CN107609690B - A method of load active management decision optimization - Google Patents

Abstract

The invention discloses a method for load active management decision optimization, comprising the following steps: step 1, establishing a model according to daily load historical data to carry out load forecasting; step 2, obtaining N households in the next twenty-four through load forecasting. The power set at m moments within an hour, and the flexible power adjustable range at each moment is obtained by calculating the flexible power; step 3, on the premise of keeping the total power of each household unchanged in the next 24 hours, within the adjustable range Internally adjust the power of each of the N households at m times in the next 24 hours, so that the difference between the peak value and the valley value of the total power of the N households at m times in the next 24 hours is the smallest; The predicted electricity price at m moments in four hours makes the electricity cost of each household the least in the next 24 hours.

Description

A method of load active management decision optimization

technical field

The invention relates to the field of active power distribution network, in particular to a method for load active management decision optimization.

Background technique

With the development of society and economy, the demand for power load is also increasing continuously. At the same time, traditional energy sources, mainly coal-fired thermal power, are facing increasingly serious problems of resource depletion and environmental pollution. In this context, renewable clean energy represented by wind power and solar energy has received extensive attention and rapid development, and is often connected to the power grid as a distributed energy source. Influence, such as changing the voltage level, increasing the short-circuit capacity, increasing the complexity of the relay protection, affecting the reliability of the power supply, etc.; in response to this situation, the active distribution network technology came into being. Active distribution network has the ability to control various distributed energy sources, controllable loads, energy storage, demand side management, etc. Power quality and power supply reliability are a development model of future smart grids.

Through active distribution network management technology, it is possible to flexibly access large-scale distributed energy sources, flexibly arrange operation modes, and flexibly arrange load power consumption; interactivity is one of the important characteristics of active distribution networks, and load-side demand response has flexible scheduling methods , it has the characteristics of huge potential to participate in the peak regulation of the system, so it has very significant practical significance to realize the two-way interaction between the load and the power grid. Traditional distribution network load control has been difficult to promote due to the limitation of equipment and measurement data; in recent years, with the rapid development of new sensor technology and communication technology, the demand-side advanced measurement system technology applied in active distribution network has been applied to direct loads. The development of control technology provides more possibilities.

SUMMARY OF THE INVENTION

The invention discloses a method for active load management decision optimization, which can actively manage and adjust various types of loads, and can realize the optimization of loads or electricity charges.

The present invention is achieved through the following technical solutions:

A method for active load management decision optimization, comprising the following steps:

Step 1. Build a model according to the daily load historical data for load forecasting;

Step 2: Obtain the power sets of each of the N households at m moments in the next 24 hours through load forecasting, and obtain the flexible power adjustable range at each moment through flexible power calculation;

Step 3. On the premise of keeping the total power of each household unchanged in the next 24 hours, adjust the power of each of the N households at m times in the next 24 hours within the adjustable range, so that the N households are The difference between the peak value and the valley value of the total power at m moments in the next 24 hours is the smallest;

Or on the premise of keeping the total power of each household unchanged in the next 24 hours, combined with the predicted electricity price at m moments in the next 24 hours, adjust the N households in the next 24 hours within the adjustable range. The power at m times within the next 24 hours will minimize the electricity consumption of each household in the next 24 hours.

表示日负荷历史数据中第x天的m个时刻的功率集合，则Dx=[Dx₁，Dx₂，……Dx_t，……Dx_m]，t=1，2……m，以P表示预测得到的未来二十四小时内m个时刻的功率集合，以a₁~a_n分别对应表示当天的拟合系数，得到模型：A further scheme of the present invention is that step 1 starts with

Represents the power set of m moments on the xth day in the daily load historical data, then Dx=[Dx ₁ , Dx ₂ ,...Dx _t ,...Dx _m ], t=1, 2...m, represented by P The predicted power sets of _m moments in the next 24 hours are represented by a ₁ ~an corresponding to the fitting coefficients of the current day, and the model is obtained:

P=a ₁ D1+a ₂ D2+…+a _x Dx+…+a _n Dn,

x=1,2,...,n,

s. t. a ₁ +a ₂ +...+a _x +...+a _n =1;

_Denote the set of a ₁ ~an values as α, and find the optimal α.

