 5, characterized in that, Construct a reactive power compensation investment model according to the compensation capacity of the reactive power compensation device, the compensation decisions of each compensation branch in the distribution network, and the power losses of each compensation branch in the distribution network, including: Construct an objective function according to the compensation decisions of each compensation branch in the distribution network and the power losses of each compensation branch in the distribution network; Determine the constraint conditions according to the compensation capacity of the reactive power compensation device; Construct a reactive power compensation investment model according to the objective function and the constraint conditions.

7. A reactive power compensation system for solving the power factor electricity cost of a regulating pump station according to claim 6, characterized in that, The expression of the objective function is as follows: Among them, L1 represents the objective function, θ and κ represent the preset weights, and q req and q comp According to the compensation decision, determine the reactive power demand and the actually provided reactive power of the compensation branch that requires reactive power compensation, q nom represents the compensation capacity of the reactive power compensation device, P loss and P base represent the power loss and the reference power of the compensation branch that requires reactive power compensation determined according to the compensation decision.

8. A reactive power compensation system for solving the power factor electricity cost of a regulating pump station according to claim 7, characterized in that The expression of the constraint conditions is as follows: Among them, q STATCOM represents the capacity allocated by the static synchronous compensator, q SVG represents the capacity allocated by the static var compensator, q FC represents the capacity allocated by the reactive power compensation capacitor, q STATCPM_MAX represents the maximum capacity of the static synchronous compensator, q SVG_MAX represents the maximum capacity of the static var compensator, q step represents the capacity step of the reactive power compensation capacitor, sat(q req -q STATCOM ,0,q SVG_MAX ) represents restricting q req -q STATCOM between 0 and q SVG_MAX , and round represents the rounding function.

9. A reactive power compensation system for solving the power factor electricity cost of a regulating pump station according to claim 8, characterized in that, Solve the reactive power compensation investment model to obtain the investment actions of each compensation branch in the distribution network, including: Randomly initialize the individual population, and the position of each individual represents the combination of each decision variable in the reactive power compensation investment model; Calculate the output of the objective function according to the combination of each decision variable represented by each individual, and determine the fitness of the individual according to the output of the objective function; Randomly generate a position update coefficient, which is a constant with a range of 0 to 1. Judge whether the current iteration number is greater than the preset iteration number threshold, and judge whether the position update coefficient is greater than 0.

5. If the current iteration number is greater than the preset iteration number and the position update coefficient is greater than 0.5, update the position of the individual through the first position formula. If the current iteration number is greater than the preset iteration number and the position update coefficient is not greater than 0.5, update the position of the individual through the second position formula; Obtain the optimal individual based on the updated positions of all individuals. If the iteration number reaches the maximum value, the iteration ends and the optimal combination of each decision variable in the reactive power compensation investment model is output. Otherwise, the iteration number is incremented by 1.

10. A reactive power compensation system for solving the power factor electricity cost of a pumping station according to claim 9, characterized in that, The expression of the first position formula is as follows: Among them, X1(t + 1) represents the first position of an individual at the (t + 1)-th iteration, X best (t) represents the position of the current optimal individual, t represents the current iteration number, T represents the total number of iterations, X M (t) represents the average position of the current individual at the t-th iteration, S rand represents the position update coefficient; The expression of the second position formula is as follows: X2(t + 1) = X best (t) * L(D) + X R (t) - S rand ; Among them, X2(t + 1) represents the second position of the individual at the (t + 1)-th iteration, L(D) represents the Levy flight distribution function, and X R (t) represents the worst position of the current individual at the t-th iteration.

Images