A further solution of the present invention is that the enumeration method is used to solve the optimal α, the value range of a ₁ ~an is ₀ to 1, the minimum interval is 0.05, and the judgment criterion of the optimal α is the first several moments of P How close the predicted value is to the actual value.

A further solution of the present invention is that in step 2, P _i represents the power set at m moments in the next 24 hours of the ith household, and P _i(t) represents the t moment in the next 24 hours of the ith household The power of , t=1, 2...m, get:

P _i = [ P _{i (1)} , P _{i (2)} , ... P _{i (t)} , ... P _{i (m)} ],

i=1, 2, 3,...N;

表示第i户家庭在未来二十四小时内第t时刻的柔性功率可调节最大值，得到：The adjustable range of the flexible power at each moment is obtained by calculating the flexible power, and P _i(t)min represents the minimum value of the adjustable power of the i-th household at the t-th moment in the next 24 hours.

Represents the maximum adjustable flexible power of the i-th household at the t-th time in the next 24 hours, and obtains:

≤

，P _{i (t) min} ≤

≤

,

i=1, 2, 3,...N;

*表示第i户家庭在未来二十四小时内第t时刻调节后的功率，by

*represents the adjusted power of the i-th household at the t-th time in the next 24 hours,

*=∑

。st ∑

*=∑

.

*表示第i户家庭在未来二十四小时调节后的m个时刻功率集合，以

_total*表示N户家庭在未来二十四小时调节后的m个时刻总功率集合，

_total*=Σ

*；以（

）_max和

_min分别表示N户家庭在未来二十四小时调节后的m个时刻总功率集合中的最大功率值和最小功率值，求解总功率峰谷差最小化目标函数：min（（

）_max-

_min）。The further scheme of the present invention is, step 3 is with

*represents the power set of m moments adjusted by the i-th household in the next 24 hours, with

_total * represents the total power set of m moments adjusted by N households in the next 24 hours,

_total *=Σ

*;by(

) _max and

_min respectively represents the maximum power value and the minimum power value in the total power set at m moments adjusted by N households in the next 24 hours. To solve the objective function of minimizing the peak-to-valley difference of the total power: min((

) _max -

_min ).

A further scheme of the present invention is to solve the total power peak-valley difference minimization objective function with particle swarm algorithm, including:

a. K represents the total number of particles, X _i represents the position of the ith particle, and X _ij represents the power set of the jth household in the position of the ith particle in the next 24 hours, and we get:

X _i =[X _i1 , X _i2 ,...,X _ij ,...,X _iN ],

i=1, 2, 3,...K,

j=1, 2, 3, ... N;

Taking P _ij(t) to represent the power of the j-th household at the ith particle position at the t-th time in the next twenty-four hours, we get:

X _ij =[ P _{ij (1)} , P _{ij (2)} , ... P _{ij (t)} , ... P _{ij (m)} ],

t=1, 2, 3, ... m;

b. Randomly adjust the X _ij obtained through load forecasting within the flexible power adjustable range to obtain the adjusted power set X _ij *, and sum X _ij and X _ij * respectively to obtain the respective total power X _{i (total)} =ΣX _ij and Xi _(total) *=ΣX _ij *, if Xi _(total) * is less than the total power of Xi _(total) , then randomly extract the power at a certain moment in Xi _ij * to adjust upward, if adjusted to If the maximum value of the flexible power adjustable range cannot balance Xi _(total) and Xi _(total) *, then randomly select other times except this moment for adjustment until Xi _(total) and Xi ₍ total) are satisfied. ₎ * the same constraints; if X _i(total) * is greater than X _{i (total)} , the power at a certain moment in X _ij * is randomly selected and adjusted downward, if it is adjusted to the minimum value of the flexible power adjustable range Balance Xi _(total) and Xi _(total) *, then randomly select other times except this moment for adjustment until the same constraints of Xi _(total) and Xi _(total) * are satisfied;

c. Denote the velocity corresponding to the i-th particle by V _i , and denote the component velocity of the i-th particle in the j-th dimension by V _ij , we get:

V _i =[V _i1 , V _i2 ,...,V _ij ,...,V _iN ],

i=1, 2, 3,...K,

j=1, 2, 3, ... N;

Taking V _ij(t) to represent the velocity of the i-th particle at the t-th moment of the velocity in the j-th dimension, we get:

V _ij =[V _i1(t) , V _i2(t) ,...,V _ij(t) ,...,V _iN(t) ],

t=1, 2, 3, ... m;

Initialize the particle velocity, and initialize the sub-velocity in N dimensions and the velocity at m times corresponding to the sub-velocity to 0;

d. Substitute the position X _i of the ith particle into the objective function to obtain the fitness value Fit(i) of the ith particle, which is expressed as:

_min；Fit(i)=(X _i(total) *) _max -

_min ;

The smaller the fitness value Fit(i) is, the better the position of the ith particle is, so that the historical optimal position X _pi searched by the particle itself and the optimal position X _pg searched by the entire particle swarm are calculated;

e. The update formula of particle velocity and position after the k+1th iteration is expressed as:

V _i ^k+1 =ω · V _i ^k + c ₁ ε · (X _pi ^k -X _i ^k ) + c ₂ η · (X _pg ^k -X _i ^k ),

X _i ^k+1 = X _i ^k +r·V _i ^k+1 ;

where ω is the inertia weight, c _{1 and} c ₂ are the coefficients of the historical optimal value searched by the particle tracking itself and the optimal value searched by all the particles; ε and η are values between 0 and 1. Random number, r is the position update coefficient;

f. Carry out the load optimization calculation with the smallest difference between the peak value and the valley value using the standard particle swarm algorithm, and obtain the power at m times in the next 24 hours for the load optimization.

A further solution of the present invention is that in step 3, price represents the set of electricity prices at m moments in the next 24 hours, and p _(t) represents the electricity price at the t-th moment in the next 24 hours, to obtain:

price=[P ₍₁₎ , P ₍₂₎ ,...P _(t) ,...P _(m) ],

t=1, 2, 3, ... m;

*·price）。Solve the objective function of minimization of electricity cost: min (

* price).

A further solution of the present invention is that solving the objective function of minimizing electricity charges includes:

a. Arrange the power P _i(t) of the ith household at time t in the next 24 hours in ascending order of electricity price at time t to form a new set P _in =[P _in1 , P _in2 ,...P _inz , ...P _inm ], z = 1, 2, ... m;

b. Adjust the power P _inz at each moment to the minimum value of the flexible power adjustable range, the adjusted power set P _in *=[ P _in1(min) , P _in2(min) ,...P _inz(min) ,... …P _inm(min) ], and then adjust the power from P _in1 to the maximum value of the flexible power adjustable range P _in1(max) until the adjusted power P _{in z} * is less than P _inz(max) ,To meet the conditions:

P _in1 +P _in2 +...+P _inm= P _in1(max) +P _in2(max) +...P _{in z} *+...+P _inm(min) ;

c. Rearrange Pin1(max), Pin2( _{max )} ,... _Pinz *,..., _Pinm(min) in chronological order to obtain the power at m moments in the next twenty-four hours for electricity tariff optimization.

A further scheme of the present invention further includes step 4: obtaining the total power Q of the i-th household in the next twenty-four hours through load prediction, and then calculating the carbon emission E according to the total power Q.

A further solution of the present invention is that the calculation formula of the carbon emission amount E is:

E=Q·(η _re · e _re +η _fossil · e _fossil );

表示用户的总功率Q中新能源所占比例，

表示化石能源所占比例，

是新能源的单位碳排放量，

是化石能源单位碳排放量；新能源的单位碳排放量

由总的新能源中各种新能源所占用电量百分比决定：in

Represents the proportion of new energy in the total power Q of the user,

represents the proportion of fossil energy,

is the unit carbon emission of new energy,

is the unit carbon emission of fossil energy; the unit carbon emission of new energy

It is determined by the percentage of electricity occupied by various new energy sources in the total new energy sources:

=Σe_i·a_i；i=1，2，……，M；

=Σe _i ·a _i ; i=1, 2, ..., M;

Among them, a _i represents the percentage of the electricity of the ith type of new energy in the total electricity of the new energy sources, and e _i represents the unit carbon emission of the ith type of new energy.

The advantages of the present invention compared with the prior art are:

1. Carry out load optimization in terms of regional total load peak shaving and valley filling and electricity economy of each household; first, the model is abstracted for load prediction, and the enumeration method is used to realize it, and then the load prediction based on the minimum peak-valley load difference is carried out. The relevant mathematical model is established for the load optimization of the power consumption, and the particle swarm algorithm is used to solve the optimization problem of the nonlinear objective function. have strong adaptability;

2. Defined and quantitatively analyzed carbon emissions, which is of great significance to the development of green power grids;

3. It is widely used and has significant social and economic benefits.

Description of drawings

FIG. 1 is a flow chart of the present invention.

FIG. 2 is a flowchart of step 1 of the present invention.

FIG. 3 is a graph obtained by fitting a user's power set in the next twenty-four hours through load prediction in the embodiment.

FIG. 4 is a flow chart showing the minimization of the total power peak-to-valley difference.

FIG. 5 is a schematic diagram of a flexible power adjustable range curve of a curve fitted by a user's power set in the next twenty-four hours in an embodiment.

FIG. 6 is a flowchart of solving the objective function of minimizing the total power peak-to-valley difference by the particle swarm algorithm.

FIG. 7 is a comparison diagram of total power curves of all users in the next 24 hours before and after load optimization in the embodiment.

FIG. 8 is a graph obtained by fitting a set of electricity prices at m times in the next twenty-four hours in the embodiment.

FIG. 9 is a comparison diagram before and after the minimization of electricity costs in the embodiment.

FIG. 10 is a graph obtained by fitting the power sets before and after the electricity cost is minimized in the embodiment.

Detailed ways

As shown in Figure 1, a load active management decision optimization method includes the following steps:

Step 1. As shown in Figure 2, build a model according to the daily load historical data for load forecasting;

表示日负荷历史数据中第x天的24个时刻的功率集合，则Dx=[Dx₁，Dx₂，……Dx_t，……Dx₂₄]，t=1，2……24，以P表示预测得到的未来二十四小时内24个时刻的功率集合，以a₁~a_n分别对应表示当天的拟合系数，得到模型：by

Represents the power set of 24 moments on the xth day in the daily load historical data, then Dx=[Dx ₁ , Dx ₂ ,...Dx _t ,...Dx ₂₄ ], t=1, 2...24, represented by P The predicted power set of 24 moments in the next 24 hours _is represented by a ₁ ~an corresponding to the fitting coefficient of the day, and the model is obtained:

P=a ₁ D1+a ₂ D2+…+a _x Dx+…+a _n Dn,

x=1,2,...,n,

s. t. a ₁ +a ₂ +...+a _x +...+a _n =1;

Denote the set of a ₁ ~an values as α, and use the enumeration method to find the optimal α. The value range of a ₁ ~an _{is 0} to 1, the minimum interval is 0.05, and the criterion for the optimal α is P The closeness between the predicted value and the actual value at the first two moments of , obtain the optimal α solution and then substitute it into the model, thus predicting the power set for the next 24 hours as shown in Figure 3.

Step 2 As shown in Figure 4, through load prediction, the power sets of 10 households at 24 moments in the next 24 hours are obtained respectively, and P _i represents the ith household's power at 24 moments in the next 24 hours. Power set, with P _i(t) representing the power of the i-th household at the t-th time in the next 24 hours, t=1, 2...24, we get:

P _i = [ P _{i (1)} , P _{i (2)} , ... P _{i (t)} , ... P _{i (24)} ],

i=1, 2, 3,...N;

表示第i户家庭在未来二十四小时内第t时刻的柔性功率可调节最大值，得到：The smart user terminal will obtain the adjustable range of flexible power at each moment through flexible power calculation as shown in Figure 5, and P _{i (t) min} represents the flexible power adjustable range of the i-th household at the t-th moment in the next 24 hours. minimum, with

Represents the maximum adjustable flexible power of the i-th household at the t-th time in the next 24 hours, and obtains:

≤

，P _{i (t) min} ≤

≤

,

i=1, 2, 3,...N;

*表示第i户家庭在未来二十四小时内第t时刻调节后的功率，Step 3. The power at each moment on the power curve of each household in the next 24 hours can be adjusted up and down within the flexible power adjustable range. On the premise that the total power remains unchanged within 14 hours, adjust the power of each of the 10 households at 24 moments in the next 24 hours within the adjustable range, namely:

*represents the adjusted power of the i-th household at the t-th time in the next 24 hours,

*=∑

。st ∑

*=∑

.

*表示第i户家庭在未来二十四小时调节后的24个时刻功率集合，以

_total*表示10户家庭在未来二十四小时调节后的24个时刻总功率集合，

_total*=Σ

*；以（

）_max和

_min分别表示10户家庭在未来二十四小时调节后的24个时刻总功率集合中的最大功率值和最小功率值，以粒子群算法求解总功率峰谷差最小化目标函数：min（（

）_max-

_min），如图6所示，包括：Minimize the difference between the peak value and the valley value of the total power of 10 households at 24 moments in the next 24 hours, namely:

*Indicates the power collection of the ith household at 24 moments adjusted in the next 24 hours, with

_total * represents the total power set of 10 households at 24 moments adjusted in the next 24 hours,

_total *=Σ

*;by(

) _max and

_min represents the maximum power value and the minimum power value of the total power set at 24 moments adjusted by 10 households in the next 24 hours respectively. The particle swarm algorithm is used to solve the objective function of minimizing the peak-to-valley difference of the total power: min ((

) _max -

_min ), as shown in Figure 6, including:

a. K represents the total number of particles, X _i represents the position of the ith particle, and X _ij represents the power set of the jth household in the position of the ith particle in the next 24 hours, and we get:

X _i =[X _i1 , X _i2 ,...,X _ij ,...,X _iN ],

i=1, 2, 3,...K,

j=1, 2, 3, ... N;

Taking P _ij(t) to represent the power of the j-th household at the ith particle position at the t-th time in the next twenty-four hours, we get:

X _ij = [ P _{ij (1)} , P _{ij (2)} , ... P _{ij (t)} , ... P _{ij (24)} ],

t=1, 2, 3, ... 24;

b. Randomly adjust the X _ij obtained through load forecasting within the flexible power adjustable range to obtain the adjusted power set X _ij *, and sum X _ij and X _ij * respectively to obtain the respective total power X _{i (total)} =ΣX _ij and Xi _(total) *=ΣX _ij *, if Xi _(total) * is less than the total power of Xi _(total) , then randomly extract the power at a certain moment in Xi _ij * to adjust upward, if adjusted to If the maximum value of the flexible power adjustable range cannot balance Xi _(total) and Xi _(total) *, then randomly select other times except this moment for adjustment until Xi _(total) and Xi ₍ total) are satisfied. ₎ * the same constraints; if X _i(total) * is greater than X _{i (total)} , the power at a certain moment in X _ij * is randomly selected and adjusted downward, if it is adjusted to the minimum value of the flexible power adjustable range Balance Xi _(total) and Xi _(total) *, then randomly select other times except this moment for adjustment until the same constraints of Xi _(total) and Xi _(total) * are satisfied;

c. Denote the velocity corresponding to the i-th particle by V _i , and denote the component velocity of the i-th particle in the j-th dimension by V _ij , we get:

V _i =[V _i1 , V _i2 ,...,V _ij ,...,V _iN ],

i=1, 2, 3,...K,

j=1, 2, 3, ... N;

Taking V _ij(t) to represent the velocity of the i-th particle at the t-th moment of the velocity in the j-th dimension, we get:

V _ij =[V _i1(t) , V _i2(t) ,...,V _ij(t) ,...,V _iN(t) ],

t=1, 2, 3, ... 24;

Initialize the particle velocity, and initialize the sub-velocity in N dimensions and the velocity at 24 moments corresponding to the sub-velocity to 0;

d. Substitute the position X _i of the ith particle into the objective function to obtain the fitness value Fit(i) of the ith particle, which is expressed as:

_min；Fit(i)=(X _i(total) *) _max -

_min ;

The smaller the fitness value Fit(i) is, the better the position of the ith particle is, so that the historical optimal position X _pi searched by the particle itself and the optimal position X _pg searched by the entire particle swarm are calculated;

e. The update formula of particle velocity and position after the k+1th iteration is expressed as:

V _i ^k+1 =ω · V _i ^k + c ₁ ε · (X _pi ^k -X _i ^k ) + c ₂ η · (X _pg ^k -X _i ^k ),

X _i ^k+1 = X _i ^k +r·V _i ^k+1 ;

where ω is the inertia weight, c _{1 and} c ₂ are the coefficients of the historical optimal value searched by the particle tracking itself and the optimal value searched by all the particles; ε and η are values between 0 and 1. Random number, r is the position update coefficient;

f. Use the standard particle swarm algorithm to carry out the load optimization calculation with the smallest difference between the peak value and the valley value, and obtain the load optimization power at 24 times in the next 24 hours as shown in Figure 7, and optimize the peak value of the total load of the top 10 households The valley difference is 14.26kW, and the peak-to-valley difference after optimization is 8.63kW, a reduction of 39.5%.

The main station of the distribution network will calculate the electricity price at 24 times in the next 24 hours through the electricity price forecast, and issue it to the upper-level system for active load management. On the premise of keeping the total power of each household unchanged in the next 24 hours, combined with the predicted electricity price at 24 moments in the next 24 hours, adjust the 10 households within the next 24 hours within the adjustable range. The power of 24 moments makes the electricity cost of each household in the next 24 hours to be the least, that is: as shown in Figure 8, price represents the set of electricity prices at 24 moments in the next 24 hours, and p _{(t )} represents the electricity price at time t in the next twenty-four hours, and we get:

price=[P ₍₁₎ , P ₍₂₎ ,...P _(t) ,...P ₍₂₄₎ ],

t=1, 2, 3, ... 24;

*·price），包括：Solve the objective function of minimization of electricity cost: min (

* price), including:

a. Arrange the power P _i(t) of the ith household at time t in the next 24 hours in ascending order of electricity price at time t to form a new set P _in =[P _in1 , P _in2 ,...P _inz , ...P _in24 ], z = 1, 2, ... 24;

b. Adjust the power P _inz at each moment to the minimum value of the flexible power adjustable range, the adjusted power set P _in *=[ P _in1(min) , P _in2(min) ,...P _inz(min) ,... ...P _in24(min) ], and then adjust the power from P _in1 to the maximum value of the flexible power adjustable range P _in1(max) , until the adjusted power P _{in z} * is less than P _{inz (max)} , as shown in Figure 9, the conditions are met:

P _in1 +P _in2 +...+P _in24= P _in1(max) +P _in2(max) +...P _{in z} *+...+P _in24(min) ;

c. Rearrange Pin1(max), Pin2( _{max )} ,... _Pinz *,..., _Pin24(min) in chronological order to get the electricity tariff optimized next twenty-four hours 24 as shown in Figure 10 For the power at this moment, the electricity consumption of user A in the next 24 hours before optimization is 20.38 yuan, and the electricity consumption after optimization is 18.48 yuan, a decrease of 9.3%.

Step 4. Obtain the total power Q of the i-th household in the next 24 hours through load prediction, and then calculate the carbon emission E according to the total power Q. The calculation formula is:

E=Q·(η _re · e _re +η _fossil · e _fossil );

表示用户的总功率Q中新能源所占比例，

表示化石能源所占比例，

是新能源的单位碳排放量，

是化石能源单位碳排放量；新能源的单位碳排放量

由总的新能源中各种新能源所占用电量百分比决定：in

Represents the proportion of new energy in the total power Q of the user,

represents the proportion of fossil energy,

is the unit carbon emission of new energy,

is the unit carbon emission of fossil energy; the unit carbon emission of new energy

It is determined by the percentage of electricity occupied by various new energy sources in the total new energy sources:

=Σe_i·a_i；i=1，2，……，24；

=Σe _i ·a _i ; i=1, 2, ..., 24;

Among them, a _i represents the percentage of the electricity of the ith in the 24 new energy sources in the total electricity of new energy, and e _i represents the unit carbon emission of the ith new energy.

Claims (6)

1. a method for active load management decision optimization, is characterized in that comprising the steps: Step 1. Build a model according to the daily load historical data for load forecasting; Step 2: Obtain the power sets of each of the N households at m moments in the next 24 hours through load forecasting, and obtain the flexible power adjustable range at each moment through flexible power calculation; Step 3. On the premise of keeping the total power of each household unchanged in the next 24 hours, adjust the power of each of the N households at m times in the next 24 hours within the adjustable range, so that the N households are The difference between the peak value and the valley value of the total power at m moments in the next 24 hours is the smallest; Or on the premise of keeping the total power of each household unchanged in the next 24 hours, combined with the predicted electricity price at m moments in the next 24 hours, adjust the N households in the next 24 hours within the adjustable range. The power at m times within the range, so that the electricity consumption of each household in the next 24 hours is the least; Step 3: P _i * represents the power set of m moments adjusted by the ith household in the next 24 hours, and P _total * represents the total power set of m moments adjusted by N households in the next 24 hours, P _total *=ΣP _i *; with (P _total *) _max and (P _total *) _min respectively represent the maximum power value and the minimum power in the total power set of N households adjusted at m moments in the next 24 hours value, solve the objective function of minimizing the total power peak-to-valley difference: min((P _total *) _max -(P _total *) _min ); The objective function of minimizing the total power peak-valley difference is solved by particle swarm algorithm, including: a. K represents the total number of particles, X _i represents the position of the ith particle, and X _ij represents the power set of the jth household in the position of the ith particle in the next 24 hours, and we get: X _i =[X _i1 , X _i2 ,...,X _ij ,...,X _iN ], i=1, 2, 3,...K, j=1, 2, 3, ... N; Taking P _ij(t) to represent the power of the j-th household at the ith particle position at the t-th time in the next twenty-four hours, we get: X _ij =[ P _{ij (1)} , P _{ij (2)} , ... P _{ij (t)} , ... P _{ij (m)} ], t=1, 2, 3, ... m; b. Randomly adjust the X _ij obtained through load forecasting within the flexible power adjustable range to obtain the adjusted power set X _ij *, and sum X _ij and X _ij * respectively to obtain the respective total power X _{i (total)} =ΣX _ij and Xi _(total) *=ΣX _ij *, if Xi _(total) * is less than the total power of Xi _(total) , then randomly extract the power at a certain moment in Xi _ij * to adjust upward, if adjusted to If the maximum value of the flexible power adjustable range cannot balance Xi _(total) and Xi _(total) *, then randomly select other times except this moment for adjustment until Xi _(total) and Xi ₍ total) are satisfied. ₎ * the same constraints; if X _i(total) * is greater than X _{i (total)} , the power at a certain moment in X _ij * is randomly selected and adjusted downward, if it is adjusted to the minimum value of the flexible power adjustable range Balance Xi _(total) and Xi _(total) *, then randomly select other times except this moment for adjustment until the same constraints of Xi _(total) and Xi _(total) * are satisfied; c. Denote the velocity corresponding to the i-th particle by V _i , and denote the component velocity of the i-th particle in the j-th dimension by V _ij , we get: V _i =[V _i1 , V _i2 ,...,V _ij ,...,V _iN ], i=1, 2, 3,...K, j=1, 2, 3, ... N; Taking V _ij(t) to represent the velocity of the i-th particle at the t-th moment of the velocity in the j-th dimension, we get: V _ij =[V _i1(t) , V _i2(t) ,...,V _ij(t) ,...,V _iN(t) ], t=1, 2, 3, ... m; Initialize the particle velocity, and initialize the sub-velocity in N dimensions and the velocity at m times corresponding to the sub-velocity to 0; d. Substitute the position X _i of the ith particle into the objective function to obtain the fitness value Fit(i) of the ith particle, which is expressed as: Fit(i) = (X _i(total) *) _max - (X _i(total) *) _min ; The smaller the fitness value Fit(i) is, the better the position of the ith particle is, so that the historical optimal position X _pi searched by the particle itself and the optimal position X _pg searched by the entire particle swarm are calculated; e. The update formula of particle velocity and position after the k+1th iteration is expressed as: V _i ^k+1 =ω · V _i ^k + c ₁ ε · (X _pi ^k -X _i ^k ) + c ₂ η · (X _pg ^k -X _i ^k ), X _i ^k+1 = X _i ^k +r·V _i ^k+1 ; where ω is the inertia weight, c _{1 and} c ₂ are the coefficients of the historical optimal value searched by the particle tracking itself and the optimal value searched by all the particles; ε and η are values between 0 and 1. Random number, r is the position update coefficient; f. Use the standard particle swarm algorithm to carry out the load optimization calculation with the smallest difference between the peak value and the valley value, and obtain the power of the load optimization at m moments in the next 24 hours; In step 3, price is used to represent the set of electricity prices at m times in the next 24 hours, and p _(t) is used to represent the electricity price at time t in the next 24 hours, and we get: price=[P ₍₁₎ , P ₍₂₎ ,...P _(t) ,...P _(m) ], t=1, 2, 3, ... m; Solve the objective function of minimizing electricity cost: min(P _i(t) * price); The objective function to solve the electricity cost minimization includes: a. Arrange the power P _i(t) of the ith household at time t in the next 24 hours in ascending order of electricity price at time t to form a new set P _in =[P _in1 , P _in2 ,...P _inz , ...P _inm ], z = 1, 2, ... m; b. Adjust the power P _inz at each moment to the minimum value of the flexible power adjustable range, the adjusted power set P _in *=[ P _in1(min) , P _in2(min) ,...P _inz(min) ,... …P _inm(min) ], and then adjust the power from P _in1 to the maximum value of the flexible power adjustable range P _in1(max) , until the adjusted power P _inz * is less than P _{inz (max)} , it satisfies condition: P _in1 +P _in2 +...+P _inm= P _in1(max) +P _in2(max) +...P _inz *+...+P _inm(min) ; c. Rearrange P _in1(max) , P _in2(max) ,...P _inz *,...,P _inm(min) in chronological order to obtain the power at m moments in the next twenty-four hours for electricity tariff optimization. 2. a kind of load active management decision optimization method as claimed in claim 1 is characterized in that: step 1 is with

Represents the power set of m moments on the xth day in the daily load historical data, then Dx=[Dx ₁ , Dx ₂ ,...Dx _t ,...Dx _m ], t=1, 2...m, represented by P The predicted power sets of _m moments in the next 24 hours are represented by a ₁ ~an corresponding to the fitting coefficients of the current day, and the model is obtained: P=a ₁ D1+a ₂ D2+…+a _x Dx+…+a _n Dn, x=1,2,...,n, s. t. a ₁ +a ₂ +...+a _x +...+a _n =1; _Denote the set of a ₁ ~an values as α, and find the optimal α. 3. a kind of load active management decision-making optimization method as claimed in claim 2 is characterized in that: adopt enumeration method to solve optimal α, the value range of a ₁ ~ a _n is 0 to 1, and the minimum interval is 0.05 , the criterion for the optimal α is the closeness of the predicted value of P at the first few moments to the actual value. 4. a kind of load active management decision-making optimization method as claimed in claim 1 is characterized in that: step 3 represents the power set of m moments in the next 24 hours of the i-th household with P _i , with P _{i ( t)} represents the power of the i-th household at the t-th time in the next 24 hours, t=1, 2...m, we get: P _i = [ P _{i (1)} , P _{i (2)} , ... P _{i (t)} , ... P _{i (m)} ], i=1, 2, 3,...N; The adjustable range of the flexible power at each moment is obtained by calculating the flexible power, and P _i(t)min represents the minimum value of the adjustable power of the i-th household at the t-th moment in the next 24 hours, and P _{i(t) max} represents the maximum adjustable flexible power of the i-th household at the t-th time in the next 24 hours, and we get: P _i(t)min ≤P _i(t) ≤P _i(t)max , i=1, 2, 3,...N; Denote the adjusted power of the i-th household at the t-th time in the next 24 hours by P _i(t) *, st ∑P _i(t) *=∑P _i(t) . 5. The method for active load management decision optimization according to claim 1, further comprising step 4, obtaining the total power Q of the i-th household in the next twenty-four hours through load prediction, and then according to the total power Q. Power Q calculates carbon emissions E. 6. The method for active load management decision optimization according to claim 5, characterized in that: the calculation formula of carbon emission E is: E=Q·(η _re · e _re +η _fossil · e _fossil ); Among them, η _re represents the proportion of new energy in the total power Q of the user, η _fossil represents the proportion of fossil energy, e _re is the unit carbon emission of new energy, and e _fossil is the unit carbon emission of fossil energy; The carbon emission e _re is determined by the percentage of electricity occupied by various new energy sources in the total new energy sources: e _re =Σe _i ·a _i ; i=1, 2, ..., M; Among them, a _i represents the percentage of the electricity of the ith type of new energy in the total electricity of the new energy sources, and e _i represents the unit carbon emission of the ith type of new energy.

Images